diff --git a/.gitattributes b/.gitattributes index 195842ab6152a8142b6db73bd28c3244c845cf6e..3d6c595e89878b4b616d9f10e467dc8c0997f943 100644 --- a/.gitattributes +++ b/.gitattributes @@ -767,3 +767,45 @@ fuzzy-topk-three-arm-ladder/60m-double-fuzzy-h16d-q0p25d-k4d-l10-3tpp-bs16-lr2e- fuzzy-topk-three-arm-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text fuzzy-topk-three-arm-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text fuzzy-topk-three-arm-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/training.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/training.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/training.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec filter=lfs diff=lfs merge=lfs -text +fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec filter=lfs diff=lfs merge=lfs -text diff --git a/README.md b/README.md index f22c56cdc2358a5f695ec9a59ce060fedd86b3f2..9a460b98e517f50d8b6de0f2be4dbfec8765953d 100644 --- a/README.md +++ b/README.md @@ -12,7 +12,7 @@ pretty_name: rig training logs # rig training logs Complete training logs for every GPT pretraining run in -[honglu2875/rig](https://github.com/honglu2875/rig) — 451 runs across nineteen +[honglu2875/rig](https://github.com/honglu2875/rig) — 463 runs across twenty studies, at full recorded resolution. Loss and learning-rate curves at every optimizer step; per-layer parameter, gradient, and update statistics at every diagnostic step. The raw logs themselves are not downsampled. @@ -31,7 +31,10 @@ their canonical GitHub locations. result.json configuration, final metrics, provenance metrics.json compact final metrics validation.csv held-out loss + fuzzy_sparsity.rigvec optional full per-feature vector history + fuzzy_sparsity_lossy.rigvec optional full-feature, widened-time companion records.jsonl one ledger line per run + lossy_records.jsonl optional companion hashes and retained steps snapshot.json.gz compact selected curves, for lightweight consumers full.json.gz every recorded point, loaded explicitly by the browser ``` @@ -82,6 +85,7 @@ and the dashboards show them beside each run. | `moe-gumbel-local-125M` | TPU v4 — 4 processes, 16 chips | | `sparse-autoencoder-eqflop-60M` | TPU v4 — 4 processes, 16 chips | | `fuzzy-topk-three-arm-ladder` | TPU v4 — 4 processes, 16 chips | +| `fuzzy-topk-sparsity-diagnostics-ladder` | TPU v4 — 4 processes, 16 chips | `batch-sweep-500M` is the only study spanning two chip types, and the split follows its 5- against 20-tokens-per-parameter arms. Those are separate @@ -111,6 +115,13 @@ log.series("grad.l2_norm", "block", 7) # per-layer, from diagnostics from the definitions the writer uses, so a reader in another language can be built without reading the Python. +`.rigvec` is the dense companion for metrics with a feature axis too large for +the scalar column table. Its fixed records are +`step int32 + float32[metric, layer, feature]`; the header records permanent +metric ids, shape, group size, tokens per step, and FLOPs per token. It is +memory-mapped by `rig.vectorlog.read_vector_log`, so multi-gigabyte histories +can be reduced one capture at a time. + --- Every dashboard here, the runs behind it, and the command that reproduces it. @@ -121,7 +132,7 @@ exact. ## The logs live on HuggingFace **[huggingface.co/datasets/quintic/rig-logs](https://huggingface.co/datasets/quintic/rig-logs)** -— 451 runs across nineteen studies, laid out as `//`, at full +— 463 runs across twenty studies, laid out as `//`, at full recorded resolution. That is the archive of record; its [dataset card](https://huggingface.co/datasets/quintic/rig-logs/blob/main/README.md) mirrors the GitHub report catalog and adds archive and reproduction metadata. @@ -166,6 +177,7 @@ by-product: | gumbel-local-moe | exact endpoints + mechanism reductions | — | 0.03 MB | | sparse-autoencoder-eqflop | exact endpoints + compute derivation | — | 0.02 MB | | fuzzy-topk-three-arm-ladder | exact endpoints + paired-seed/compute tables | — | 0.03 MB | +| fuzzy-topk-sparsity-diagnostics | exact mechanism reductions + 40.19 GiB raw vectors | — | 0.02 MB | The two large ones carry layer detail because gradient spikes are visible in it, and studying them is the point. This is deliberate discretion, not a @@ -228,17 +240,23 @@ total active matrix FLOPs. This dataset retains every complete trajectory, both browser payloads, exact source/config/data provenance, and the explicit incomplete-250M evidence boundary. +The [fuzzy TopK sparsity report](https://github.com/honglu2875/rig/blob/research/fuzzy-topk-autoencoder/docs/reports/fuzzy-topk-sparsity-diagnostics.html) +is a static mechanism study over twelve fuzzy-only runs from 60M through 500M. +This dataset retains 40.19 GiB of cadence-10 full-neuron vectors, 0.891 GiB of +exact widening-step companions, and the one-off interactive dead-layer, +ridgeline, histogram, and positive-quantile views. + Which metrics get charted is a declared list in `rig/report.py`, separate from the metric registry, because how a quantity should be drawn is a judgement the -registry cannot make. Everything so far is a line against the time axis; a -distribution rather than a scalar — a routing histogram, say — wants bars -against expert index and would arrive as a new chart kind rather than being -bent into a timeline. +registry cannot make. Ordinary metrics remain lines against time or layer. The +fuzzy sparsity study is the one explicit exception: zero-aware log-frequency +histograms and positive quantiles render as dedicated histogram, ridgeline, +and heatmap views rather than as scalar timelines. ## The study browser [`study-browser.html`](https://github.com/honglu2875/rig/blob/main/docs/reports/study-browser.html) -carries no run data at all — about 64 KB. It +carries no run data at all — about 80 KB. It lists the studies, renders each one's card from the dataset, and fetches only that study's overview (0.05–1.1 MB) when you pick one. A separately labelled action loads `full.json.gz`, containing every recorded point, and states its @@ -283,6 +301,7 @@ study that mixes the two topologies. | [gumbel-local-moe](https://github.com/honglu2875/rig/blob/research/gumbel-local-moe/docs/reports/gumbel-local-moe.html) | 8 | 125M | Gumbel-routed local MoE steps × seed | [`moe-gumbel-local-125M`](https://huggingface.co/datasets/quintic/rig-logs/tree/main/moe-gumbel-local-125M) | | [sparse-autoencoder-eqflop](https://github.com/honglu2875/rig/blob/research/sparse-autoencoder/docs/reports/sparse-autoencoder-eqflop.html) | 13 | 60M geometry | dictionary width × retained width at equal algorithmic FLOPs | [`sparse-autoencoder-eqflop-60M`](https://huggingface.co/datasets/quintic/rig-logs/tree/main/sparse-autoencoder-eqflop-60M) | | [fuzzy-topk-three-arm-ladder](https://github.com/honglu2875/rig/blob/research/fuzzy-topk-autoencoder/docs/reports/fuzzy-topk-three-arm-ladder.html) | 24 | 60M/125M/250M | dense vs fuzzy TopK vs double-fuzzy at matched active FLOPs | [`fuzzy-topk-three-arm-ladder`](https://huggingface.co/datasets/quintic/rig-logs/tree/main/fuzzy-topk-three-arm-ladder) | +| [fuzzy-topk-sparsity-diagnostics](https://github.com/honglu2875/rig/blob/research/fuzzy-topk-autoencoder/docs/reports/fuzzy-topk-sparsity-diagnostics.html) | 12 | 60M/125M/250M/500M | per-feature fuzzy TopK activity over training | [`fuzzy-topk-sparsity-diagnostics-ladder`](https://huggingface.co/datasets/quintic/rig-logs/tree/main/fuzzy-topk-sparsity-diagnostics-ladder) | | [transfer-charts](https://github.com/honglu2875/rig/blob/main/docs/reports/transfer-charts.html) | — | — | derived figures, not a run dashboard | — | Each study also carries a compact `snapshot.json.gz` (0.05–1.1 MB of thinned @@ -706,6 +725,38 @@ contains every endpoint, design/throughput table, interpretation, and limitation. This study's own card contains the source/config/data hashes, canonical manifest links, exclusion, and reproduction command. +## fuzzy-topk-sparsity-diagnostics.html + +Twelve verified fuzzy-only runs record full per-feature winner and positive +activation behavior from 60M through 500M. Every block's `H=16D` stored +features are observed on a 131,072-token global batch at step 1, every ten +steps, and final. The raw `.rigvec` files total 40.19 GiB. Exact full-feature +companions keep every capture through step 200, then 300, 500, 900, 1700, and +doubled gaps plus final, totaling 0.891 GiB. + +The observer is trajectory-neutral: all nine 60M–250M `training.riglog` files +are byte-identical to the same-tier, same-seed uninstrumented fuzzy runs in the +three-arm study. Almost every feature activates by step 1 and all do by step +10, but persistent inactivity is created later. At final, the fraction absent +from ten consecutive sampled batches averages 34.33%, 35.76%, 33.87%, and +36.34% over layers and seeds at 60M, 125M, 250M, and 500M. + +The effect is layer-structured. Block 1's dead fraction rises from 69.69% to +85.62% across the ladder while its normalized within-group winner entropy falls +from 0.424 to 0.103; later blocks recover activity. The +[static report](https://github.com/honglu2875/rig/blob/research/fuzzy-topk-autoencoder/docs/reports/fuzzy-topk-sparsity-diagnostics.html) +summarizes these reductions. The study browser carries the one-off tier/seed/ +layer selectors, dead heatmap, activation-frequency ridgeline, draggable-step +histogram, and positive-frequency quantiles. + +“Dead” is defined on sampled observer batches and is not proof that a feature +never fires elsewhere in the corpus. The study measures under-use; it does not +test a revival, pruning, or reallocation intervention. `records.jsonl` retains +the final 500M seed's notebook-dependency-only `pyproject.toml`/`uv.lock` dirt; +trainer, shared Python tree, config, data, runtime, and topology identities are +unchanged. The study card inside the folder gives exact endpoints, hashes, +commands, metric definitions, limitations, and loading examples. + ## transfer-charts.html Not a run dashboard. It is a self-contained derived visualization in the diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..dfe51ce8ad2f2f3ae2db33892293f5e7af3116eb --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c76ce3a3561c9ec8dacc95382e027cc5b0d2058e5db409635a906d9565340b96 +size 512740 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..8510ea25fd30af013f9d31a476fa53462cb2cd22 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f50eb0698169aca5045ad10b0a3636632a7af98f4360b54d1dcdaeef91276c20 +size 841648004 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..a6cb650029196c0698bf6fdc3f55d57fbcc3a583 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:747a2b8211685e5edcab1f62e06b0583afb571b76e54f42c2d59f462eba527c8 +size 48660644 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d404ca45cb7ae1e46fd248aab84bba2bde1d1fea --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.591915398836136, + "perplexity": 36.30354514272369, + "scored_tokens": 1048576, + "seconds": 0.517010869923979 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 466, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.6666666666666667 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 4656, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "train_tokens": 610271232 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1084.1801071023378, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 64.66830940404907, + "diagnostic_point_count": 467, + "diagnostics_every": 10, + "estimated_total_flops": 1176245951036129280, + "eval_compile_seconds": 13.15049199713394, + "final_validation_seconds": 0.517010869923979, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 7086696038400, + "dot_general": 3033331138560, + "tpu_flash_causal_attention_bwd_dkv": 1889785610240, + "tpu_flash_causal_attention_bwd_dq": 1889785610240, + "tpu_flash_causal_attention_fwd": 1889785610240 + }, + "elementwise_per_sequence": 6690029490, + "matmul_per_sequence": 15789384007680, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 1927415040, + "mfu_estimate": 0.24640456979598588, + "model_tier": "125m", + "parameter_count": 226753280, + "parameters": 226753280, + "schedule_steps": 4656, + "sparsity_diagnostic_compile_seconds": 12.556366964941844, + "sparsity_diagnostic_point_count": 467, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "tokens_per_second": 562504.7457875693, + "tokens_processed": 610271232, + "total_compile_seconds": 155.68586288695224, + "train_compile_seconds": 65.31069452082738, + "train_loss": 3.641507625579834, + "train_seconds": 1084.9174812659621, + "training_data_epochs": 0.07724972079725992, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 4656, + "training_token_budget": 610271232, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.591915398836136, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/result.json new file mode 100644 index 0000000000000000000000000000000000000000..d404ca45cb7ae1e46fd248aab84bba2bde1d1fea --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.591915398836136, + "perplexity": 36.30354514272369, + "scored_tokens": 1048576, + "seconds": 0.517010869923979 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 466, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.6666666666666667 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 4656, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "train_tokens": 610271232 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1084.1801071023378, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 64.66830940404907, + "diagnostic_point_count": 467, + "diagnostics_every": 10, + "estimated_total_flops": 1176245951036129280, + "eval_compile_seconds": 13.15049199713394, + "final_validation_seconds": 0.517010869923979, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 7086696038400, + "dot_general": 3033331138560, + "tpu_flash_causal_attention_bwd_dkv": 1889785610240, + "tpu_flash_causal_attention_bwd_dq": 1889785610240, + "tpu_flash_causal_attention_fwd": 1889785610240 + }, + "elementwise_per_sequence": 6690029490, + "matmul_per_sequence": 15789384007680, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 1927415040, + "mfu_estimate": 0.24640456979598588, + "model_tier": "125m", + "parameter_count": 226753280, + "parameters": 226753280, + "schedule_steps": 4656, + "sparsity_diagnostic_compile_seconds": 12.556366964941844, + "sparsity_diagnostic_point_count": 467, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "tokens_per_second": 562504.7457875693, + "tokens_processed": 610271232, + "total_compile_seconds": 155.68586288695224, + "train_compile_seconds": 65.31069452082738, + "train_loss": 3.641507625579834, + "train_seconds": 1084.9174812659621, + "training_data_epochs": 0.07724972079725992, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 4656, + "training_token_budget": 610271232, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.591915398836136, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..a2f585ecd73df7c4b6397d2b235f48d82e798851 Binary files /dev/null and b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/training.riglog differ diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..a069f4d7d591f50b1a45411f3225f7104699ef3d --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +4656,610271232,fineweb,fineweb,1048576,3.591915398836136,36.30354514272369,0.517010869923979,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..d233ddfc8c057630198d5569ce2e7df2163b6b21 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d61d77acd38a1d2ca4d21dc07a2f69fa9f659ac47b707966f8b494d4b536a190 +size 512740 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..1cd40898b497397f046d43484fb5e43acaecbee5 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74997501b57e7d4563b194e998c974476763b7882c7377c21a059d9fbc10bb3f +size 841648004 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..1e886e1cf7377d87eb6d18c1f1454eb5def7fef3 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7d84886e72b5c16e159bb5f9ce770a894b08859dc894b2d7ec1e9c750fb5f9b +size 48660644 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f144348e5094a9414f6466db8468029e97969cfa --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.581060439348221, + "perplexity": 35.91160273893408, + "scored_tokens": 1048576, + "seconds": 0.5136202480643988 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 466, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.6666666666666667 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 4656, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "train_tokens": 610271232 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1083.86685823526, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 64.37414500606246, + "diagnostic_point_count": 467, + "diagnostics_every": 10, + "estimated_total_flops": 1176245951036129280, + "eval_compile_seconds": 13.222909380914643, + "final_validation_seconds": 0.5136202480643988, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 7086696038400, + "dot_general": 3033331138560, + "tpu_flash_causal_attention_bwd_dkv": 1889785610240, + "tpu_flash_causal_attention_bwd_dq": 1889785610240, + "tpu_flash_causal_attention_fwd": 1889785610240 + }, + "elementwise_per_sequence": 6690029490, + "matmul_per_sequence": 15789384007680, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 1927415040, + "mfu_estimate": 0.24633337687165, + "model_tier": "125m", + "parameter_count": 226753280, + "parameters": 226753280, + "schedule_steps": 4656, + "sparsity_diagnostic_compile_seconds": 11.524073926033452, + "sparsity_diagnostic_point_count": 467, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "tokens_per_second": 562342.2229989759, + "tokens_processed": 610271232, + "total_compile_seconds": 154.58197558182292, + "train_compile_seconds": 65.46084726881236, + "train_loss": 3.6206488609313965, + "train_seconds": 1085.2310337740928, + "training_data_epochs": 0.07724972079725992, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 4656, + "training_token_budget": 610271232, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.581060439348221, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/result.json new file mode 100644 index 0000000000000000000000000000000000000000..f144348e5094a9414f6466db8468029e97969cfa --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.581060439348221, + "perplexity": 35.91160273893408, + "scored_tokens": 1048576, + "seconds": 0.5136202480643988 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 466, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.6666666666666667 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 4656, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "train_tokens": 610271232 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1083.86685823526, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 64.37414500606246, + "diagnostic_point_count": 467, + "diagnostics_every": 10, + "estimated_total_flops": 