qnn/sm8750 4k: HuggingFaceTB/SmolLM2-135M-Instruct@12fd25f77366

README.md

@@ -11,6 +11,7 @@ tags:
 - mtk
 - on-device
 - openweights
+- qnn
 - vulkan
 - xnnpack
 ---
@@ -21,7 +22,7 @@ ExecuTorch exports of [HuggingFaceTB/SmolLM2-135M-Instruct](https://huggingface.
 
 ## Files
 
-Windows in this repo: XNNPACK (CPU) at 2k to 32k; Vulkan (GPU) at 2k to 32k; MediaTek NeuroPilot MT6991 (Dimensity 9400) at 2k to 16k. The window is fixed inside the file: the runtime allocates the whole KV cache at load, so pick the largest window the device can hold (`fits_phone_budget` in each folder's `config.json` is the estimate against a 5 GB budget).
+Windows in this repo: XNNPACK (CPU) at 2k to 32k; Vulkan (GPU) at 2k to 32k; Qualcomm QNN (HTP) SM8750 (Snapdragon 8 Elite) at 4k; MediaTek NeuroPilot MT6991 (Dimensity 9400) at 2k to 16k. The window is fixed inside the file: the runtime allocates the whole KV cache at load, so pick the largest window the device can hold (`fits_phone_budget` in each folder's `config.json` is the estimate against a 5 GB budget).
 
 | Backend | Target | File | Window | Size | Smoke test |
 |---|---|---|---|---|---|
@@ -35,6 +36,7 @@ Windows in this repo: XNNPACK (CPU) at 2k to 32k; Vulkan (GPU) at 2k to 32k; Med
 | Vulkan (GPU) | any arm64 | [`vulkan/SmolLM2-135M-Instruct-vulkan-8da4w-8k.pte`](vulkan/SmolLM2-135M-Instruct-vulkan-8da4w-8k.pte) | 8,192 tokens | 0.14 GB | structure checked (no host NPU runtime) |
 | Vulkan (GPU) | any arm64 | [`vulkan/SmolLM2-135M-Instruct-vulkan-8da4w-16k.pte`](vulkan/SmolLM2-135M-Instruct-vulkan-8da4w-16k.pte) | 16,384 tokens | 0.15 GB | structure checked (no host NPU runtime) |
 | Vulkan (GPU) | any arm64 | [`vulkan/SmolLM2-135M-Instruct-vulkan-8da4w-32k.pte`](vulkan/SmolLM2-135M-Instruct-vulkan-8da4w-32k.pte) | 32,768 tokens | 0.17 GB | structure checked (no host NPU runtime) |
+| Qualcomm QNN (HTP) | SM8750 (Snapdragon 8 Elite) | [`qnn/sm8750/SmolLM2-135M-Instruct-qnn-hybrid-4k.pte`](qnn/sm8750/SmolLM2-135M-Instruct-qnn-hybrid-4k.pte) | 4,096 tokens | 0.20 GB | structure checked (no host NPU runtime) |
 | MediaTek NeuroPilot | MT6991 (Dimensity 9400) | [`mtk/mt6991/SmolLM2-135M-Instruct-neuropilot-a16w8-2k-chunk1of3.pte`](mtk/mt6991/SmolLM2-135M-Instruct-neuropilot-a16w8-2k-chunk1of3.pte) | 2,048 tokens | 0.04 GB | structure checked (no host NPU runtime) |
 | MediaTek NeuroPilot | MT6991 (Dimensity 9400) | [`mtk/mt6991/SmolLM2-135M-Instruct-neuropilot-a16w8-2k-chunk2of3.pte`](mtk/mt6991/SmolLM2-135M-Instruct-neuropilot-a16w8-2k-chunk2of3.pte) | 2,048 tokens | 0.04 GB | structure checked (no host NPU runtime) |
 | MediaTek NeuroPilot | MT6991 (Dimensity 9400) | [`mtk/mt6991/SmolLM2-135M-Instruct-neuropilot-a16w8-2k-chunk3of3.pte`](mtk/mt6991/SmolLM2-135M-Instruct-neuropilot-a16w8-2k-chunk3of3.pte) | 2,048 tokens | 0.07 GB | structure checked (no host NPU runtime) |
@@ -69,6 +71,7 @@ Tokenizer: [`tokenizer.json`](tokenizer.json), copied unchanged from the source
 