1176245951036129280, + "eval_compile_seconds": 13.222909380914643, + "final_validation_seconds": 0.5136202480643988, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 7086696038400, + "dot_general": 3033331138560, + "tpu_flash_causal_attention_bwd_dkv": 1889785610240, + "tpu_flash_causal_attention_bwd_dq": 1889785610240, + "tpu_flash_causal_attention_fwd": 1889785610240 + }, + "elementwise_per_sequence": 6690029490, + "matmul_per_sequence": 15789384007680, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 1927415040, + "mfu_estimate": 0.24633337687165, + "model_tier": "125m", + "parameter_count": 226753280, + "parameters": 226753280, + "schedule_steps": 4656, + "sparsity_diagnostic_compile_seconds": 11.524073926033452, + "sparsity_diagnostic_point_count": 467, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "tokens_per_second": 562342.2229989759, + "tokens_processed": 610271232, + "total_compile_seconds": 154.58197558182292, + "train_compile_seconds": 65.46084726881236, + "train_loss": 3.6206488609313965, + "train_seconds": 1085.2310337740928, + "training_data_epochs": 0.07724972079725992, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 4656, + "training_token_budget": 610271232, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.581060439348221, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..2fab5790fce2e556c8e9fbdf482751cc032c4bd0 Binary files /dev/null and b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/training.riglog differ diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..c7b8dcf8b2dbcfac9fec38250e3d58951c2317e2 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +4656,610271232,fineweb,fineweb,1048576,3.581060439348221,35.91160273893408,0.5136202480643988,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..75ccddbceb644fd8f2243db1a39f9c508cb9a8bc --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62b57e17b6624f065ac0b0fb5dcb7a7fc7c2b9d3670f2faec885a0538fcbbe09 +size 512740 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..211d56af2cf5fb7188f3a926b6142b9216cd98f8 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af03bbd8c3aa08f7f406368d502c226bf71b24f9be1d43eff74af7c7dc5a081f +size 841648004 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..956d382077b92648d99c290f0b0e4660367bac92 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f1f249d2f9c878bacc42332ab759f4f308f1c1ffbe872a0543ccaa3fd88b16e +size 48660644 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d7ccb005cdcf9588084a813b3e0659c2f38b9aac --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.584172636270523, + "perplexity": 36.023540814770165, + "scored_tokens": 1048576, + "seconds": 0.5274990880861878 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 466, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.6666666666666667 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 4656, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "train_tokens": 610271232 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1083.9200870397517, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 63.24943733587861, + "diagnostic_point_count": 467, + "diagnostics_every": 10, + "estimated_total_flops": 1176245951036129280, + "eval_compile_seconds": 12.438345218077302, + "final_validation_seconds": 0.5274990880861878, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 7086696038400, + "dot_general": 3033331138560, + "tpu_flash_causal_attention_bwd_dkv": 1889785610240, + "tpu_flash_causal_attention_bwd_dq": 1889785610240, + "tpu_flash_causal_attention_fwd": 1889785610240 + }, + "elementwise_per_sequence": 6690029490, + "matmul_per_sequence": 15789384007680, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 1927415040, + "mfu_estimate": 0.2463454743272163, + "model_tier": "125m", + "parameter_count": 226753280, + "parameters": 226753280, + "schedule_steps": 4656, + "sparsity_diagnostic_compile_seconds": 11.386971961008385, + "sparsity_diagnostic_point_count": 467, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "tokens_per_second": 562369.8396790302, + "tokens_processed": 610271232, + "total_compile_seconds": 135.26415866799653, + "train_compile_seconds": 48.18940415303223, + "train_loss": 3.5505714416503906, + "train_seconds": 1085.1777405920438, + "training_data_epochs": 0.07724972079725992, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 4656, + "training_token_budget": 610271232, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.584172636270523, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/result.json new file mode 100644 index 0000000000000000000000000000000000000000..d7ccb005cdcf9588084a813b3e0659c2f38b9aac --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.584172636270523, + "perplexity": 36.023540814770165, + "scored_tokens": 1048576, + "seconds": 0.5274990880861878 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 640, + "heads": 10, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 2560, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "125m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 466, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.6666666666666667 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 4656, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "train_tokens": 610271232 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1083.9200870397517, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 63.24943733587861, + "diagnostic_point_count": 467, + "diagnostics_every": 10, + "estimated_total_flops": 1176245951036129280, + "eval_compile_seconds": 12.438345218077302, + "final_validation_seconds": 0.5274990880861878, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 7086696038400, + "dot_general": 3033331138560, + "tpu_flash_causal_attention_bwd_dkv": 1889785610240, + "tpu_flash_causal_attention_bwd_dq": 1889785610240, + "tpu_flash_causal_attention_fwd": 1889785610240 + }, + "elementwise_per_sequence": 6690029490, + "matmul_per_sequence": 15789384007680, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 1927415040, + "mfu_estimate": 0.2463454743272163, + "model_tier": "125m", + "parameter_count": 226753280, + "parameters": 226753280, + "schedule_steps": 4656, + "sparsity_diagnostic_compile_seconds": 11.386971961008385, + "sparsity_diagnostic_point_count": 467, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.6913446720594294, + "tokens_per_parameter": 2.6913446720594294, + "tokens_per_second": 562369.8396790302, + "tokens_processed": 610271232, + "total_compile_seconds": 135.26415866799653, + "train_compile_seconds": 48.18940415303223, + "train_loss": 3.5505714416503906, + "train_seconds": 1085.1777405920438, + "training_data_epochs": 0.07724972079725992, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 4656, + "training_token_budget": 610271232, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.584172636270523, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..d481ee89186a88aefa8a90a61afd4a55afba1195 Binary files /dev/null and b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/training.riglog differ diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..eeabbced9b49d464f4a314aa0a8062733dd6a853 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +4656,610271232,fineweb,fineweb,1048576,3.584172636270523,36.023540814770165,0.5274990880861878,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..192ab6b981804dbe8ae2978f5e841932f3797113 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6c350e8a3a3b73522c617c30bad2dfcd13c42ee48ec6fe6150300deba4ba0da +size 1232408 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..9099bddfbd69db5e1bdb98847dc231f84546500f --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a300d96d0c1a4bc069975ce48f5fbca9c3e0e94fd86110360cc23e05a4e516d +size 3025014512 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..5c31c0465b62d747a3ff07e8bdc8f26dd797662e --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f88d3aba739b15e5f9481deee7081550aa183bc6683f01192c4cc72eacf5119f +size 89915560 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..be4c129752c76d14d8e7472119d6f8c4fdf6abbd --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.314135879278183, + "perplexity": 27.498621586202635, + "scored_tokens": 1048576, + "seconds": 0.9843828401062638 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 940, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.1666666666666667, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 2.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 9402, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "train_tokens": 1232338944 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1122.39495655022, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 76.33016821881756, + "diagnostic_point_count": 942, + "diagnostics_every": 10, + "estimated_total_flops": 4963669455340044288, + "eval_compile_seconds": 12.449057026999071, + "final_validation_seconds": 0.9843828401062638, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 17678085390336, + "dot_general": 5216252461056, + "tpu_flash_causal_attention_bwd_dkv": 3367254360064, + "tpu_flash_causal_attention_bwd_dq": 3367254360064, + "tpu_flash_causal_attention_fwd": 3367254360064 + }, + "elementwise_per_sequence": 8887879538, + "matmul_per_sequence": 32996100931584, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 4027844352, + "mfu_estimate": 0.25508976285232277, + "model_tier": "250m", + "parameter_count": 495053440, + "parameters": 495053440, + "schedule_steps": 9402, + "sparsity_diagnostic_compile_seconds": 14.994110438041389, + "sparsity_diagnostic_point_count": 942, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "tokens_per_second": 278658.9695286766, + "tokens_processed": 1232338944, + "total_compile_seconds": 164.92245301185176, + "train_compile_seconds": 61.149117327993736, + "train_loss": 3.3761191368103027, + "train_seconds": 4422.391090027988, + "training_data_epochs": 0.15599267073364215, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 9402, + "training_token_budget": 1232338944, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.314135879278183, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/result.json new file mode 100644 index 0000000000000000000000000000000000000000..be4c129752c76d14d8e7472119d6f8c4fdf6abbd --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.314135879278183, + "perplexity": 27.498621586202635, + "scored_tokens": 1048576, + "seconds": 0.9843828401062638 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 940, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.1666666666666667, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 2.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 9402, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "train_tokens": 1232338944 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1122.39495655022, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 76.33016821881756, + "diagnostic_point_count": 942, + "diagnostics_every": 10, + "estimated_total_flops": 4963669455340044288, + "eval_compile_seconds": 12.449057026999071, + "final_validation_seconds": 0.9843828401062638, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 17678085390336, + "dot_general": 5216252461056, + "tpu_flash_causal_attention_bwd_dkv": 3367254360064, + "tpu_flash_causal_attention_bwd_dq": 3367254360064, + "tpu_flash_causal_attention_fwd": 3367254360064 + }, + "elementwise_per_sequence": 8887879538, + "matmul_per_sequence": 32996100931584, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 4027844352, + "mfu_estimate": 0.25508976285232277, + "model_tier": "250m", + "parameter_count": 495053440, + "parameters": 495053440, + "schedule_steps": 9402, + "sparsity_diagnostic_compile_seconds": 14.994110438041389, + "sparsity_diagnostic_point_count": 942, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "tokens_per_second": 278658.9695286766, + "tokens_processed": 1232338944, + "total_compile_seconds": 164.92245301185176, + "train_compile_seconds": 61.149117327993736, + "train_loss": 3.3761191368103027, + "train_seconds": 4422.391090027988, + "training_data_epochs": 0.15599267073364215, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 9402, + "training_token_budget": 1232338944, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.314135879278183, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..70ee7a9ac765959985cd75421554418ecb574c22 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/training.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34e549a8d940eb2afcd495773e93aa9a2071943cca85bce4736980d9898e670d +size 150536 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..e463cc90fef0eaa6f0962a9bd8388add22479bff --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +9402,1232338944,fineweb,fineweb,1048576,3.314135879278183,27.498621586202635,0.9843828401062638,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..82556ceb16cb110c44a2b48f1f34c3423bb05cde --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e90ea4b61a38334d46f127cc46689466bc48505961e625f0d3afe109c91430e +size 1232408 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..394e6a48285d9e6a1357ec50985f0ac348b436b5 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:228bb4a2438541a9a3cbc86ed3d3d9383279cf361af9c4c7b3566f20669427c7 +size 3025014512 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..fa0fc045d0114b6e24c39f4d0dc296a891b2ff82 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c6eb8512c8cb5867d9390626a5d99a60d515bdb5ac4518d8b365f9e3d6ce9c7 +size 89915560 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..df107c1bcf5582dd4aecca5b89e85167127a3ada --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.3040049374103546, + "perplexity": 27.221441070036004, + "scored_tokens": 1048576, + "seconds": 0.9745558970607817 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 940, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.1666666666666667, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 2.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 9402, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "train_tokens": 1232338944 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1122.3941557798357, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 81.67842828086577, + "diagnostic_point_count": 942, + "diagnostics_every": 10, + "estimated_total_flops": 4963669455340044288, + "eval_compile_seconds": 12.553831937955692, + "final_validation_seconds": 0.9745558970607817, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 17678085390336, + "dot_general": 5216252461056, + "tpu_flash_causal_attention_bwd_dkv": 3367254360064, + "tpu_flash_causal_attention_bwd_dq": 3367254360064, + "tpu_flash_causal_attention_fwd": 3367254360064 + }, + "elementwise_per_sequence": 8887879538, + "matmul_per_sequence": 32996100931584, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 4027844352, + "mfu_estimate": 0.2550895808590536, + "model_tier": "250m", + "parameter_count": 495053440, + "parameters": 495053440, + "schedule_steps": 9402, + "sparsity_diagnostic_compile_seconds": 15.16223825304769, + "sparsity_diagnostic_point_count": 942, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "tokens_per_second": 278658.77072000515, + "tokens_processed": 1232338944, + "total_compile_seconds": 166.9548121576663, + "train_compile_seconds": 57.560313685797155, + "train_loss": 3.1848177909851074, + "train_seconds": 4422.39424517611, + "training_data_epochs": 0.15599267073364215, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 9402, + "training_token_budget": 1232338944, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.3040049374103546, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/result.json new file mode 100644 index 0000000000000000000000000000000000000000..df107c1bcf5582dd4aecca5b89e85167127a3ada --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.3040049374103546, + "perplexity": 27.221441070036004, + "scored_tokens": 1048576, + "seconds": 0.9745558970607817 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 940, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.1666666666666667, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 2.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 9402, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "train_tokens": 1232338944 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1122.3941557798357, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 81.67842828086577, + "diagnostic_point_count": 942, + "diagnostics_every": 10, + "estimated_total_flops": 4963669455340044288, + "eval_compile_seconds": 12.553831937955692, + "final_validation_seconds": 0.9745558970607817, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 17678085390336, + "dot_general": 5216252461056, + "tpu_flash_causal_attention_bwd_dkv": 3367254360064, + "tpu_flash_causal_attention_bwd_dq": 3367254360064, + "tpu_flash_causal_attention_fwd": 3367254360064 + }, + "elementwise_per_sequence": 8887879538, + "matmul_per_sequence": 32996100931584, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 4027844352, + "mfu_estimate": 0.2550895808590536, + "model_tier": "250m", + "parameter_count": 495053440, + "parameters": 495053440, + "schedule_steps": 9402, + "sparsity_diagnostic_compile_seconds": 15.16223825304769, + "sparsity_diagnostic_point_count": 942, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "tokens_per_second": 278658.77072000515, + "tokens_processed": 1232338944, + "total_compile_seconds": 166.9548121576663, + "train_compile_seconds": 57.560313685797155, + "train_loss": 3.1848177909851074, + "train_seconds": 4422.39424517611, + "training_data_epochs": 0.15599267073364215, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 9402, + "training_token_budget": 1232338944, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.3040049374103546, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..2595a1302df557bf1e5b3b39e6a0fd3f23615e94 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/training.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70c0ffdd65ec23cc21e1f21f5d919c167b7c30242c2e0c5e7ee4926fc9a64831 +size 150536 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..dc958488979b477c5a9f8882b0feb6459db850ec --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +9402,1232338944,fineweb,fineweb,1048576,3.3040049374103546,27.221441070036004,0.9745558970607817,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..ebb5d47bc30e7af9acf3a27b1e8f90c57fe798e2 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d5ec373ee087cda4d56c63d826dc29b4a042458ef3f1e72a5e83a8e2b29ee8a +size 1232408 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..49d3b595f0769b8b5eae1ba5176fb94d675856f9 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:822c4975de87808dd44684e2cc6f52650e7da926b76624a9bce0284bc87ddaed +size 3025014512 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..a79014113cdb8fbb394811d1cac6e337b4fba959 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f15f58782ff3ee42fbf1208790f45f2a95dc054ad36ab583aafd4b28a35e2c4 +size 89915560 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..497e637352ed0184ebfd41690b79f82ed03aba0b --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.306984007358551, + "perplexity": 27.302656560322916, + "scored_tokens": 1048576, + "seconds": 0.9671843170654029 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 940, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.1666666666666667, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 2.