 - XNNPACK (CPU) any arm64: ExecuTorch 1.4.0 `export_llm`: 8-bit dynamic activations and 4-bit weights in groups of 32, int8 per-channel embeddings, XNNPACK with extended ops, prefill chunk 2048, fp32 KV cache. Built by [run 1](https://github.com/ExperimentalMachines/execupack/actions/runs/37158331242).
 - Vulkan (GPU) any arm64: ExecuTorch 1.4.0 `export_llm`: 8-bit dynamic activations and 4-bit weights in groups of 32, int8 per-channel embeddings, the Vulkan delegate, prefill chunk 2048, fp32 KV cache. Built by [run 1](https://github.com/ExperimentalMachines/execupack/actions/runs/37157380448).
+- Qualcomm QNN (HTP) SM8750 (Snapdragon 8 Elite): ExecuTorch 1.5.1 Qualcomm static LLM (`examples/qualcomm/oss_scripts/llama`, `--decoder_model smollm2_135m`): the quantization recipe ExecuTorch registers for this model, calibrated on wikitext (1 sample), hybrid prefill (128 tokens per step) and decode graphs compiled with QAIRT 2.37.0.250724 for SM8750 (Snapdragon 8 Elite). Built by [run 1](https://github.com/ExperimentalMachines/execupack/actions/runs/37643682104).
 - MediaTek NeuroPilot MT6991 (Dimensity 9400): ExecuTorch 1.4.0 MediaTek LLM export (`examples/mediatek`, `llama.py`): A16W8 (16-bit activations, 8-bit weights) calibrated on MediaTek's `alpaca.txt` prompts (9 of 9) in the qwen.json chat template, cut into 3 chunks, with a 128-token prompt graph and a one-token generation graph over a 2048-token cache, compiled with MediaTek NeuroPilot Express SDK (mtk_converter 8.13.0+public, mtk_neuron 8.2.23) for MT6991 (Dimensity 9400). Built by [run 1](https://github.com/ExperimentalMachines/execupack/actions/runs/37158191680).
 - MediaTek NeuroPilot MT6991 (Dimensity 9400): ExecuTorch 1.4.0 MediaTek LLM export (`examples/mediatek`, `llama.py`): A16W8 (16-bit activations, 8-bit weights) calibrated on MediaTek's `alpaca.txt` prompts (9 of 9) in the qwen.json chat template, cut into 3 chunks, with a 128-token prompt graph and a one-token generation graph over a 4096-token cache, compiled with MediaTek NeuroPilot Express SDK (mtk_converter 8.13.0+public, mtk_neuron 8.2.23) for MT6991 (Dimensity 9400). Built by [run 1](https://github.com/ExperimentalMachines/execupack/actions/runs/37158191680).
 - MediaTek NeuroPilot MT6991 (Dimensity 9400): ExecuTorch 1.4.0 MediaTek LLM export (`examples/mediatek`, `llama.py`): A16W8 (16-bit activations, 8-bit weights) calibrated on MediaTek's `alpaca.txt` prompts (9 of 9) in the qwen.json chat template, cut into 3 chunks, with a 128-token prompt graph and a one-token generation graph over a 8192-token cache, compiled with MediaTek NeuroPilot Express SDK (mtk_converter 8.13.0+public, mtk_neuron 8.2.23) for MT6991 (Dimensity 9400). Built by [run 1](https://github.com/ExperimentalMachines/execupack/actions/runs/37158191680).
@@ -78,4 +81,6 @@ Tokenizer: [`tokenizer.json`](tokenizer.json), copied unchanged from the source
 
 A quantized derivative of [HuggingFaceTB/SmolLM2-135M-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM2-135M-Instruct), distributed under the same terms (apache-2.0).
 
+The `qnn/` folders hold QNN HTP context binaries compiled with the Qualcomm AI Runtime SDK (QAIRT) 2.37.0.250724 from Qualcomm Technologies, Inc., used under its AI Stack License. No Qualcomm SDK or runtime library is included; running them needs the matching QNN runtime (for example `executorch-android-qnn` 1.5.1, which depends on `qnn-runtime` 2.37.0).
+
 The `mtk/` folders hold model binaries compiled with the MediaTek NeuroPilot Express SDK 8.0.8-build20250925 (mtk_converter 8.13.0+public, mtk_neuron 8.2.23) from MediaTek Inc., used under MediaTek's license terms for that SDK. No MediaTek SDK or runtime library is included. They run on MediaTek's LLM runner from ExecuTorch (`examples/mediatek/executor_runner`) with the device's NeuroPilot runtime; the settings it needs are in each folder's `config.json`, under each variant's `runner`.