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 9402, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "train_tokens": 1232338944 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1122.6806043428, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 83.23276506806724, + "diagnostic_point_count": 942, + "diagnostics_every": 10, + "estimated_total_flops": 4963669455340044288, + "eval_compile_seconds": 12.998791442951187, + "final_validation_seconds": 0.9671843170654029, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 17678085390336, + "dot_general": 5216252461056, + "tpu_flash_causal_attention_bwd_dkv": 3367254360064, + "tpu_flash_causal_attention_bwd_dq": 3367254360064, + "tpu_flash_causal_attention_fwd": 3367254360064 + }, + "elementwise_per_sequence": 8887879538, + "matmul_per_sequence": 32996100931584, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 4027844352, + "mfu_estimate": 0.25515468280518183, + "model_tier": "250m", + "parameter_count": 495053440, + "parameters": 495053440, + "schedule_steps": 9402, + "sparsity_diagnostic_compile_seconds": 15.126094905892387, + "sparsity_diagnostic_point_count": 942, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "tokens_per_second": 278729.88780843536, + "tokens_processed": 1232338944, + "total_compile_seconds": 181.17454498284496, + "train_compile_seconds": 69.81689356593415, + "train_loss": 3.2581043243408203, + "train_seconds": 4421.265884650871, + "training_data_epochs": 0.15599267073364215, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 9402, + "training_token_budget": 1232338944, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.306984007358551, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/result.json new file mode 100644 index 0000000000000000000000000000000000000000..497e637352ed0184ebfd41690b79f82ed03aba0b --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.306984007358551, + "perplexity": 27.302656560322916, + "scored_tokens": 1048576, + "seconds": 0.9671843170654029 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 896, + "heads": 14, + "layers": 14, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 3584, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "250m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 940, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.1666666666666667, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 2.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 9402, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "train_tokens": 1232338944 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1122.6806043428, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 83.23276506806724, + "diagnostic_point_count": 942, + "diagnostics_every": 10, + "estimated_total_flops": 4963669455340044288, + "eval_compile_seconds": 12.998791442951187, + "final_validation_seconds": 0.9671843170654029, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 17678085390336, + "dot_general": 5216252461056, + "tpu_flash_causal_attention_bwd_dkv": 3367254360064, + "tpu_flash_causal_attention_bwd_dq": 3367254360064, + "tpu_flash_causal_attention_fwd": 3367254360064 + }, + "elementwise_per_sequence": 8887879538, + "matmul_per_sequence": 32996100931584, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 4027844352, + "mfu_estimate": 0.25515468280518183, + "model_tier": "250m", + "parameter_count": 495053440, + "parameters": 495053440, + "schedule_steps": 9402, + "sparsity_diagnostic_compile_seconds": 15.126094905892387, + "sparsity_diagnostic_point_count": 942, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.489304879893371, + "tokens_per_parameter": 2.489304879893371, + "tokens_per_second": 278729.88780843536, + "tokens_processed": 1232338944, + "total_compile_seconds": 181.17454498284496, + "train_compile_seconds": 69.81689356593415, + "train_loss": 3.2581043243408203, + "train_seconds": 4421.265884650871, + "training_data_epochs": 0.15599267073364215, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 9402, + "training_token_budget": 1232338944, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.306984007358551, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..3a2ec61e98a3357cb327cac546998fbe7b671ad0 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/training.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6062f911a5a81665bc9fdd1259b13bc63f66d91a9523161437631f1a195ffc2 +size 150536 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..90571a4f4b1be498de8f8cebe37be991ac9758b0 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +9402,1232338944,fineweb,fineweb,1048576,3.306984007358551,27.302656560322916,0.9671843170654029,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..9a3e264224a9a535c75f83c8ab67180824cdeb5c --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ccd0cb279b9405d22db1b4a5a9c328a93070e5670ff52711a9a693016fd2e5ea +size 2836024 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..8a135ae78cd2419ca7960b5ae8b5c1bae10f9e4f --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e982261923a00d3cd5962a7adcff671c2600400cff132fdeb3d6660383046ea +size 10265566928 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..e9e86424241d9e450c6ee787c230070fdb79ca5a --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0c72cb5b650b5426e55a6b38c53a288611413020d62e00aefb8f5256fad31a7 +size 152043692 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..989dff4e20221f5e5080a23b3c9a52387c220be7 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.093101680278778, + "perplexity": 22.04534966885452, + "scored_tokens": 1048576, + "seconds": 1.7293408529367298 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 1956, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.3333333333333333, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 3.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 19561, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "train_tokens": 2563899392 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1164.0399494191674, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 92.70696510304697, + "diagnostic_point_count": 1958, + "diagnostics_every": 10, + "estimated_total_flops": 21023633708112936960, + "eval_compile_seconds": 16.746170554775745, + "final_validation_seconds": 1.7293408529367298, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 41231686041600, + "dot_general": 9448949022720, + "tpu_flash_causal_attention_bwd_dkv": 5497558138880, + "tpu_flash_causal_attention_bwd_dq": 5497558138880, + "tpu_flash_causal_attention_fwd": 5497558138880 + }, + "elementwise_per_sequence": 12057100530, + "matmul_per_sequence": 67173309480960, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 8199866880, + "mfu_estimate": 0.2645545339589017, + "model_tier": "500m", + "parameter_count": 1072968960, + "parameters": 1072968960, + "schedule_steps": 19561, + "sparsity_diagnostic_compile_seconds": 19.950090633006766, + "sparsity_diagnostic_point_count": 1958, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "tokens_per_second": 141958.39596595592, + "tokens_processed": 2563899392, + "total_compile_seconds": 203.57062846887857, + "train_compile_seconds": 74.16740217804909, + "train_loss": 3.0271120071411133, + "train_seconds": 18060.921120966086, + "training_data_epochs": 0.32454505767079067, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 19561, + "training_token_budget": 2563899392, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.093101680278778, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/result.json new file mode 100644 index 0000000000000000000000000000000000000000..989dff4e20221f5e5080a23b3c9a52387c220be7 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.093101680278778, + "perplexity": 22.04534966885452, + "scored_tokens": 1048576, + "seconds": 1.7293408529367298 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 1956, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.3333333333333333, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 3.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 19561, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "train_tokens": 2563899392 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1164.0399494191674, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 92.70696510304697, + "diagnostic_point_count": 1958, + "diagnostics_every": 10, + "estimated_total_flops": 21023633708112936960, + "eval_compile_seconds": 16.746170554775745, + "final_validation_seconds": 1.7293408529367298, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 41231686041600, + "dot_general": 9448949022720, + "tpu_flash_causal_attention_bwd_dkv": 5497558138880, + "tpu_flash_causal_attention_bwd_dq": 5497558138880, + "tpu_flash_causal_attention_fwd": 5497558138880 + }, + "elementwise_per_sequence": 12057100530, + "matmul_per_sequence": 67173309480960, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 8199866880, + "mfu_estimate": 0.2645545339589017, + "model_tier": "500m", + "parameter_count": 1072968960, + "parameters": 1072968960, + "schedule_steps": 19561, + "sparsity_diagnostic_compile_seconds": 19.950090633006766, + "sparsity_diagnostic_point_count": 1958, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "tokens_per_second": 141958.39596595592, + "tokens_processed": 2563899392, + "total_compile_seconds": 203.57062846887857, + "train_compile_seconds": 74.16740217804909, + "train_loss": 3.0271120071411133, + "train_seconds": 18060.921120966086, + "training_data_epochs": 0.32454505767079067, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 19561, + "training_token_budget": 2563899392, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.093101680278778, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..7d77cf8170c3cf994b80826f86486127d230620a --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/training.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e6ede524e9c5b6ee96944b98f792696f086025eadf0ca64da88cebd1a51cb6a +size 313080 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..305be19f8a657e397dc05f8321c7e458d5930338 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +19561,2563899392,fineweb,fineweb,1048576,3.093101680278778,22.04534966885452,1.7293408529367298,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..c4d330f080a0099ae9097d410831c08d6e7aa8fb --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:123ffd0c707339ffbc1cf8f1d045af8ed1f06c0b097e654f6aa1e5a6188b3268 +size 2836024 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..600f4163f2c5ddcc3523bce1f393db7ca8e2269e --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05bf65f865e67c9fad86c291ee942ab6b6f39cbe3286a1e6eeb15f23586ca13e +size 10265566928 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..e95a3f0194142cca2702e5ace756f2a5f9cb1536 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29345feda05f0a5708cf8cb21656f15987f21a6c78ece2bd66cb1054cf43ac35 +size 152043692 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1d4573a71c503ad76485b88860ee21a6deb0d449 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.080546021461487, + "perplexity": 21.770286194982404, + "scored_tokens": 1048576, + "seconds": 1.743782521924004 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 1956, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.3333333333333333, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 3.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 19561, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "train_tokens": 2563899392 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1163.9397576841584, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 94.47339812200516, + "diagnostic_point_count": 1958, + "diagnostics_every": 10, + "estimated_total_flops": 21023633708112936960, + "eval_compile_seconds": 17.084043910959736, + "final_validation_seconds": 1.743782521924004, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 41231686041600, + "dot_general": 9448949022720, + "tpu_flash_causal_attention_bwd_dkv": 5497558138880, + "tpu_flash_causal_attention_bwd_dq": 5497558138880, + "tpu_flash_causal_attention_fwd": 5497558138880 + }, + "elementwise_per_sequence": 12057100530, + "matmul_per_sequence": 67173309480960, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 8199866880, + "mfu_estimate": 0.264531763110036, + "model_tier": "500m", + "parameter_count": 1072968960, + "parameters": 1072968960, + "schedule_steps": 19561, + "sparsity_diagnostic_compile_seconds": 19.11918603302911, + "sparsity_diagnostic_point_count": 1958, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "tokens_per_second": 141946.1772633263, + "tokens_processed": 2563899392, + "total_compile_seconds": 203.83854955085553, + "train_compile_seconds": 73.16192148486152, + "train_loss": 3.1012890338897705, + "train_seconds": 18062.475801963126, + "training_data_epochs": 0.32454505767079067, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 19561, + "training_token_budget": 2563899392, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.080546021461487, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/result.json new file mode 100644 index 0000000000000000000000000000000000000000..1d4573a71c503ad76485b88860ee21a6deb0d449 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.080546021461487, + "perplexity": 21.770286194982404, + "scored_tokens": 1048576, + "seconds": 1.743782521924004 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 1956, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.3333333333333333, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 3.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 19561, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "train_tokens": 2563899392 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1163.9397576841584, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 94.47339812200516, + "diagnostic_point_count": 1958, + "diagnostics_every": 10, + "estimated_total_flops": 21023633708112936960, + "eval_compile_seconds": 17.084043910959736, + "final_validation_seconds": 1.743782521924004, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 41231686041600, + "dot_general": 9448949022720, + "tpu_flash_causal_attention_bwd_dkv": 5497558138880, + "tpu_flash_causal_attention_bwd_dq": 5497558138880, + "tpu_flash_causal_attention_fwd": 5497558138880 + }, + "elementwise_per_sequence": 12057100530, + "matmul_per_sequence": 67173309480960, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 8199866880, + "mfu_estimate": 0.264531763110036, + "model_tier": "500m", + "parameter_count": 1072968960, + "parameters": 1072968960, + "schedule_steps": 19561, + "sparsity_diagnostic_compile_seconds": 19.11918603302911, + "sparsity_diagnostic_point_count": 1958, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "tokens_per_second": 141946.1772633263, + "tokens_processed": 2563899392, + "total_compile_seconds": 203.83854955085553, + "train_compile_seconds": 73.16192148486152, + "train_loss": 3.1012890338897705, + "train_seconds": 18062.475801963126, + "training_data_epochs": 0.32454505767079067, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 19561, + "training_token_budget": 2563899392, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.080546021461487, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..a6af5bbbe8cc049b3f22b266e2b1a2032e0acb8f --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/training.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2faaa344826c61eec1b727a37a3035180afc61584694b464dd101ec6941675c +size 313080 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..4a545dd70a6c9d82722a0f09ca056948294f15ec --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +19561,2563899392,fineweb,fineweb,1048576,3.080546021461487,21.770286194982404,1.743782521924004,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..8981f79034e82d98b135bb7dd8409512c7e88462 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af9d34d62f7f2931698edeeb3dd78253380c282409c2fccf583b1fb810a528d2 +size 2836024 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..df1947dc1ac7853ac08b062e7394b691b0b2c2a7 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82f3cfc8d254e6eebb77330c07b394fe8385db90aec5252d4b362177639db7e0 +size 10265566928 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..41337f18083a9346d7e873904cd4469bacd4fcb0 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:523f999acff33b5c3e925db4fd5658321b4485faa99c4076c10686b7fe21804e +size 152043692 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..21e8a14a194b0ef81b8323b7f2ede7842e32ecfb --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.1003285944461823, + "perplexity": 22.20524660348485, + "scored_tokens": 1048576, + "seconds": 1.7348054798785597 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 1956, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.3333333333333333, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 3.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 19561, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "train_tokens": 2563899392 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1164.112342725145, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 96.65750715904869, + "diagnostic_point_count": 1958, + "diagnostics_every": 10, + "estimated_total_flops": 21023633708112936960, + "eval_compile_seconds": 17.654935136903077, + "final_validation_seconds": 1.7348054798785597, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 41231686041600, + "dot_general": 9448949022720, + "tpu_flash_causal_attention_bwd_dkv": 5497558138880, + "tpu_flash_causal_attention_bwd_dq": 5497558138880, + "tpu_flash_causal_attention_fwd": 5497558138880 + }, + "elementwise_per_sequence": 12057100530, + "matmul_per_sequence": 67173309480960, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 8199866880, + "mfu_estimate": 0.26457098698298753, + "model_tier": "500m", + "parameter_count": 1072968960, + "parameters": 1072968960, + "schedule_steps": 19561, + "sparsity_diagnostic_compile_seconds": 19.68290765210986, + "sparsity_diagnostic_point_count": 1958, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "tokens_per_second": 141967.2245612291, + "tokens_processed": 2563899392, + "total_compile_seconds": 224.390300540952, + "train_compile_seconds": 90.39495059289038, + "train_loss": 2.885099411010742, + "train_seconds": 18059.797956353053, + "training_data_epochs": 0.32454505767079067, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 19561, + "training_token_budget": 2563899392, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.1003285944461823, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/result.json new file mode 100644 index 0000000000000000000000000000000000000000..21e8a14a194b0ef81b8323b7f2ede7842e32ecfb --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.1003285944461823, + "perplexity": 22.20524660348485, + "scored_tokens": 1048576, + "seconds": 1.7348054798785597 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 1280, + "heads": 20, + "layers": 16, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 5120, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "500m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 1956, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 1.3333333333333333, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 3.3333333333333335 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 19561, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "train_tokens": 2563899392 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 1164.112342725145, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 96.65750715904869, + "diagnostic_point_count": 1958, + "diagnostics_every": 10, + "estimated_total_flops": 21023633708112936960, + "eval_compile_seconds": 17.654935136903077, + "final_validation_seconds": 1.7348054798785597, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 41231686041600, + "dot_general": 9448949022720, + "tpu_flash_causal_attention_bwd_dkv": 5497558138880, + "tpu_flash_causal_attention_bwd_dq": 5497558138880, + "tpu_flash_causal_attention_fwd": 5497558138880 + }, + "elementwise_per_sequence": 12057100530, + "matmul_per_sequence": 67173309480960, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 8199866880, + "mfu_estimate": 0.26457098698298753, + "model_tier": "500m", + "parameter_count": 1072968960, + "parameters": 1072968960, + "schedule_steps": 19561, + "sparsity_diagnostic_compile_seconds": 19.68290765210986, + "sparsity_diagnostic_point_count": 1958, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 2.389537337594556, + "tokens_per_parameter": 2.389537337594556, + "tokens_per_second": 141967.2245612291, + "tokens_processed": 2563899392, + "total_compile_seconds": 224.390300540952, + "train_compile_seconds": 90.39495059289038, + "train_loss": 2.885099411010742, + "train_seconds": 18059.797956353053, + "training_data_epochs": 0.32454505767079067, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 19561, + "training_token_budget": 2563899392, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.1003285944461823, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..aec9b4f1a2a5ce46f6b481a41df08cdd3bec9596 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/training.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:006b65cd1635eafe32fa277c75b11080b855c61a382cf6d166cf355b17eadebd +size 313080 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..34acb49eac7b2b835f82c37ee464285cd11c36d7 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +19561,2563899392,fineweb,fineweb,1048576,3.1003285944461823,22.20524660348485,1.7348054798785597,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..1aa348a9841ae8b233fa5bfa67315205cf60ecd3 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06bb668b3646f7e7630d6c3c420a6dd16c67a2d9eb7517fb1ec0a96a02e33fea +size 259084 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..a03c6620156a17de5d6e34016f5d26786f87582e --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:158e7d03f0f88d4916f9283e0ca6a4441142fd7061dc095288a742c25dd7518b +size 251954140 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..b3ebce79662bb3dc6bb917814533403f20733bbf --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b78a7f0c5af38acc3c7f4d308b455591947a1677280b58260215144d9dbf7711 +size 28115104 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cf81fc1b14491943635634f43bb3879e3c0357d3 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.9218744337558746, + "perplexity": 50.49500566944533, + "scored_tokens": 1048576, + "seconds": 0.3247907080221921 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 232, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.0 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 2316, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "train_tokens": 303562752 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 990.6911316196339, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 60.6359039300587, + "diagnostic_point_count": 233, + "diagnostics_every": 10, + "estimated_total_flops": 280151604005437440, + "eval_compile_seconds": 10.184669547015801, + "final_validation_seconds": 0.3247907080221921, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 2551210573824, + "dot_general": 1607397801984, + "tpu_flash_causal_attention_bwd_dkv": 1133871366144, + "tpu_flash_causal_attention_bwd_dq": 1133871366144, + "tpu_flash_causal_attention_fwd": 1133871366144 + }, + "elementwise_per_sequence": 5532367282, + "matmul_per_sequence": 7560222474240, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 922878720, + "mfu_estimate": 0.22515707536809862, + "model_tier": "60m", + "parameter_count": 97123584, + "parameters": 97123584, + "schedule_steps": 2316, + "sparsity_diagnostic_compile_seconds": 10.30694063496776, + "sparsity_diagnostic_point_count": 233, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "tokens_per_second": 1073479.2233801142, + "tokens_processed": 303562752, + "total_compile_seconds": 119.93457057909109, + "train_compile_seconds": 38.807056467048824, + "train_loss": 3.8996334075927734, + "train_seconds": 282.78400307008997, + "training_data_epochs": 0.03842576318008032, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 2316, + "training_token_budget": 303562752, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.9218744337558746, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/result.json new file mode 100644 index 0000000000000000000000000000000000000000..cf81fc1b14491943635634f43bb3879e3c0357d3 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.9218744337558746, + "perplexity": 50.49500566944533, + "scored_tokens": 1048576, + "seconds": 0.3247907080221921 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 232, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.0 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 2316, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "train_tokens": 303562752 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 990.6911316196339, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 60.6359039300587, + "diagnostic_point_count": 233, + "diagnostics_every": 10, + "estimated_total_flops": 280151604005437440, + "eval_compile_seconds": 10.184669547015801, + "final_validation_seconds": 0.3247907080221921, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 2551210573824, + "dot_general": 1607397801984, + "tpu_flash_causal_attention_bwd_dkv": 1133871366144, + "tpu_flash_causal_attention_bwd_dq": 1133871366144, + "tpu_flash_causal_attention_fwd": 1133871366144 + }, + "elementwise_per_sequence": 5532367282, + "matmul_per_sequence": 7560222474240, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 922878720, + "mfu_estimate": 0.22515707536809862, + "model_tier": "60m", + "parameter_count": 97123584, + "parameters": 97123584, + "schedule_steps": 2316, + "sparsity_diagnostic_compile_seconds": 10.30694063496776, + "sparsity_diagnostic_point_count": 233, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "tokens_per_second": 1073479.2233801142, + "tokens_processed": 303562752, + "total_compile_seconds": 119.93457057909109, + "train_compile_seconds": 38.807056467048824, + "train_loss": 3.8996334075927734, + "train_seconds": 282.78400307008997, + "training_data_epochs": 0.03842576318008032, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 2316, + "training_token_budget": 303562752, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.9218744337558746, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1337, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..e658abfd6a1cb0178e77478ebdee96819ae0c68f Binary files /dev/null and b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/training.riglog differ diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..a408d49c7b8e931aa3111ecff8d7f102b8cb2a20 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +2316,303562752,fineweb,fineweb,1048576,3.9218744337558746,50.49500566944533,0.3247907080221921,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..4bbbf657d73ae4212a66614523f164ee997ec086 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d66db2838496c252f36163cb7e8776effa82d0622750d0a3c4f01d3dc1f9641 +size 259084 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..e22ef93bf4226f363f1eb4d4fa6bf128ead6c5bf --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e324a586cca82df9638914c1dd4cad2ede8c83ea18441cab1b31db8742a5bfea +size 251954140 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..9d0f04d3d0686e80a67ec074cac32d08dea0ec36 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9029bea602b2f4c0e94cb825a9ba309d7514bce9a408a175983afa3a89453f9c +size 28115104 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5203ddfe41f41581e80bb74e661912112d930cc3 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.9359951615333557, + "perplexity": 51.213089902791445, + "scored_tokens": 1048576, + "seconds": 0.2988725209143013 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 232, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.0 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 2316, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "train_tokens": 303562752 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 993.3663244029677, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 62.959355283062905, + "diagnostic_point_count": 233, + "diagnostics_every": 10, + "estimated_total_flops": 280151604005437440, + "eval_compile_seconds": 9.965425219153985, + "final_validation_seconds": 0.2988725209143013, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 2551210573824, + "dot_general": 1607397801984, + "tpu_flash_causal_attention_bwd_dkv": 1133871366144, + "tpu_flash_causal_attention_bwd_dq": 1133871366144, + "tpu_flash_causal_attention_fwd": 1133871366144 + }, + "elementwise_per_sequence": 5532367282, + "matmul_per_sequence": 7560222474240, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 922878720, + "mfu_estimate": 0.2257650737279472, + "model_tier": "60m", + "parameter_count": 97123584, + "parameters": 97123584, + "schedule_steps": 2316, + "sparsity_diagnostic_compile_seconds": 10.03020230308175, + "sparsity_diagnostic_point_count": 233, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "tokens_per_second": 1076377.971314549, + "tokens_processed": 303562752, + "total_compile_seconds": 122.27075585327111, + "train_compile_seconds": 39.31577304797247, + "train_loss": 4.031710147857666, + "train_seconds": 282.0224494461436, + "training_data_epochs": 0.03842576318008032, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 2316, + "training_token_budget": 303562752, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.9359951615333557, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/result.json new file mode 100644 index 0000000000000000000000000000000000000000..5203ddfe41f41581e80bb74e661912112d930cc3 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.9359951615333557, + "perplexity": 51.213089902791445, + "scored_tokens": 1048576, + "seconds": 0.2988725209143013 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 232, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.0 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 2316, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "train_tokens": 303562752 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 993.3663244029677, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 62.959355283062905, + "diagnostic_point_count": 233, + "diagnostics_every": 10, + "estimated_total_flops": 280151604005437440, + "eval_compile_seconds": 9.965425219153985, + "final_validation_seconds": 0.2988725209143013, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 2551210573824, + "dot_general": 1607397801984, + "tpu_flash_causal_attention_bwd_dkv": 1133871366144, + "tpu_flash_causal_attention_bwd_dq": 1133871366144, + "tpu_flash_causal_attention_fwd": 1133871366144 + }, + "elementwise_per_sequence": 5532367282, + "matmul_per_sequence": 7560222474240, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 922878720, + "mfu_estimate": 0.2257650737279472, + "model_tier": "60m", + "parameter_count": 97123584, + "parameters": 97123584, + "schedule_steps": 2316, + "sparsity_diagnostic_compile_seconds": 10.03020230308175, + "sparsity_diagnostic_point_count": 233, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "tokens_per_second": 1076377.971314549, + "tokens_processed": 303562752, + "total_compile_seconds": 122.27075585327111, + "train_compile_seconds": 39.31577304797247, + "train_loss": 4.031710147857666, + "train_seconds": 282.0224494461436, + "training_data_epochs": 0.03842576318008032, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 2316, + "training_token_budget": 303562752, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.9359951615333557, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1338, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..88da251d86a749dd4c9e14ca1fb491c955b90ef7 Binary files /dev/null and b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/training.riglog differ diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..952b8954789579c02ece369c578e290db093239c --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +2316,303562752,fineweb,fineweb,1048576,3.9359951615333557,51.213089902791445,0.2988725209143013,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog new file mode 100644 index 0000000000000000000000000000000000000000..aebd1dd9fb506397735cc04a36a4ba6b74e47743 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/diagnostics.riglog @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c1329f4bccc27326e06964359f174798665dbda61a31a5c9c259d755bb28132 +size 259084 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..f43eb532e9c4899f70c97c845f2600c42e52b26a --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:363ffd6b00598a4b4b41821a456342df116109384fdaf95be92738bb0bc65723 +size 251954140 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec new file mode 100644 index 0000000000000000000000000000000000000000..e274ee1580a4e6da88d14c5dc1515373b0f73d56 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/fuzzy_sparsity_lossy.rigvec @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35763aacab1c4c8b8a544aa6ad55260a21d049e135c2096d4cc83167f03e2374 +size 28115104 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/metrics.json b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b1acc533cf12488c69adadd73f2b0518fea0fe64 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/metrics.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.9170133471488953, + "perplexity": 50.25014071073495, + "scored_tokens": 1048576, + "seconds": 0.2946780801285058 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 232, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.0 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 2316, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "train_tokens": 303562752 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 995.9657301481654, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 65.19332287181169, + "diagnostic_point_count": 233, + "diagnostics_every": 10, + "estimated_total_flops": 280151604005437440, + "eval_compile_seconds": 10.52271445398219, + "final_validation_seconds": 0.2946780801285058, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 2551210573824, + "dot_general": 1607397801984, + "tpu_flash_causal_attention_bwd_dkv": 1133871366144, + "tpu_flash_causal_attention_bwd_dq": 1133871366144, + "tpu_flash_causal_attention_fwd": 1133871366144 + }, + "elementwise_per_sequence": 5532367282, + "matmul_per_sequence": 7560222474240, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 922878720, + "mfu_estimate": 0.2263558477609467, + "model_tier": "60m", + "parameter_count": 97123584, + "parameters": 97123584, + "schedule_steps": 2316, + "sparsity_diagnostic_compile_seconds": 10.412538775010034, + "sparsity_diagnostic_point_count": 233, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "tokens_per_second": 1079194.5989914746, + "tokens_processed": 303562752, + "total_compile_seconds": 124.88443722180091, + "train_compile_seconds": 38.755861120997, + "train_loss": 3.888942241668701, + "train_seconds": 281.28638920513913, + "training_data_epochs": 0.03842576318008032, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 2316, + "training_token_budget": 303562752, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.9170133471488953, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/result.json b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/result.json new file mode 100644 index 0000000000000000000000000000000000000000..b1acc533cf12488c69adadd73f2b0518fea0fe64 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/result.json @@ -0,0 +1,256 @@ +{ + "artifacts": { + "diagnostics": "diagnostics.riglog", + "fuzzy_sparsity": "fuzzy_sparsity.rigvec", + "training_curve": "training.riglog", + "validation_curve": "validation.csv" + }, + "checkpoint": null, + "contract": { + "context_preset": "8k", + "dataset_id": "fineweb-8b-gpt2", + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", + "sequence_length": 8192, + "tokenizer_id": "gpt2" + }, + "evaluations": { + "fineweb": { + "canonical": true, + "loss": 3.9170133471488953, + "perplexity": 50.25014071073495, + "scored_tokens": 1048576, + "seconds": 0.2946780801285058 + } + }, + "implementation": { + "attention_backend": "tpu_flash", + "attention_tuning": { + "key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", + "resolution_source": "heuristic", + "tiles": { + "block_kv": 512, + "block_kv_compute": 256, + "block_kv_dkv": 512, + "block_kv_dkv_compute": 256, + "block_kv_dq": 512, + "block_kv_dq_compute": 256, + "block_q": 512, + "block_q_dkv": 512, + "block_q_dkv_compute": 256, + "block_q_dq": 256 + }, + "tune_seconds": 0.0 + }, + "configuration": { + "context_preset": "8k", + "path": "dev.yaml", + "profile": "dev", + "resolved": { + "evaluation": { + "eval_batches": 8, + "final_predictions": 1048576, + "probe_predictions": 0, + "val_every": 0, + "val_probe_batches": 0 + }, + "kernels": { + "attention_backend": "tpu_flash", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048 + }, + "logging": { + "diagnostics_every": 10, + "log_every": 10, + "sparsity_diagnostics_every": 10 + }, + "model": { + "d_model": 384, + "heads": 6, + "layers": 11, + "mlp_activation": "fuzzy_topk_relu", + "mlp_mult": 16, + "mlp_top_k": 1536, + "normalization": "rms_norm", + "parameterization": "completep_fixed_tpp_v1", + "position_encoding": "rope_base_10000", + "semantic_vocab_size": 