qnn/sm8750/SmolLM2-135M-Instruct-qnn-hybrid-4k.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83b6a2f4877064abed160a5b927bfa5c6be790f7efcd4bc2c15e14a210d19cfd
+size 203697792

qnn/sm8750/config.json

@@ -0,0 +1,28 @@
+{
+  "runtime": "executorch",
+  "runtime_version": "1.5.1",
+  "backend": "qnn",
+  "target": "sm8750",
+  "tokenizer": "tokenizer.json",
+  "source_model": "HuggingFaceTB/SmolLM2-135M-Instruct",
+  "source_revision": "12fd25f77366fa6b3b4b768ec3050bf629380bac",
+  "variants": [
+    {
+      "file": "SmolLM2-135M-Instruct-qnn-hybrid-4k.pte",
+      "size_bytes": 203697792,
+      "sha256": "83b6a2f4877064abed160a5b927bfa5c6be790f7efcd4bc2c15e14a210d19cfd",
+      "context": 4096,
+      "source_revision": "12fd25f77366fa6b3b4b768ec3050bf629380bac",
+      "quantization": "QNN HTP, ExecuTorch smollm2_135m recipe",
+      "fits_phone_budget": null,
+      "methods": {
+        "get_max_context_len": 4096,
+        "get_max_seq_len": 4096,
+        "get_bos_id": 1,
+        "get_vocab_size": 49152,
+        "get_n_layers": 30
+      }
+    }
+  ],
+  "qnn_sdk_version": "2.37.0.250724"
+}