50304, + "tied_embeddings": false, + "tier": "60m", + "vocab_size": 50304 + }, + "optimizer": { + "adam_epsilon": 1e-08, + "beta1": 0.9, + "beta2": 0.95, + "effective": { + "adam_epsilon_horizon_multiplier": 1.0, + "beta1": 0.9, + "beta2": 0.95, + "global_peak_learning_rate": 0.00390625, + "weight_decay_horizon_multiplier": 1.0 + }, + "grad_clip": 0.0, + "learning_rate": 0.00390625, + "min_lr_ratio": 0.1, + "warmup_steps": 232, + "weight_decay": 0.1 + }, + "parameterization": { + "attention_scale": "inverse_head_dim", + "base_depth": 12, + "base_width": 384, + "batch_ratio": 1.0, + "depth_alpha": 1.0, + "depth_multiplier": 0.9166666666666666, + "embeddings": "untied", + "init_std": 0.02, + "ladder_data_multiplier": 1.0, + "name": "completep_fixed_tpp_v1", + "width_multiplier": 1.0 + }, + "training": { + "batch_size": 16, + "dtype": "bfloat16", + "sampling": "shuffled_epochs", + "seq_len": 8192, + "steps": 2316, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "train_tokens": 303562752 + } + }, + "schema_version": 6, + "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4" + }, + "context_preset": "8k", + "document_masking": true, + "loss_backend": "tiled", + "sparse_mlp_backend": "choicewise", + "vocab_tile_size": 2048, + "weight_decay_policy": "weights_and_embeddings_only_v2" + }, + "metrics": { + "achieved_tflops": 995.9657301481654, + "attention_tune_seconds": 0.0, + "base_learning_rate": 0.00390625, + "diagnostic_compile_seconds": 65.19332287181169, + "diagnostic_point_count": 233, + "diagnostics_every": 10, + "estimated_total_flops": 280151604005437440, + "eval_compile_seconds": 10.52271445398219, + "final_validation_seconds": 0.2946780801285058, + "flop_accounting": { + "by_site": { + "_choicewise_fuzzy_topk_mlp": 2551210573824, + "dot_general": 1607397801984, + "tpu_flash_causal_attention_bwd_dkv": 1133871366144, + "tpu_flash_causal_attention_bwd_dq": 1133871366144, + "tpu_flash_causal_attention_fwd": 1133871366144 + }, + "elementwise_per_sequence": 5532367282, + "matmul_per_sequence": 7560222474240, + "method": "traced-jaxpr", + "warnings": [] + }, + "flops_per_token": 922878720, + "mfu_estimate": 0.2263558477609467, + "model_tier": "60m", + "parameter_count": 97123584, + "parameters": 97123584, + "schedule_steps": 2316, + "sparsity_diagnostic_compile_seconds": 10.412538775010034, + "sparsity_diagnostic_point_count": 233, + "sparsity_diagnostics_every": 10, + "stop_after_step": null, + "target_tokens_per_parameter": 3.125530787661213, + "tokens_per_parameter": 3.125530787661213, + "tokens_per_second": 1079194.5989914746, + "tokens_processed": 303562752, + "total_compile_seconds": 124.88443722180091, + "train_compile_seconds": 38.755861120997, + "train_loss": 3.888942241668701, + "train_seconds": 281.28638920513913, + "training_data_epochs": 0.03842576318008032, + "training_data_sharding": "rank_disjoint_shuffled_windows", + "training_sampling": "shuffled_epochs", + "training_steps": 2316, + "training_token_budget": 303562752, + "training_usable_tokens_per_epoch": 7899979776, + "validation_loss": 3.9170133471488953, + "validation_probe_count": 0, + "validation_probe_seconds": 0.0, + "validation_tokens": 1048576 + }, + "profile": "dev", + "schema_version": 1, + "seed": 1339, + "status": "ok", + "system": { + "controller_process_index": 1, + "device_count": 16, + "device_ids": [ + 0, + 1, + 2, + 3, + 4, + 5, + 6, + 7, + 8, + 9, + 10, + 11, + 12, + 13, + 14, + 15 + ], + "device_kinds": [ + "TPU v4" + ], + "jax_version": "0.11.0", + "jaxlib_version": "0.11.0", + "libtpu_version": "0.0.44.1", + "local_device_count": 4, + "platform": "tpu", + "process_count": 4, + "process_indices": [ + 0, + 0, + 0, + 0, + 1, + 1, + 1, + 1, + 2, + 2, + 2, + 2, + 3, + 3, + 3, + 3 + ], + "python_version": "3.12.13" + }, + "track": "open" +} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/training.riglog b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/training.riglog new file mode 100644 index 0000000000000000000000000000000000000000..7489175c1965e1faf77893b31801ac16ec78d0a9 Binary files /dev/null and b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/training.riglog differ diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/validation.csv b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/validation.csv new file mode 100644 index 0000000000000000000000000000000000000000..960428d28d91395eabec2a24ea191e18366d04cc --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339/validation.csv @@ -0,0 +1,2 @@ +step,tokens_processed,kind,domain,validation_tokens,validation_loss,perplexity,validation_seconds,canonical +2316,303562752,fineweb,fineweb,1048576,3.9170133471488953,50.25014071073495,0.2946780801285058,true diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/README.md b/fuzzy-topk-sparsity-diagnostics-ladder/README.md new file mode 100644 index 0000000000000000000000000000000000000000..10c589d9fddf436ebeb19b7518f99c1ffb634276 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/README.md @@ -0,0 +1,197 @@ +# Fuzzy TopK sparsity diagnostics ladder + +Twelve completed fuzzy-TopK pretraining runs: four dense-equivalent active- +compute tiers (60M, 125M, 250M, 500M) × seeds 1337–1339. This is a mechanism +dataset for the successful `H=16D, K=4D` fuzzy arm, not another architecture +comparison. It records every stored latent's competition and positive-ReLU +activity in every transformer block on one sampled global batch at step 1, +every 10 optimizer steps, and the exact final step. + +The related [three-arm study](../fuzzy-topk-three-arm-ladder/) established the +quality result: fuzzy TopK beat dense GELU in all nine paired 60M–250M seeds, +while adding the doubly-fuzzy input selector lost that gain. This archive asks +how the outer fuzzy dictionary is actually used, and extends its observations +through 500M. + +## Exact setup and endpoints + +All runs use FineWeb 8B GPT-2 tokens (canonical manifest SHA-256 +`34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804`), +8,192-token context with document masking, global batch 16 (131,072 tokens per +step), base learning rate `2^-8`, AdamW (`β1=0.9`, `β2=0.95`, weight decay +0.1), bfloat16, 1,048,576 fixed FineWeb validation predictions, and no +checkpoint. Hardware is four processes and sixteen TPU v4 chips (`v4-32`). + +| tier | L | D | stored H | retained K | stored parameters | steps | train tokens | captures/run | raw vector/run | validation mean ± sample SD | +|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:| +| 60M | 11 | 384 | 6,144 | 1,536 | 97,123,584 | 2,316 | 303,562,752 | 233 | 240.28 MiB | 3.924961 ± 0.009860 | +| 125M | 11 | 640 | 10,240 | 2,560 | 226,753,280 | 4,656 | 610,271,232 | 467 | 802.66 MiB | 3.585716 ± 0.005590 | +| 250M | 14 | 896 | 14,336 | 3,584 | 495,053,440 | 9,402 | 1,232,338,944 | 942 | 2.82 GiB | 3.308375 ± 0.005207 | +| 500M | 16 | 1,280 | 20,480 | 5,120 | 1,072,968,960 | 19,561 | 2,563,899,392 | 1,958 | 9.56 GiB | 3.091325 ± 0.010010 | + +Exact validation loss by seed: + +| tier | seed 1337 | seed 1338 | seed 1339 | +|---|---:|---:|---:| +| 60M | 3.921874 | 3.935995 | 3.917013 | +| 125M | 3.591915 | 3.581060 | 3.584173 | +| 250M | 3.314136 | 3.304005 | 3.306984 | +| 500M | 3.093102 | 3.080546 | 3.100329 | + +The 60M–250M endpoints and all per-step scalar training records reproduce the +earlier fuzzy study exactly: all nine `training.riglog` files are byte-identical +to the same-tier, same-seed uninstrumented runs. The forward-only observer did +not perturb the optimizer trajectory. + +## What is recorded + +`fuzzy_sparsity.rigvec` is an append-only float32 tensor with logical shape +`[capture, metric, layer, stored_feature]`. The four permanent metrics are: + +- `fuzzy.winner_frequency`: fraction of sampled tokens for which a feature is + the pre-ReLU winner in its fixed group of four; +- `fuzzy.activation_frequency`: fraction for which it wins and its retained + ReLU value is strictly positive; +- `fuzzy.activation_mean`: unconditional retained-activation mean, including + zero on inactive tokens; +- `fuzzy.activation_rms`: unconditional retained-activation RMS, including + zero on inactive tokens. + +Positive activation frequency is current-batch structural data-gradient support +for the corresponding UP column and DOWN row. It does not mean the AdamW +parameter delta is sparse: momentum and weight decay can move parameters whose +current-batch data gradient is zero. + +The twelve authoritative vector logs contain 10,800 captures and +43,152,550,752 bytes (40.19 GiB). Each run also has +`fuzzy_sparsity_lossy.rigvec`, which keeps every metric, layer, feature identity, +and float32 value at a reduced time grid: + +```text +step 1; every recorded 10-step capture through 200; +300, 500, 900, 1700, 3300, 6500, 12900, ...; exact final step +``` + +The later gap doubles each time. These companions retain 26, 27, 28, and 29 +captures per 60M, 125M, 250M, and 500M run and total 956,205,000 bytes +(0.891 GiB). `lossy_records.jsonl` gives their selected steps, SHA-256 hashes, +source hashes, and byte-equality verification. They are lossy only along time; +the cadence-10 originals remain authoritative. + +`snapshot.json.gz` is the 0.48 MB lightweight browser view. +`full.json.gz` is the 28.33 MB explicit full-report view. Both include the +widening-step distribution explorer; the full view additionally keeps every +cadence-10 layer summary. The explorer provides: + +- model-size, seed, and layer selectors; +- a draggable retained-step selector; +- current-batch, trailing-ten-capture, and ever-observed dead-latent heatmaps; +- an activation-frequency ridgeline through training; +- a selected-step histogram with exact zero separated from the log-binned + positive tail; +- positive activation-frequency quantiles across layers. + +## Main observation + +This is not a population that never wakes up. Almost every stored feature is +positive at least once in the step-1 global batch; the remaining few are seen +by step 10. Features become inactive later. At the final step, the fraction +inactive in each of the current and previous nine sampled batches averages: + +| tier | mean over layers and seeds | block 1 mean | current-batch mean | +|---|---:|---:|---:| +| 60M | 34.33% | 69.69% | 34.47% | +| 125M | 35.76% | 77.16% | 35.78% | +| 250M | 33.87% | 83.78% | 33.88% | +| 500M | 36.34% | 85.62% | 36.35% | + +Current-batch and trailing-window values are nearly identical at the endpoint, +so inactivity is persistent rather than a one-batch fluctuation. It is also +strongly layer-structured: block 0 remains almost fully covered, block 1 is the +most under-used, and later blocks progressively recover activity. + +Competition concentrates with scale in that same early block. Mean normalized +within-group winner entropy in block 1 falls from 0.424 at 60M to 0.103 at +500M; the fraction of groups with a positive winner falls from 1.25% to 0.32%. +The previously observed fuzzy quality gain therefore coexists with stable, +highly nonuniform dictionary under-use. This dataset measures the phenomenon; +it does not test a reallocation or anti-dead-feature intervention. + +“Dead” here always refers to the sampled observer batches. A feature absent +from ten such batches could still activate elsewhere in the corpus. Conversely, +the cumulative-ever metric cannot detect later death after an early activation; +the trailing-ten-capture view is the relevant diagnostic for creation of +persistent inactivity. + +## Provenance + +Every run passes `rig verify`; `records.jsonl` preserves the exact invoked +command, dataset identity, hardware/runtime versions, source hashes, artifact +hashes, and dirty status. The trainer SHA-256 is +`8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1`, +the resolved config SHA-256 is +`35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4`, +and the shared Python tree SHA-256 is +`66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc` +for all twelve runs. + +The clean run commits are `247c0ef7523c1b120245526bfe289f06b0d72414` +and `170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d`; they differ only by the +500M study manifest/README extension, not the trainer or shared Python tree. +The final 500M seed records commit +`6d3ad351837e5b0d2e5320105b4ee4d5a250e1cb` with only `pyproject.toml` and +`uv.lock` notebook-dependency dirt. It invoked the existing `.venv/bin/python`; +trainer/shared/config hashes and recorded JAX, JAXlib, libtpu, data, and TPU +topology are unchanged. The exception is preserved, not sanitized. + +The 120-step 500M feasibility run is systems evidence only and is excluded. + +## Reading the vectors + +```python +from pathlib import Path +from huggingface_hub import hf_hub_download +from rig import vectorlog + +path = hf_hub_download( + "quintic/rig-logs", + "fuzzy-topk-sparsity-diagnostics-ladder/500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337/fuzzy_sparsity_lossy.rigvec", + repo_type="dataset", +) +log = vectorlog.read_vector_log(Path(path)) +log.steps +frequency = log.metric("fuzzy.activation_frequency") +frequency.shape # [29 captures, 16 layers, 20,480 stored features] +``` + +Replace `fuzzy_sparsity_lossy.rigvec` with `fuzzy_sparsity.rigvec` only when +the full cadence-10 history is required; the 500M source is 9.56 GiB per run. + +## Reproduction + +Use clean commit `170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d`. +The tier-specific overrides are: + +| tier | overrides after `--` | +|---|---| +| 60M | `--sparse-layers 11 --sparse-top-k 1536 --sparse-training-steps 2316` | +| 125M | `--sparse-layers 11 --sparse-top-k 2560 --sparse-training-steps 4656` | +| 250M | `--sparse-layers 14 --sparse-top-k 3584 --sparse-training-steps 9402` | +| 500M | `--sparse-layers 16 --sparse-top-k 5120 --sparse-training-steps 19561` | + +Representative 60M command (repeat seeds 1337–1339 and substitute the table +for other tiers): + +```bash +git checkout 170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d +uv run --frozen --no-sync rig run fuzzy_topk_autoencoder \ + --cluster v4-32 --profile dev --tier 60m --context 8k \ + --batch-size 16 --base-learning-rate 0.00390625 --seed 1337 \ + --checkpoint-policy none --name 60m-fuzzy-sparsity-c10-s1337 -- \ + --sparse-layers 11 --sparse-top-k 1536 \ + --sparse-training-steps 2316 --sparsity-diagnostics-every 10 +``` + +GitHub: [static findings report](https://github.com/honglu2875/rig/blob/research/fuzzy-topk-autoencoder/docs/reports/fuzzy-topk-sparsity-diagnostics.html), +[recipe and manifests](https://github.com/honglu2875/rig/tree/research/fuzzy-topk-autoencoder/recipes/fuzzy_topk_autoencoder), +and [interactive study browser](https://github.com/honglu2875/rig/blob/research/fuzzy-topk-autoencoder/docs/reports/study-browser.html). diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/full.json.gz b/fuzzy-topk-sparsity-diagnostics-ladder/full.json.gz new file mode 100644 index 0000000000000000000000000000000000000000..3d3e80422e659d8be5d8323ba214cc3b3fed82d8 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/full.json.gz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90c09af899fcf79dea231a33e13d0edcc9a4cf9e5f537c5db5686e13f8d1a847 +size 28331092 diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/lossy_records.jsonl b/fuzzy-topk-sparsity-diagnostics-ladder/lossy_records.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..d50b11a49e4377629be50a281111f024be794fdc --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/lossy_records.jsonl @@ -0,0 +1,12 @@ +{"schema_version":1,"run_id":"20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc","folder":"60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337","source":{"path":"fuzzy_sparsity.rigvec","bytes":251954140,"sha256":"158e7d03f0f88d4916f9283e0ca6a4441142fd7061dc095288a742c25dd7518b","captures":233},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":28115104,"sha256":"b78a7f0c5af38acc3c7f4d308b455591947a1677280b58260215144d9dbf7711","captures":26,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,2316]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b","folder":"60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338","source":{"path":"fuzzy_sparsity.rigvec","bytes":251954140,"sha256":"e324a586cca82df9638914c1dd4cad2ede8c83ea18441cab1b31db8742a5bfea","captures":233},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":28115104,"sha256":"9029bea602b2f4c0e94cb825a9ba309d7514bce9a408a175983afa3a89453f9c","captures":26,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,2316]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f","folder":"60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339","source":{"path":"fuzzy_sparsity.rigvec","bytes":251954140,"sha256":"363ffd6b00598a4b4b41821a456342df116109384fdaf95be92738bb0bc65723","captures":233},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":28115104,"sha256":"35763aacab1c4c8b8a544aa6ad55260a21d049e135c2096d4cc83167f03e2374","captures":26,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,2316]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763","folder":"125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337","source":{"path":"fuzzy_sparsity.rigvec","bytes":841648004,"sha256":"f50eb0698169aca5045ad10b0a3636632a7af98f4360b54d1dcdaeef91276c20","captures":467},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":48660644,"sha256":"747a2b8211685e5edcab1f62e06b0583afb571b76e54f42c2d59f462eba527c8","captures":27,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,4656]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385","folder":"125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338","source":{"path":"fuzzy_sparsity.rigvec","bytes":841648004,"sha256":"74997501b57e7d4563b194e998c974476763b7882c7377c21a059d9fbc10bb3f","captures":467},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":48660644,"sha256":"c7d84886e72b5c16e159bb5f9ce770a894b08859dc894b2d7ec1e9c750fb5f9b","captures":27,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,4656]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79","folder":"125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339","source":{"path":"fuzzy_sparsity.rigvec","bytes":841648004,"sha256":"af03bbd8c3aa08f7f406368d502c226bf71b24f9be1d43eff74af7c7dc5a081f","captures":467},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":48660644,"sha256":"1f1f249d2f9c878bacc42332ab759f4f308f1c1ffbe872a0543ccaa3fd88b16e","captures":27,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,4656]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a","folder":"250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337","source":{"path":"fuzzy_sparsity.rigvec","bytes":3025014512,"sha256":"9a300d96d0c1a4bc069975ce48f5fbca9c3e0e94fd86110360cc23e05a4e516d","captures":942},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":89915560,"sha256":"f88d3aba739b15e5f9481deee7081550aa183bc6683f01192c4cc72eacf5119f","captures":28,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,6500,9402]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be","folder":"250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338","source":{"path":"fuzzy_sparsity.rigvec","bytes":3025014512,"sha256":"228bb4a2438541a9a3cbc86ed3d3d9383279cf361af9c4c7b3566f20669427c7","captures":942},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":89915560,"sha256":"2c6eb8512c8cb5867d9390626a5d99a60d515bdb5ac4518d8b365f9e3d6ce9c7","captures":28,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,6500,9402]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264","folder":"250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339","source":{"path":"fuzzy_sparsity.rigvec","bytes":3025014512,"sha256":"822c4975de87808dd44684e2cc6f52650e7da926b76624a9bce0284bc87ddaed","captures":942},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":89915560,"sha256":"0f15f58782ff3ee42fbf1208790f45f2a95dc054ad36ab583aafd4b28a35e2c4","captures":28,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,6500,9402]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2","folder":"500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337","source":{"path":"fuzzy_sparsity.rigvec","bytes":10265566928,"sha256":"9e982261923a00d3cd5962a7adcff671c2600400cff132fdeb3d6660383046ea","captures":1958},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":152043692,"sha256":"a0c72cb5b650b5426e55a6b38c53a288611413020d62e00aefb8f5256fad31a7","captures":29,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,6500,12900,19561]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841","folder":"500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338","source":{"path":"fuzzy_sparsity.rigvec","bytes":10265566928,"sha256":"05bf65f865e67c9fad86c291ee942ab6b6f39cbe3286a1e6eeb15f23586ca13e","captures":1958},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":152043692,"sha256":"29345feda05f0a5708cf8cb21656f15987f21a6c78ece2bd66cb1054cf43ac35","captures":29,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,6500,12900,19561]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} +{"schema_version":1,"run_id":"20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5","folder":"500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339","source":{"path":"fuzzy_sparsity.rigvec","bytes":10265566928,"sha256":"82f3cfc8d254e6eebb77330c07b394fe8385db90aec5252d4b362177639db7e0","captures":1958},"lossy":{"path":"fuzzy_sparsity_lossy.rigvec","bytes":152043692,"sha256":"523f999acff33b5c3e925db4fd5658321b4485faa99c4076c10686b7fe21804e","captures":29,"steps":[1,10,20,30,40,50,60,70,80,90,100,110,120,130,140,150,160,170,180,190,200,300,500,900,1700,3300,6500,12900,19561]},"selection":{"faithful_through":200,"first_gap":100,"rule":"all recorded captures through 200; targets 300, 500, 900, 1700, ... with doubled gaps; exact final capture"},"verification":"every retained float32 metric/layer/feature value is byte-equal to the source record"} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/records.jsonl b/fuzzy-topk-sparsity-diagnostics-ladder/records.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..02d9ed5c4532b4d8b3aa6563c313ab86a3a40834 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/records.jsonl @@ -0,0 +1,12 @@ +{"artifacts": {"diagnostics": {"bytes": 259084, "path": "diagnostics.riglog", "sha256": "06bb668b3646f7e7630d6c3c420a6dd16c67a2d9eb7517fb1ec0a96a02e33fea"}, "fuzzy_sparsity": {"bytes": 251954140, "path": "fuzzy_sparsity.rigvec", "sha256": "158e7d03f0f88d4916f9283e0ca6a4441142fd7061dc095288a742c25dd7518b"}, "training_curve": {"bytes": 37160, "path": "training.riglog", "sha256": "4bb0ccc20560ad443e0457265f85eee6becc012fdc1f1686be706ba9b71f68d8"}, "validation_curve": {"bytes": 208, "path": "validation.csv", "sha256": "468d3836e5ce147353d8b69337a76917b1ca49f450e7043402c72d87a3855c07"}}, "checkpoint": null, "cohort": {"cohort_id": "c404b54642dde8102f6155f732b8de5b8b9c72feb740f6d34f8d2182acb1297f", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 97123584, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "3.12553078766121"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "60m"}, "cohort_id": "c404b54642dde8102f6155f732b8de5b8b9c72feb740f6d34f8d2182acb1297f", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc --seed 1337 --profile dev --omit-checkpoint --tier 60m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 11 --sparse-top-k 1536 --sparse-training-steps 2316 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 303562752, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 384, "heads": 6, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 1536, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "60m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.9218744337558746, "perplexity": 50.49500566944533, "scored_tokens": 1048576, "seconds": 0.3247907080221921}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 384, "heads": 6, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 1536, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "60m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 232, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 0.9166666666666666, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 1.0}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 2316, "target_tokens_per_parameter": 3.125530787661213, "tokens_per_parameter": 3.125530787661213, "train_tokens": 303562752}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "b1db3f303c13db8d4adf96607e53ef09deab14fce7b4ade1132475575290ec7e", "stdout": "stdout.log", "stdout_sha256": "b3739c28459a5f739ed54213d1430cc5fdb0f245fcf9a68962424ce86f8cd770"}, "metrics": {"achieved_tflops": 990.6911316196339, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 60.6359039300587, "diagnostic_point_count": 233, "diagnostics_every": 10, "estimated_total_flops": 280151604005437440, "eval_compile_seconds": 10.184669547015801, "final_validation_seconds": 0.3247907080221921, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 2551210573824, "dot_general": 1607397801984, "tpu_flash_causal_attention_bwd_dkv": 1133871366144, "tpu_flash_causal_attention_bwd_dq": 1133871366144, "tpu_flash_causal_attention_fwd": 1133871366144}, "elementwise_per_sequence": 5532367282, "matmul_per_sequence": 7560222474240, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 922878720, "mfu_estimate": 0.22515707536809862, "model_tier": "60m", "parameter_count": 97123584, "parameters": 97123584, "schedule_steps": 2316, "sparsity_diagnostic_compile_seconds": 10.30694063496776, "sparsity_diagnostic_point_count": 233, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 3.125530787661213, "tokens_per_parameter": 3.125530787661213, "tokens_per_second": 1073479.2233801142, "tokens_processed": 303562752, "total_compile_seconds": 119.93457057909109, "train_compile_seconds": 38.807056467048824, "train_loss": 3.8996334075927734, "train_seconds": 282.78400307008997, "training_data_epochs": 0.03842576318008032, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 2316, "training_token_budget": 303562752, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.9218744337558746, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337", "plan": {"achieved_tokens_per_parameter": 3.125530787661213, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 97123584, "document_masking": true, "expected_tokens": 303562752, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 303562752, "profile": "dev", "run_kind": "full", "schedule_steps": 2316, "schema_version": 3, "sequence_length": 8192, "sha256": "58a539992adca00d8630e63c5d5f7b8f25c9256278ba348c209da735f51dfeee", "stop_after_step": null, "target_tokens_per_parameter": 3.125530787661213, "tier": "60m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "247c0ef7523c1b120245526bfe289f06b0d72414", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc", "run_kind": "full", "seed": 1337, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T01:07:05.347531+00:00", "started_at": "2026-08-28T01:00:14.077417+00:00"}, "timing": {"observed_wall_seconds": 411.27010611607693}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T010014.076667Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1337-910ebbcc", "--seed", "1337", "--profile", "dev", "--omit-checkpoint", "--tier", "60m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "11", "--sparse-top-k", "1536", "--sparse-training-steps", "2316", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337"} +{"artifacts": {"diagnostics": {"bytes": 259084, "path": "diagnostics.riglog", "sha256": "0d66db2838496c252f36163cb7e8776effa82d0622750d0a3c4f01d3dc1f9641"}, "fuzzy_sparsity": {"bytes": 251954140, "path": "fuzzy_sparsity.rigvec", "sha256": "e324a586cca82df9638914c1dd4cad2ede8c83ea18441cab1b31db8742a5bfea"}, "training_curve": {"bytes": 37160, "path": "training.riglog", "sha256": "5b08334b58eb4be95245cbb6b760f941baca7de6d5f81d641217d61ccf41756e"}, "validation_curve": {"bytes": 209, "path": "validation.csv", "sha256": "e59fcd92c462897e332084238fd1544ef34e9f37c762cd20f4ed269f131158bf"}}, "checkpoint": null, "cohort": {"cohort_id": "c404b54642dde8102f6155f732b8de5b8b9c72feb740f6d34f8d2182acb1297f", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 97123584, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "3.12553078766121"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "60m"}, "cohort_id": "c404b54642dde8102f6155f732b8de5b8b9c72feb740f6d34f8d2182acb1297f", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b --seed 1338 --profile dev --omit-checkpoint --tier 60m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 11 --sparse-top-k 1536 --sparse-training-steps 2316 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 303562752, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 384, "heads": 6, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 1536, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "60m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.9359951615333557, "perplexity": 51.213089902791445, "scored_tokens": 1048576, "seconds": 0.2988725209143013}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 384, "heads": 6, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 1536, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "60m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 232, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 0.9166666666666666, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 1.0}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 2316, "target_tokens_per_parameter": 3.125530787661213, "tokens_per_parameter": 3.125530787661213, "train_tokens": 303562752}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "8fe8571704bf5aac615d3af7053f3a6f8efe3dd0cd08ae7dc50c4a5b3a4478ad", "stdout": "stdout.log", "stdout_sha256": "affa55d89a547bdc5a78d66862c2799e319341338fa122bfb249a1c71e6d1d25"}, "metrics": {"achieved_tflops": 993.3663244029677, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 62.959355283062905, "diagnostic_point_count": 233, "diagnostics_every": 10, "estimated_total_flops": 280151604005437440, "eval_compile_seconds": 9.965425219153985, "final_validation_seconds": 0.2988725209143013, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 2551210573824, "dot_general": 1607397801984, "tpu_flash_causal_attention_bwd_dkv": 1133871366144, "tpu_flash_causal_attention_bwd_dq": 1133871366144, "tpu_flash_causal_attention_fwd": 1133871366144}, "elementwise_per_sequence": 5532367282, "matmul_per_sequence": 7560222474240, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 922878720, "mfu_estimate": 0.2257650737279472, "model_tier": "60m", "parameter_count": 97123584, "parameters": 97123584, "schedule_steps": 2316, "sparsity_diagnostic_compile_seconds": 10.03020230308175, "sparsity_diagnostic_point_count": 233, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 3.125530787661213, "tokens_per_parameter": 3.125530787661213, "tokens_per_second": 1076377.971314549, "tokens_processed": 303562752, "total_compile_seconds": 122.27075585327111, "train_compile_seconds": 39.31577304797247, "train_loss": 4.031710147857666, "train_seconds": 282.0224494461436, "training_data_epochs": 0.03842576318008032, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 2316, "training_token_budget": 303562752, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.9359951615333557, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338", "plan": {"achieved_tokens_per_parameter": 3.125530787661213, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 97123584, "document_masking": true, "expected_tokens": 303562752, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 303562752, "profile": "dev", "run_kind": "full", "schedule_steps": 2316, "schema_version": 3, "sequence_length": 8192, "sha256": "58a539992adca00d8630e63c5d5f7b8f25c9256278ba348c209da735f51dfeee", "stop_after_step": null, "target_tokens_per_parameter": 3.125530787661213, "tier": "60m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "247c0ef7523c1b120245526bfe289f06b0d72414", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b", "run_kind": "full", "seed": 1338, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T01:14:13.328585+00:00", "started_at": "2026-08-28T01:07:20.730881+00:00"}, "timing": {"observed_wall_seconds": 412.5976973278448}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T010720.730293Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1338-cdd37f5b", "--seed", "1338", "--profile", "dev", "--omit-checkpoint", "--tier", "60m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "11", "--sparse-top-k", "1536", "--sparse-training-steps", "2316", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338"} +{"artifacts": {"diagnostics": {"bytes": 259084, "path": "diagnostics.riglog", "sha256": "6c1329f4bccc27326e06964359f174798665dbda61a31a5c9c259d755bb28132"}, "fuzzy_sparsity": {"bytes": 251954140, "path": "fuzzy_sparsity.rigvec", "sha256": "363ffd6b00598a4b4b41821a456342df116109384fdaf95be92738bb0bc65723"}, "training_curve": {"bytes": 37160, "path": "training.riglog", "sha256": "572543f35348f0c57ae84df683aa7f3c862168639e8ce4e44af1968c2762d388"}, "validation_curve": {"bytes": 208, "path": "validation.csv", "sha256": "220536d26a131076cf932f2f7751019c9f9960d4e75719901f372367d04941ce"}}, "checkpoint": null, "cohort": {"cohort_id": "c404b54642dde8102f6155f732b8de5b8b9c72feb740f6d34f8d2182acb1297f", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 97123584, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "3.12553078766121"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "60m"}, "cohort_id": "c404b54642dde8102f6155f732b8de5b8b9c72feb740f6d34f8d2182acb1297f", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f --seed 1339 --profile dev --omit-checkpoint --tier 60m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 11 --sparse-top-k 1536 --sparse-training-steps 2316 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 303562752, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 384, "heads": 6, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 1536, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "60m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.9170133471488953, "perplexity": 50.25014071073495, "scored_tokens": 1048576, "seconds": 0.2946780801285058}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "2945ec93b8930bc2d6c9f44ec094f261987b0041d7dfb3bb4c76d8b2fb00dbb5", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 384, "heads": 6, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 1536, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "60m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 232, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 0.9166666666666666, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 1.0}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 2316, "target_tokens_per_parameter": 3.125530787661213, "tokens_per_parameter": 3.125530787661213, "train_tokens": 303562752}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "b4a50c25c5998d1fd9aa7332665eaae62b514fecabb2ac7eea0bcff9f449a4af", "stdout": "stdout.log", "stdout_sha256": "4ebf4ad6ec712c8a4ffc037ca4001ce975fbd0efb2faef209babb71ee0cbab0d"}, "metrics": {"achieved_tflops": 995.9657301481654, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 65.19332287181169, "diagnostic_point_count": 233, "diagnostics_every": 10, "estimated_total_flops": 280151604005437440, "eval_compile_seconds": 10.52271445398219, "final_validation_seconds": 0.2946780801285058, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 2551210573824, "dot_general": 1607397801984, "tpu_flash_causal_attention_bwd_dkv": 1133871366144, "tpu_flash_causal_attention_bwd_dq": 1133871366144, "tpu_flash_causal_attention_fwd": 1133871366144}, "elementwise_per_sequence": 5532367282, "matmul_per_sequence": 7560222474240, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 922878720, "mfu_estimate": 0.2263558477609467, "model_tier": "60m", "parameter_count": 97123584, "parameters": 97123584, "schedule_steps": 2316, "sparsity_diagnostic_compile_seconds": 10.412538775010034, "sparsity_diagnostic_point_count": 233, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 3.125530787661213, "tokens_per_parameter": 3.125530787661213, "tokens_per_second": 1079194.5989914746, "tokens_processed": 303562752, "total_compile_seconds": 124.88443722180091, "train_compile_seconds": 38.755861120997, "train_loss": 3.888942241668701, "train_seconds": 281.28638920513913, "training_data_epochs": 0.03842576318008032, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 2316, "training_token_budget": 303562752, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.9170133471488953, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339", "plan": {"achieved_tokens_per_parameter": 3.125530787661213, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 97123584, "document_masking": true, "expected_tokens": 303562752, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 303562752, "profile": "dev", "run_kind": "full", "schedule_steps": 2316, "schema_version": 3, "sequence_length": 8192, "sha256": "58a539992adca00d8630e63c5d5f7b8f25c9256278ba348c209da735f51dfeee", "stop_after_step": null, "target_tokens_per_parameter": 3.125530787661213, "tier": "60m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "247c0ef7523c1b120245526bfe289f06b0d72414", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f", "run_kind": "full", "seed": 1339, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T01:21:24.179431+00:00", "started_at": "2026-08-28T01:14:28.637664+00:00"}, "timing": {"observed_wall_seconds": 