qnn/sm8750/export-report-4k.json

@@ -0,0 +1,692 @@
+{
+  "backend": "qnn",
+  "target": "sm8750",
+  "target_name": "Snapdragon 8 Elite",
+  "tokenizer": "tokenizer.json",
+  "output_repo": "experimentalmachines/SmolLM2-135M-Instruct-ExecuTorch",
+  "source": {
+    "id": "HuggingFaceTB/SmolLM2-135M-Instruct",
+    "sha": "12fd25f77366fa6b3b4b768ec3050bf629380bac",
+    "created_at": "2024-10-31T13:41:10+00:00",
+    "total_params": 134515008,
+    "family": "llama",
+    "variant": "instruct",
+    "license": {
+      "license": "apache-2.0"
+    },
+    "license_files": []
+  },
+  "toolchain": {
+    "executorch": "1.5.1",
+    "torch": "2.14.0+cpu",
+    "torchao": "0.18.0",
+    "transformers": "5.0.0rc1",
+    "lm_eval": "0.4.5",
+    "qairt": "2.37.0.250724"
+  },
+  "recipe": {
+    "decoder_model": "smollm2_135m",
+    "params": "executorch: examples/models/smollm2/135M_config.json",
+    "model_mode": "hybrid",
+    "prefill_ar_len": 128,
+    "max_context_len": 4096,
+    "calibration": {
+      "tasks": [
+        "wikitext"
+      ],
+      "limit": 1
+    },
+    "label": "QNN HTP, ExecuTorch smollm2_135m recipe",
+    "description": "ExecuTorch 1.5.1 Qualcomm static LLM (`examples/qualcomm/oss_scripts/llama`, `--decoder_model smollm2_135m`): the quantization recipe ExecuTorch registers for this model, calibrated on wikitext (1 sample), hybrid prefill (128 tokens per step) and decode graphs compiled with QAIRT 2.37.0.250724 for SM8750 (Snapdragon 8 Elite)."
+  },
+  "window": {
+    "context": 4096,
+    "reason": "forced with --context (static NPU graphs)",
+    "kv_cache_bytes_per_token": null
+  },
+  "files": [
+    {
+      "path": "qnn/sm8750/SmolLM2-135M-Instruct-qnn-hybrid-4k.pte",
+      "bytes": 203697792,
+      "sha256": "83b6a2f4877064abed160a5b927bfa5c6be790f7efcd4bc2c15e14a210d19cfd"
+    }
+  ],
+  "host": {
+    "nproc": 32,
+    "mem_total_bytes": 729889779712,
+    "swap_total_bytes": 0,
+    "cpu_model": "unknown",
+    "peak_rss_export_bytes": 10922528768,
+    "peak_swap_used_bytes": 0,
+    "min_mem_available_bytes": 720223883264,
+    "peak_in_use_bytes": 9665896448,
+    "peak_in_use_at": "2026-10-07T15:57:19Z",
+    "peak_cgroup_bytes": 9665654784,
+    "export_seconds": 1707.5,
+    "total_seconds": 1733.6
+  },
+  "metadata": {
+    "get_ar_len": 1,
+    "get_bos_id": 1,
+    "get_dim": 576,
+    "get_eos_id": 2,
+    "get_head_dim": 64,
+    "get_kv_io_bit_width": 16,
+    "get_kv_output_0_quant_attr": [
+      1.5267929484252818e-05,
+      39862,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_10_quant_attr": [
+      0.00010673665383365005,
+      34317,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_11_quant_attr": [
+      0.0002642688632477075,
+      33578,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_12_quant_attr": [
+      0.00010371369717177004,
+      35940,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_13_quant_attr": [
+      0.00028053202549926937,
+      31988,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_14_quant_attr": [
+      0.00015581764455419034,
+      34378,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_15_quant_attr": [
+      0.0003261002420913428,
+      35515,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_16_quant_attr": [
+      0.00015474561951123178,
+      32272,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_17_quant_attr": [
+      0.0002924219297710806,
+      32578,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_18_quant_attr": [
+      0.00015396317758131772,
+      32918,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_19_quant_attr": [
+      0.0003059203736484051,
+      31890,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_1_quant_attr": [
+      0.00030670216074213386,
+      30269,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_20_quant_attr": [
+      0.00013694606604985893,
+      29973,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_21_quant_attr": [
+      0.0002858389634639025,
+      33457,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_22_quant_attr": [
+      0.0001531876332592219,
+      35649,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_23_quant_attr": [
+      0.0005066198064014316,
+      33836,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_24_quant_attr": [
+      0.00012952715042047203,
+      35486,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_25_quant_attr": [
+      0.0003104919451288879,
+      31998,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_26_quant_attr": [
+      0.00010727052722359076,
+      28316,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_27_quant_attr": [
+      0.00033561201416887343,
+      34965,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_28_quant_attr": [
+      0.00016533012967556715,
+      28573,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_29_quant_attr": [
+      0.0003243841347284615,
+      32269,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_2_quant_attr": [
+      9.455693361815065e-05,
+      34333,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_30_quant_attr": [
+      0.00015277363127097487,
+      29647,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_31_quant_attr": [