415.54175907187164}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T011428.636932Z-fuzzy_topk_autoencoder-60m-fuzzy-sparsity-c10-s1339-7f8b426f", "--seed", "1339", "--profile", "dev", "--omit-checkpoint", "--tier", "60m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "11", "--sparse-top-k", "1536", "--sparse-training-steps", "2316", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "60m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339"} +{"artifacts": {"diagnostics": {"bytes": 512740, "path": "diagnostics.riglog", "sha256": "c76ce3a3561c9ec8dacc95382e027cc5b0d2058e5db409635a906d9565340b96"}, "fuzzy_sparsity": {"bytes": 841648004, "path": "fuzzy_sparsity.rigvec", "sha256": "f50eb0698169aca5045ad10b0a3636632a7af98f4360b54d1dcdaeef91276c20"}, "training_curve": {"bytes": 74600, "path": "training.riglog", "sha256": "41771c68b55c66ac11085317755bb65e82514a1064c5a67eb7e270c895349cee"}, "validation_curve": {"bytes": 206, "path": "validation.csv", "sha256": "1516ce41e982fe1f8ff11f25fc08ff9bab7f4ce41245aa7b170cc2328bf1da3b"}}, "checkpoint": null, "cohort": {"cohort_id": "11edc5d4efe7c46dcc8792e7dd8ccf2cb8dad9f361b741ead99ed11259c05efe", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 226753280, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.69134467205943"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "125m"}, "cohort_id": "11edc5d4efe7c46dcc8792e7dd8ccf2cb8dad9f361b741ead99ed11259c05efe", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763 --seed 1337 --profile dev --omit-checkpoint --tier 125m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 11 --sparse-top-k 2560 --sparse-training-steps 4656 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 610271232, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 640, "heads": 10, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 2560, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "125m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.591915398836136, "perplexity": 36.30354514272369, "scored_tokens": 1048576, "seconds": 0.517010869923979}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 640, "heads": 10, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 2560, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "125m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 466, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 0.9166666666666666, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 1.6666666666666667}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 4656, "target_tokens_per_parameter": 2.6913446720594294, "tokens_per_parameter": 2.6913446720594294, "train_tokens": 610271232}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "c9eff5154bf07f1eb87154f41187b06fc4a646eae023298869ce0ee5229e6d4b", "stdout": "stdout.log", "stdout_sha256": "39a9c10501d4583173017fed90538751570f9860c4263919846fcaa998b2b7e8"}, "metrics": {"achieved_tflops": 1084.1801071023378, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 64.66830940404907, "diagnostic_point_count": 467, "diagnostics_every": 10, "estimated_total_flops": 1176245951036129280, "eval_compile_seconds": 13.15049199713394, "final_validation_seconds": 0.517010869923979, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 7086696038400, "dot_general": 3033331138560, "tpu_flash_causal_attention_bwd_dkv": 1889785610240, "tpu_flash_causal_attention_bwd_dq": 1889785610240, "tpu_flash_causal_attention_fwd": 1889785610240}, "elementwise_per_sequence": 6690029490, "matmul_per_sequence": 15789384007680, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 1927415040, "mfu_estimate": 0.24640456979598588, "model_tier": "125m", "parameter_count": 226753280, "parameters": 226753280, "schedule_steps": 4656, "sparsity_diagnostic_compile_seconds": 12.556366964941844, "sparsity_diagnostic_point_count": 467, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.6913446720594294, "tokens_per_parameter": 2.6913446720594294, "tokens_per_second": 562504.7457875693, "tokens_processed": 610271232, "total_compile_seconds": 155.68586288695224, "train_compile_seconds": 65.31069452082738, "train_loss": 3.641507625579834, "train_seconds": 1084.9174812659621, "training_data_epochs": 0.07724972079725992, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 4656, "training_token_budget": 610271232, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.591915398836136, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337", "plan": {"achieved_tokens_per_parameter": 2.6913446720594294, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 226753280, "document_masking": true, "expected_tokens": 610271232, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 610271232, "profile": "dev", "run_kind": "full", "schedule_steps": 4656, "schema_version": 3, "sequence_length": 8192, "sha256": "9a5613bd2b0b71f81d981d8e7a596ce07ab7a616aac44d87160b6ed1d138ccaf", "stop_after_step": null, "target_tokens_per_parameter": 2.6913446720594294, "tier": "125m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "247c0ef7523c1b120245526bfe289f06b0d72414", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763", "run_kind": "full", "seed": 1337, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T01:42:34.258307+00:00", "started_at": "2026-08-28T01:21:39.585107+00:00"}, "timing": {"observed_wall_seconds": 1254.6731929120142}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T012139.584469Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1337-304a7763", "--seed", "1337", "--profile", "dev", "--omit-checkpoint", "--tier", "125m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "11", "--sparse-top-k", "2560", "--sparse-training-steps", "4656", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1337"} +{"artifacts": {"diagnostics": {"bytes": 512740, "path": "diagnostics.riglog", "sha256": "d61d77acd38a1d2ca4d21dc07a2f69fa9f659ac47b707966f8b494d4b536a190"}, "fuzzy_sparsity": {"bytes": 841648004, "path": "fuzzy_sparsity.rigvec", "sha256": "74997501b57e7d4563b194e998c974476763b7882c7377c21a059d9fbc10bb3f"}, "training_curve": {"bytes": 74600, "path": "training.riglog", "sha256": "feae8557a321336d87f82b9e889554fa3d2984520959e5fe8a888d8fca8e84ca"}, "validation_curve": {"bytes": 207, "path": "validation.csv", "sha256": "6ffc0647f2f6eb84a621fd4ca33be58831f43b66e001dc9dd7035e3e627e9c1e"}}, "checkpoint": null, "cohort": {"cohort_id": "11edc5d4efe7c46dcc8792e7dd8ccf2cb8dad9f361b741ead99ed11259c05efe", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 226753280, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.69134467205943"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "125m"}, "cohort_id": "11edc5d4efe7c46dcc8792e7dd8ccf2cb8dad9f361b741ead99ed11259c05efe", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385 --seed 1338 --profile dev --omit-checkpoint --tier 125m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 11 --sparse-top-k 2560 --sparse-training-steps 4656 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 610271232, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 640, "heads": 10, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 2560, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "125m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.581060439348221, "perplexity": 35.91160273893408, "scored_tokens": 1048576, "seconds": 0.5136202480643988}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 640, "heads": 10, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 2560, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "125m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 466, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 0.9166666666666666, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 1.6666666666666667}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 4656, "target_tokens_per_parameter": 2.6913446720594294, "tokens_per_parameter": 2.6913446720594294, "train_tokens": 610271232}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "dff1c35efb0a07b50fb742fdce5f0bc387fbcbc1a64216136e40685fd3e2e837", "stdout": "stdout.log", "stdout_sha256": "c69ff9d018599305d08be8596716e6905217cc84d56695eacbeb6c0b25c69dc2"}, "metrics": {"achieved_tflops": 1083.86685823526, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 64.37414500606246, "diagnostic_point_count": 467, "diagnostics_every": 10, "estimated_total_flops": 1176245951036129280, "eval_compile_seconds": 13.222909380914643, "final_validation_seconds": 0.5136202480643988, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 7086696038400, "dot_general": 3033331138560, "tpu_flash_causal_attention_bwd_dkv": 1889785610240, "tpu_flash_causal_attention_bwd_dq": 1889785610240, "tpu_flash_causal_attention_fwd": 1889785610240}, "elementwise_per_sequence": 6690029490, "matmul_per_sequence": 15789384007680, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 1927415040, "mfu_estimate": 0.24633337687165, "model_tier": "125m", "parameter_count": 226753280, "parameters": 226753280, "schedule_steps": 4656, "sparsity_diagnostic_compile_seconds": 11.524073926033452, "sparsity_diagnostic_point_count": 467, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.6913446720594294, "tokens_per_parameter": 2.6913446720594294, "tokens_per_second": 562342.2229989759, "tokens_processed": 610271232, "total_compile_seconds": 154.58197558182292, "train_compile_seconds": 65.46084726881236, "train_loss": 3.6206488609313965, "train_seconds": 1085.2310337740928, "training_data_epochs": 0.07724972079725992, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 4656, "training_token_budget": 610271232, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.581060439348221, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338", "plan": {"achieved_tokens_per_parameter": 2.6913446720594294, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 226753280, "document_masking": true, "expected_tokens": 610271232, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 610271232, "profile": "dev", "run_kind": "full", "schedule_steps": 4656, "schema_version": 3, "sequence_length": 8192, "sha256": "9a5613bd2b0b71f81d981d8e7a596ce07ab7a616aac44d87160b6ed1d138ccaf", "stop_after_step": null, "target_tokens_per_parameter": 2.6913446720594294, "tier": "125m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385", "run_kind": "full", "seed": 1338, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T02:03:42.958639+00:00", "started_at": "2026-08-28T01:42:50.137850+00:00"}, "timing": {"observed_wall_seconds": 1252.8207831680775}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T014250.137281Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1338-fa728385", "--seed", "1338", "--profile", "dev", "--omit-checkpoint", "--tier", "125m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "11", "--sparse-top-k", "2560", "--sparse-training-steps", "4656", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1338"} +{"artifacts": {"diagnostics": {"bytes": 512740, "path": "diagnostics.riglog", "sha256": "62b57e17b6624f065ac0b0fb5dcb7a7fc7c2b9d3670f2faec885a0538fcbbe09"}, "fuzzy_sparsity": {"bytes": 841648004, "path": "fuzzy_sparsity.rigvec", "sha256": "af03bbd8c3aa08f7f406368d502c226bf71b24f9be1d43eff74af7c7dc5a081f"}, "training_curve": {"bytes": 74600, "path": "training.riglog", "sha256": "83f31b0db8d875099ce248d0796cc7acaeda6c4afd39c2dc2628acda974a5bc2"}, "validation_curve": {"bytes": 208, "path": "validation.csv", "sha256": "75347e08a538b532ef3a6b4d9d474a4344d0eaee35f14527e07d7828872221f1"}}, "checkpoint": null, "cohort": {"cohort_id": "11edc5d4efe7c46dcc8792e7dd8ccf2cb8dad9f361b741ead99ed11259c05efe", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 226753280, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.69134467205943"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "125m"}, "cohort_id": "11edc5d4efe7c46dcc8792e7dd8ccf2cb8dad9f361b741ead99ed11259c05efe", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79 --seed 1339 --profile dev --omit-checkpoint --tier 125m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 11 --sparse-top-k 2560 --sparse-training-steps 4656 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 610271232, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 640, "heads": 10, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 2560, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "125m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.584172636270523, "perplexity": 36.023540814770165, "scored_tokens": 1048576, "seconds": 0.5274990880861878}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "3d77f3dec34a89456ba8284f3d7d149016d9913126698de06073a72e1ff1a550", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 640, "heads": 10, "layers": 11, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 2560, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "125m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 466, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 0.9166666666666666, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 1.6666666666666667}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 4656, "target_tokens_per_parameter": 2.6913446720594294, "tokens_per_parameter": 2.6913446720594294, "train_tokens": 610271232}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "da1bc21d72b82cdd70ea16a5c795e3c88a8c1d1f2fda05ebb5f3d928e2018288", "stdout": "stdout.log", "stdout_sha256": "38da40ec26ee2168e0f12db48186fe51e7c23e5c1b9d2e6ddf0008f3233a56b6"}, "metrics": {"achieved_tflops": 1083.9200870397517, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 63.24943733587861, "diagnostic_point_count": 467, "diagnostics_every": 10, "estimated_total_flops": 1176245951036129280, "eval_compile_seconds": 12.438345218077302, "final_validation_seconds": 0.5274990880861878, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 7086696038400, "dot_general": 3033331138560, "tpu_flash_causal_attention_bwd_dkv": 1889785610240, "tpu_flash_causal_attention_bwd_dq": 1889785610240, "tpu_flash_causal_attention_fwd": 1889785610240}, "elementwise_per_sequence": 6690029490, "matmul_per_sequence": 15789384007680, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 1927415040, "mfu_estimate": 0.2463454743272163, "model_tier": "125m", "parameter_count": 226753280, "parameters": 226753280, "schedule_steps": 4656, "sparsity_diagnostic_compile_seconds": 11.386971961008385, "sparsity_diagnostic_point_count": 467, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.6913446720594294, "tokens_per_parameter": 2.6913446720594294, "tokens_per_second": 562369.8396790302, "tokens_processed": 610271232, "total_compile_seconds": 135.26415866799653, "train_compile_seconds": 48.18940415303223, "train_loss": 3.5505714416503906, "train_seconds": 1085.1777405920438, "training_data_epochs": 0.07724972079725992, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 4656, "training_token_budget": 610271232, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.584172636270523, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339", "plan": {"achieved_tokens_per_parameter": 2.6913446720594294, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 226753280, "document_masking": true, "expected_tokens": 610271232, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 610271232, "profile": "dev", "run_kind": "full", "schedule_steps": 4656, "schema_version": 3, "sequence_length": 8192, "sha256": "9a5613bd2b0b71f81d981d8e7a596ce07ab7a616aac44d87160b6ed1d138ccaf", "stop_after_step": null, "target_tokens_per_parameter": 2.6913446720594294, "tier": "125m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79", "run_kind": "full", "seed": 1339, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T02:24:33.583517+00:00", "started_at": "2026-08-28T02:03:58.757293+00:00"}, "timing": {"observed_wall_seconds": 1234.8262156671844}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T020358.756671Z-fuzzy_topk_autoencoder-125m-fuzzy-sparsity-c10-s1339-3cb38c79", "--seed", "1339", "--profile", "dev", "--omit-checkpoint", "--tier", "125m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "11", "--sparse-top-k", "2560", "--sparse-training-steps", "4656", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "125m-fuzzy-h16d-k4d-l11-3tpp-bs16-lr2e-8-s1339"} +{"artifacts": {"diagnostics": {"bytes": 1232408, "path": "diagnostics.riglog", "sha256": "f6c350e8a3a3b73522c617c30bad2dfcd13c42ee48ec6fe6150300deba4ba0da"}, "fuzzy_sparsity": {"bytes": 3025014512, "path": "fuzzy_sparsity.rigvec", "sha256": "9a300d96d0c1a4bc069975ce48f5fbca9c3e0e94fd86110360cc23e05a4e516d"}, "training_curve": {"bytes": 150536, "path": "training.riglog", "sha256": "34e549a8d940eb2afcd495773e93aa9a2071943cca85bce4736980d9898e670d"}, "validation_curve": {"bytes": 209, "path": "validation.csv", "sha256": "2c1acf6f5d7affdf85cb598f2ad9525d9780c2c78891e7285f1ac39517a6f972"}}, "checkpoint": null, "cohort": {"cohort_id": "243d9a1fd97b4c1c498bdfe0ceff30b72639bea95ad20f4d080641da41c2bfe1", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 495053440, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.48930487989337"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "250m"}, "cohort_id": "243d9a1fd97b4c1c498bdfe0ceff30b72639bea95ad20f4d080641da41c2bfe1", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a --seed 1337 --profile dev --omit-checkpoint --tier 250m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 14 --sparse-top-k 3584 --sparse-training-steps 9402 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 1232338944, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 896, "heads": 14, "layers": 14, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 3584, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "250m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.314135879278183, "perplexity": 27.498621586202635, "scored_tokens": 1048576, "seconds": 0.9843828401062638}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 896, "heads": 14, "layers": 14, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 3584, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "250m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 940, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 1.1666666666666667, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 2.3333333333333335}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 9402, "target_tokens_per_parameter": 2.489304879893371, "tokens_per_parameter": 2.489304879893371, "train_tokens": 1232338944}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "544151d94033be4011ee773560ab5825e6f87add5277f6a87ce222a86bfcd63a", "stdout": "stdout.log", "stdout_sha256": "ddb546d8fc451d019409a4e07b86ebc453f36d154b9f8ccc2a346dd2efc233d3"}, "metrics": {"achieved_tflops": 1122.39495655022, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 76.33016821881756, "diagnostic_point_count": 942, "diagnostics_every": 10, "estimated_total_flops": 4963669455340044288, "eval_compile_seconds": 12.449057026999071, "final_validation_seconds": 0.9843828401062638, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 17678085390336, "dot_general": 5216252461056, "tpu_flash_causal_attention_bwd_dkv": 3367254360064, "tpu_flash_causal_attention_bwd_dq": 3367254360064, "tpu_flash_causal_attention_fwd": 