+      0.00031807637424208224,
+      33863,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_32_quant_attr": [
+      0.00013909072731621563,
+      34676,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_33_quant_attr": [
+      0.00027213257271796465,
+      33227,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_34_quant_attr": [
+      0.0001762398605933413,
+      34815,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_35_quant_attr": [
+      0.00037117855390533805,
+      34998,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_36_quant_attr": [
+      0.00027227666578255594,
+      28074,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_37_quant_attr": [
+      0.00045419292291626334,
+      33582,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_38_quant_attr": [
+      0.0002045194269157946,
+      34913,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_39_quant_attr": [
+      0.0003060080052819103,
+      33463,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_3_quant_attr": [
+      0.00030539327417500317,
+      32608,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_40_quant_attr": [
+      0.00022089778212830424,
+      29561,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_41_quant_attr": [
+      0.00029903819086030126,
+      33474,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_42_quant_attr": [
+      0.0001695959217613563,
+      34948,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_43_quant_attr": [
+      0.00036811965401284397,
+      32581,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_44_quant_attr": [
+      0.00020002422388643026,
+      31396,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_45_quant_attr": [
+      0.0003247488639317453,
+      31775,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_46_quant_attr": [
+      0.0003469733928795904,
+      32098,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_47_quant_attr": [
+      0.00034109599073417485,
+      33204,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_48_quant_attr": [
+      0.00040949569665826857,
+      32632,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_49_quant_attr": [
+      0.0003326442965772003,
+      34870,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_4_quant_attr": [
+      0.0001026399913826026,
+      31874,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_50_quant_attr": [
+      0.00028462670161388814,
+      27937,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_51_quant_attr": [
+      0.0002819089568220079,
+      34220,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_52_quant_attr": [
+      0.0003380293201189488,
+      33640,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_53_quant_attr": [
+      0.0002854775229934603,
+      32441,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_54_quant_attr": [
+      0.000266329909209162,
+      32551,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_55_quant_attr": [
+      0.00026868030545301735,
+      32587,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_56_quant_attr": [
+      0.000399931421270594,
+      30353,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_57_quant_attr": [
+      0.000237992440816015,
+      34703,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_58_quant_attr": [
+      0.0004007575334981084,
+      32312,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_59_quant_attr": [
+      0.0002558397827669978,
+      31986,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_5_quant_attr": [
+      0.00031697796657681465,
+      33672,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_6_quant_attr": [
+      0.0001312212843913585,
+      29320,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_7_quant_attr": [
+      0.0003144720394629985,
+      32406,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_8_quant_attr": [
+      0.00012001510913250968,
+      37142,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_kv_output_9_quant_attr": [
+      0.00027390732429921627,
+      33054,
+      0,
+      65535,
+      "torch.int32"
+    ],
+    "get_logits_scale": 0.001260443706996739,
+    "get_logits_zero_point": 31885,
+    "get_max_batch_size": 1,
+    "get_max_context_len": 4096,
+    "get_max_seq_len": 4096,
+    "get_n_bos": 1,
+    "get_n_eos": 1,
+    "get_n_kv_heads": 3,
+    "get_n_layers": 30,
+    "get_use_kv_cache": true,
+    "get_vocab_size": 49152,
+    "methods": [
+      "get_ar_len",
+      "get_bos_id",
+      "get_dim",
+      "get_eos_id",
+      "get_head_dim",
+      "get_kv_io_bit_width",
+      "get_kv_output_0_quant_attr",
+      "get_kv_output_10_quant_attr",
+      "get_kv_output_11_quant_attr",
+      "get_kv_output_12_quant_attr",
+      "get_kv_output_13_quant_attr",
+      "get_kv_output_14_quant_attr",
+      "get_kv_output_15_quant_attr",
+      "get_kv_output_16_quant_attr",
+      "get_kv_output_17_quant_attr",
+      "get_kv_output_18_quant_attr",
+      "get_kv_output_19_quant_attr",
+      "get_kv_output_1_quant_attr",
+      "get_kv_output_20_quant_attr",
+      "get_kv_output_21_quant_attr",
+      "get_kv_output_22_quant_attr",
+      "get_kv_output_23_quant_attr",
+      "get_kv_output_24_quant_attr",
+      "get_kv_output_25_quant_attr",
+      "get_kv_output_26_quant_attr",
+      "get_kv_output_27_quant_attr",
+      "get_kv_output_28_quant_attr",
+      "get_kv_output_29_quant_attr",
+      "get_kv_output_2_quant_attr",
+      "get_kv_output_30_quant_attr",
+      "get_kv_output_31_quant_attr",
+      "get_kv_output_32_quant_attr",
+      "get_kv_output_33_quant_attr",
+      "get_kv_output_34_quant_attr",
+      "get_kv_output_35_quant_attr",
+      "get_kv_output_36_quant_attr",
+      "get_kv_output_37_quant_attr",
+      "get_kv_output_38_quant_attr",
+      "get_kv_output_39_quant_attr",
+      "get_kv_output_3_quant_attr",
+      "get_kv_output_40_quant_attr",
+      "get_kv_output_41_quant_attr",
+      "get_kv_output_42_quant_attr",
+      "get_kv_output_43_quant_attr",
+      "get_kv_output_44_quant_attr",
+      "get_kv_output_45_quant_attr",
+      "get_kv_output_46_quant_attr",
+      "get_kv_output_47_quant_attr",
+      "get_kv_output_48_quant_attr",
+      "get_kv_output_49_quant_attr",
+      "get_kv_output_4_quant_attr",
+      "get_kv_output_50_quant_attr",
+      "get_kv_output_51_quant_attr",
+      "get_kv_output_52_quant_attr",
+      "get_kv_output_53_quant_attr",
+      "get_kv_output_54_quant_attr",
+      "get_kv_output_55_quant_attr",
+      "get_kv_output_56_quant_attr",
+      "get_kv_output_57_quant_attr",
+      "get_kv_output_58_quant_attr",
+      "get_kv_output_59_quant_attr",
+      "get_kv_output_5_quant_attr",
+      "get_kv_output_6_quant_attr",
+      "get_kv_output_7_quant_attr",
+      "get_kv_output_8_quant_attr",
+      "get_kv_output_9_quant_attr",
+      "get_logits_scale",
+      "get_logits_zero_point",
+      "get_max_batch_size",
+      "get_max_context_len",
+      "get_max_seq_len",
+      "get_n_bos",
+      "get_n_eos",
+      "get_n_kv_heads",
+      "get_n_layers",
+      "get_use_kv_cache",
+      "get_vocab_size",
+      "kv_forward",
+      "prefill_forward"
+    ]
+  },
+  "smoke": {
+    "kind": "structural",
+    "passed": true,
+    "problems": [],
+    "methods": [
+      "get_ar_len",
+      "get_bos_id",
+      "get_dim",
+      "get_eos_id",
+      "get_head_dim",
+      "get_kv_io_bit_width",
+      "get_kv_output_0_quant_attr",
+      "get_kv_output_10_quant_attr",
+      "get_kv_output_11_quant_attr",
+      "get_kv_output_12_quant_attr",
+      "get_kv_output_13_quant_attr",
+      "get_kv_output_14_quant_attr",
+      "get_kv_output_15_quant_attr",
+      "get_kv_output_16_quant_attr",
+      "get_kv_output_17_quant_attr",
+      "get_kv_output_18_quant_attr",
+      "get_kv_output_19_quant_attr",
+      "get_kv_output_1_quant_attr",
+      "get_kv_output_20_quant_attr",
+      "get_kv_output_21_quant_attr",
+      "get_kv_output_22_quant_attr",
+      "get_kv_output_23_quant_attr",
+      "get_kv_output_24_quant_attr",
+      "get_kv_output_25_quant_attr",
+      "get_kv_output_26_quant_attr",
+      "get_kv_output_27_quant_attr",
+      "get_kv_output_28_quant_attr",
+      "get_kv_output_29_quant_attr",
+      "get_kv_output_2_quant_attr",
+      "get_kv_output_30_quant_attr",
+      "get_kv_output_31_quant_attr",
+      "get_kv_output_32_quant_attr",
+      "get_kv_output_33_quant_attr",
+      "get_kv_output_34_quant_attr",
+      "get_kv_output_35_quant_attr",
+      "get_kv_output_36_quant_attr",
+      "get_kv_output_37_quant_attr",
+      "get_kv_output_38_quant_attr",
+      "get_kv_output_39_quant_attr",
+      "get_kv_output_3_quant_attr",
+      "get_kv_output_40_quant_attr",
+      "get_kv_output_41_quant_attr",
+      "get_kv_output_42_quant_attr",
+      "get_kv_output_43_quant_attr",
+      "get_kv_output_44_quant_attr",
+      "get_kv_output_45_quant_attr",
+      "get_kv_output_46_quant_attr",
+      "get_kv_output_47_quant_attr",
+      "get_kv_output_48_quant_attr",
+      "get_kv_output_49_quant_attr",
+      "get_kv_output_4_quant_attr",
+      "get_kv_output_50_quant_attr",
+      "get_kv_output_51_quant_attr",
+      "get_kv_output_52_quant_attr",
+      "get_kv_output_53_quant_attr",
+      "get_kv_output_54_quant_attr",
+      "get_kv_output_55_quant_attr",
+      "get_kv_output_56_quant_attr",
+      "get_kv_output_57_quant_attr",
+      "get_kv_output_58_quant_attr",
+      "get_kv_output_59_quant_attr",
+      "get_kv_output_5_quant_attr",
+      "get_kv_output_6_quant_attr",
+      "get_kv_output_7_quant_attr",
+      "get_kv_output_8_quant_attr",
+      "get_kv_output_9_quant_attr",
+      "get_logits_scale",
+      "get_logits_zero_point",
+      "get_max_batch_size",
+      "get_max_context_len",
+      "get_max_seq_len",
+      "get_n_bos",
+      "get_n_eos",
+      "get_n_kv_heads",
+      "get_n_layers",
+      "get_use_kv_cache",
+      "get_vocab_size",
+      "kv_forward",
+      "prefill_forward"
+    ],
+    "delegates": {
+      "kv_forward": {
+        "backends": [
+          "QnnBackend"
+        ],
+        "delegate_calls": 1
+      },
+      "prefill_forward": {
+        "backends": [
+          "QnnBackend"
+        ],
+        "delegate_calls": 1
+      }
+    }
+  },
+  "run": {
+    "url": "https://github.com/ExperimentalMachines/execupack/actions/runs/37643682104",
+    "id": "37643682104"
+  }
+}