3367254360064}, "elementwise_per_sequence": 8887879538, "matmul_per_sequence": 32996100931584, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 4027844352, "mfu_estimate": 0.25508976285232277, "model_tier": "250m", "parameter_count": 495053440, "parameters": 495053440, "schedule_steps": 9402, "sparsity_diagnostic_compile_seconds": 14.994110438041389, "sparsity_diagnostic_point_count": 942, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.489304879893371, "tokens_per_parameter": 2.489304879893371, "tokens_per_second": 278658.9695286766, "tokens_processed": 1232338944, "total_compile_seconds": 164.92245301185176, "train_compile_seconds": 61.149117327993736, "train_loss": 3.3761191368103027, "train_seconds": 4422.391090027988, "training_data_epochs": 0.15599267073364215, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 9402, "training_token_budget": 1232338944, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.314135879278183, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337", "plan": {"achieved_tokens_per_parameter": 2.489304879893371, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 495053440, "document_masking": true, "expected_tokens": 1232338944, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 1232338944, "profile": "dev", "run_kind": "full", "schedule_steps": 9402, "schema_version": 3, "sequence_length": 8192, "sha256": "15057949438217bb6e652ace1c232de456ba684396f254fc5bb6004ed10caa20", "stop_after_step": null, "target_tokens_per_parameter": 2.489304879893371, "tier": "250m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a", "run_kind": "full", "seed": 1337, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T03:41:36.466748+00:00", "started_at": "2026-08-28T02:24:49.454634+00:00"}, "timing": {"observed_wall_seconds": 4607.012098823907}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T022449.453967Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1337-9dc93f5a", "--seed", "1337", "--profile", "dev", "--omit-checkpoint", "--tier", "250m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "14", "--sparse-top-k", "3584", "--sparse-training-steps", "9402", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1337"} +{"artifacts": {"diagnostics": {"bytes": 1232408, "path": "diagnostics.riglog", "sha256": "0e90ea4b61a38334d46f127cc46689466bc48505961e625f0d3afe109c91430e"}, "fuzzy_sparsity": {"bytes": 3025014512, "path": "fuzzy_sparsity.rigvec", "sha256": "228bb4a2438541a9a3cbc86ed3d3d9383279cf361af9c4c7b3566f20669427c7"}, "training_curve": {"bytes": 150536, "path": "training.riglog", "sha256": "70c0ffdd65ec23cc21e1f21f5d919c167b7c30242c2e0c5e7ee4926fc9a64831"}, "validation_curve": {"bytes": 210, "path": "validation.csv", "sha256": "319ddd3ae17eb8a0752b1116dd96ca0285ed52036dfd24bf11c99c42f5ff10f2"}}, "checkpoint": null, "cohort": {"cohort_id": "243d9a1fd97b4c1c498bdfe0ceff30b72639bea95ad20f4d080641da41c2bfe1", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 495053440, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.48930487989337"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "250m"}, "cohort_id": "243d9a1fd97b4c1c498bdfe0ceff30b72639bea95ad20f4d080641da41c2bfe1", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be --seed 1338 --profile dev --omit-checkpoint --tier 250m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 14 --sparse-top-k 3584 --sparse-training-steps 9402 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 1232338944, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 896, "heads": 14, "layers": 14, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 3584, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "250m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.3040049374103546, "perplexity": 27.221441070036004, "scored_tokens": 1048576, "seconds": 0.9745558970607817}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 896, "heads": 14, "layers": 14, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 3584, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "250m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 940, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 1.1666666666666667, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 2.3333333333333335}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 9402, "target_tokens_per_parameter": 2.489304879893371, "tokens_per_parameter": 2.489304879893371, "train_tokens": 1232338944}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "ef3381f3cbf06fc43a543f73b7665026decfe03989b683c8c0a87e471fdcdc90", "stdout": "stdout.log", "stdout_sha256": "edabbc400260a1e29cd1611b63f81e567db09846f6929d796198da949dc9dacb"}, "metrics": {"achieved_tflops": 1122.3941557798357, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 81.67842828086577, "diagnostic_point_count": 942, "diagnostics_every": 10, "estimated_total_flops": 4963669455340044288, "eval_compile_seconds": 12.553831937955692, "final_validation_seconds": 0.9745558970607817, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 17678085390336, "dot_general": 5216252461056, "tpu_flash_causal_attention_bwd_dkv": 3367254360064, "tpu_flash_causal_attention_bwd_dq": 3367254360064, "tpu_flash_causal_attention_fwd": 3367254360064}, "elementwise_per_sequence": 8887879538, "matmul_per_sequence": 32996100931584, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 4027844352, "mfu_estimate": 0.2550895808590536, "model_tier": "250m", "parameter_count": 495053440, "parameters": 495053440, "schedule_steps": 9402, "sparsity_diagnostic_compile_seconds": 15.16223825304769, "sparsity_diagnostic_point_count": 942, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.489304879893371, "tokens_per_parameter": 2.489304879893371, "tokens_per_second": 278658.77072000515, "tokens_processed": 1232338944, "total_compile_seconds": 166.9548121576663, "train_compile_seconds": 57.560313685797155, "train_loss": 3.1848177909851074, "train_seconds": 4422.39424517611, "training_data_epochs": 0.15599267073364215, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 9402, "training_token_budget": 1232338944, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.3040049374103546, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338", "plan": {"achieved_tokens_per_parameter": 2.489304879893371, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 495053440, "document_masking": true, "expected_tokens": 1232338944, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 1232338944, "profile": "dev", "run_kind": "full", "schedule_steps": 9402, "schema_version": 3, "sequence_length": 8192, "sha256": "15057949438217bb6e652ace1c232de456ba684396f254fc5bb6004ed10caa20", "stop_after_step": null, "target_tokens_per_parameter": 2.489304879893371, "tier": "250m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be", "run_kind": "full", "seed": 1338, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T04:58:43.189161+00:00", "started_at": "2026-08-28T03:41:54.231931+00:00"}, "timing": {"observed_wall_seconds": 4608.957214114955}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T034154.231319Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1338-ba68b1be", "--seed", "1338", "--profile", "dev", "--omit-checkpoint", "--tier", "250m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "14", "--sparse-top-k", "3584", "--sparse-training-steps", "9402", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1338"} +{"artifacts": {"diagnostics": {"bytes": 1232408, "path": "diagnostics.riglog", "sha256": "0d5ec373ee087cda4d56c63d826dc29b4a042458ef3f1e72a5e83a8e2b29ee8a"}, "fuzzy_sparsity": {"bytes": 3025014512, "path": "fuzzy_sparsity.rigvec", "sha256": "822c4975de87808dd44684e2cc6f52650e7da926b76624a9bce0284bc87ddaed"}, "training_curve": {"bytes": 150536, "path": "training.riglog", "sha256": "b6062f911a5a81665bc9fdd1259b13bc63f66d91a9523161437631f1a195ffc2"}, "validation_curve": {"bytes": 209, "path": "validation.csv", "sha256": "f281786e458b96978a1ef3832c39060b9fe18da579666789a093d6cb0508c00e"}}, "checkpoint": null, "cohort": {"cohort_id": "243d9a1fd97b4c1c498bdfe0ceff30b72639bea95ad20f4d080641da41c2bfe1", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 495053440, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.48930487989337"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "250m"}, "cohort_id": "243d9a1fd97b4c1c498bdfe0ceff30b72639bea95ad20f4d080641da41c2bfe1", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264 --seed 1339 --profile dev --omit-checkpoint --tier 250m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 14 --sparse-top-k 3584 --sparse-training-steps 9402 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 1232338944, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 896, "heads": 14, "layers": 14, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 3584, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "250m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.306984007358551, "perplexity": 27.302656560322916, "scored_tokens": 1048576, "seconds": 0.9671843170654029}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "e06781c1c2978705df2ad6d078d005991aaaf8908efa50bfd764dd157326f324", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 896, "heads": 14, "layers": 14, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 3584, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "250m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 940, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 1.1666666666666667, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 2.3333333333333335}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 9402, "target_tokens_per_parameter": 2.489304879893371, "tokens_per_parameter": 2.489304879893371, "train_tokens": 1232338944}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "ea526c38fe171f381a16126d7d3db8db6e675637538636ccc67b3ca330bc5f31", "stdout": "stdout.log", "stdout_sha256": "c37b7ee85bbfb7145ed1ffcbbd28927179e0150f131c3c5ed9fc5e511fb785b2"}, "metrics": {"achieved_tflops": 1122.6806043428, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 83.23276506806724, "diagnostic_point_count": 942, "diagnostics_every": 10, "estimated_total_flops": 4963669455340044288, "eval_compile_seconds": 12.998791442951187, "final_validation_seconds": 0.9671843170654029, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 17678085390336, "dot_general": 5216252461056, "tpu_flash_causal_attention_bwd_dkv": 3367254360064, "tpu_flash_causal_attention_bwd_dq": 3367254360064, "tpu_flash_causal_attention_fwd": 3367254360064}, "elementwise_per_sequence": 8887879538, "matmul_per_sequence": 32996100931584, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 4027844352, "mfu_estimate": 0.25515468280518183, "model_tier": "250m", "parameter_count": 495053440, "parameters": 495053440, "schedule_steps": 9402, "sparsity_diagnostic_compile_seconds": 15.126094905892387, "sparsity_diagnostic_point_count": 942, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.489304879893371, "tokens_per_parameter": 2.489304879893371, "tokens_per_second": 278729.88780843536, "tokens_processed": 1232338944, "total_compile_seconds": 181.17454498284496, "train_compile_seconds": 69.81689356593415, "train_loss": 3.2581043243408203, "train_seconds": 4421.265884650871, "training_data_epochs": 0.15599267073364215, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 9402, "training_token_budget": 1232338944, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.306984007358551, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339", "plan": {"achieved_tokens_per_parameter": 2.489304879893371, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 495053440, "document_masking": true, "expected_tokens": 1232338944, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 1232338944, "profile": "dev", "run_kind": "full", "schedule_steps": 9402, "schema_version": 3, "sequence_length": 8192, "sha256": "15057949438217bb6e652ace1c232de456ba684396f254fc5bb6004ed10caa20", "stop_after_step": null, "target_tokens_per_parameter": 2.489304879893371, "tier": "250m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": false, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264", "run_kind": "full", "seed": 1339, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T06:16:02.583742+00:00", "started_at": "2026-08-28T04:59:00.979708+00:00"}, "timing": {"observed_wall_seconds": 4621.604019111022}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T045900.979160Z-fuzzy_topk_autoencoder-250m-fuzzy-sparsity-c10-s1339-f826e264", "--seed", "1339", "--profile", "dev", "--omit-checkpoint", "--tier", "250m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "14", "--sparse-top-k", "3584", "--sparse-training-steps", "9402", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "250m-fuzzy-h16d-k4d-l14-2tpp-bs16-lr2e-8-s1339"} +{"artifacts": {"diagnostics": {"bytes": 2836024, "path": "diagnostics.riglog", "sha256": "ccd0cb279b9405d22db1b4a5a9c328a93070e5670ff52711a9a693016fd2e5ea"}, "fuzzy_sparsity": {"bytes": 10265566928, "path": "fuzzy_sparsity.rigvec", "sha256": "9e982261923a00d3cd5962a7adcff671c2600400cff132fdeb3d6660383046ea"}, "training_curve": {"bytes": 313080, "path": "training.riglog", "sha256": "4e6ede524e9c5b6ee96944b98f792696f086025eadf0ca64da88cebd1a51cb6a"}, "validation_curve": {"bytes": 209, "path": "validation.csv", "sha256": "6f5b51945e8b8b71389ad4f70ed74e568d07feef8c0a16cf7d948835df6a1ed9"}}, "checkpoint": null, "cohort": {"cohort_id": "86becaef140b74d4923d2145f0a5c831ac56182778f83554ba6efdfd2be6f5c9", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 1072968960, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.38953733759456"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "500m"}, "cohort_id": "86becaef140b74d4923d2145f0a5c831ac56182778f83554ba6efdfd2be6f5c9", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2 --seed 1337 --profile dev --omit-checkpoint --tier 500m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 16 --sparse-top-k 5120 --sparse-training-steps 19561 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 2563899392, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 1280, "heads": 20, "layers": 16, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 5120, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "500m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.093101680278778, "perplexity": 22.04534966885452, "scored_tokens": 1048576, "seconds": 1.7293408529367298}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 1280, "heads": 20, "layers": 16, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 5120, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "500m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 1956, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 1.3333333333333333, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 3.3333333333333335}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 19561, "target_tokens_per_parameter": 2.389537337594556, "tokens_per_parameter": 2.389537337594556, "train_tokens": 2563899392}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "68d284db3719ec7c6cab69c1633606ee0dc39ced28cafd1d98275792f76dcc98", "stdout": "stdout.log", "stdout_sha256": "39d04b372f9c2671158e4175c56967196d0a68848e726ab2f2d02e09b26dc08f"}, "metrics": {"achieved_tflops": 1164.0399494191674, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 92.70696510304697, "diagnostic_point_count": 1958, "diagnostics_every": 10, "estimated_total_flops": 21023633708112936960, "eval_compile_seconds": 16.746170554775745, "final_validation_seconds": 1.7293408529367298, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 41231686041600, "dot_general": 9448949022720, "tpu_flash_causal_attention_bwd_dkv": 5497558138880, "tpu_flash_causal_attention_bwd_dq": 5497558138880, "tpu_flash_causal_attention_fwd": 5497558138880}, "elementwise_per_sequence": 12057100530, "matmul_per_sequence": 67173309480960, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 8199866880, "mfu_estimate": 0.2645545339589017, "model_tier": "500m", "parameter_count": 1072968960, "parameters": 1072968960, "schedule_steps": 19561, "sparsity_diagnostic_compile_seconds": 19.950090633006766, "sparsity_diagnostic_point_count": 1958, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.389537337594556, "tokens_per_parameter": 2.389537337594556, "tokens_per_second": 141958.39596595592, "tokens_processed": 2563899392, "total_compile_seconds": 203.57062846887857, "train_compile_seconds": 74.16740217804909, "train_loss": 3.0271120071411133, "train_seconds": 18060.921120966086, "training_data_epochs": 0.32454505767079067, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 19561, "training_token_budget": 2563899392, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.093101680278778, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337", "plan": {"achieved_tokens_per_parameter": 2.389537337594556, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 1072968960, "document_masking": true, "expected_tokens": 2563899392, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 2563899392, "profile": "dev", "run_kind": "full", "schedule_steps": 19561, "schema_version": 3, "sequence_length": 8192, "sha256": "2cbdd2d61dc2ea127447e7938c6fd686a20a0b6a63c42d22b6b50f34d085ff2e", "stop_after_step": null, "target_tokens_per_parameter": 2.389537337594556, "tier": "500m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": true, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2", "run_kind": "full", "seed": 1337, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T11:27:57.964800+00:00", "started_at": "2026-08-28T06:23:03.201003+00:00"}, "timing": {"observed_wall_seconds": 18294.763782619033}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T062303.200275Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1337-eed705b2", "--seed", "1337", "--profile", "dev", "--omit-checkpoint", "--tier", "500m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "16", "--sparse-top-k", "5120", "--sparse-training-steps", "19561", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1337"} +{"artifacts": {"diagnostics": {"bytes": 2836024, "path": "diagnostics.riglog", "sha256": "123ffd0c707339ffbc1cf8f1d045af8ed1f06c0b097e654f6aa1e5a6188b3268"}, "fuzzy_sparsity": {"bytes": 10265566928, "path": "fuzzy_sparsity.rigvec", "sha256": "05bf65f865e67c9fad86c291ee942ab6b6f39cbe3286a1e6eeb15f23586ca13e"}, "training_curve": {"bytes": 313080, "path": "training.riglog", "sha256": "e2faaa344826c61eec1b727a37a3035180afc61584694b464dd101ec6941675c"}, "validation_curve": {"bytes": 209, "path": "validation.csv", "sha256": "05077c043c3d5cf25922719965b619d2e5ca5d6189212c6dbcd34238d94a18c5"}}, "checkpoint": null, "cohort": {"cohort_id": "86becaef140b74d4923d2145f0a5c831ac56182778f83554ba6efdfd2be6f5c9", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 1072968960, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.38953733759456"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "500m"}, "cohort_id": "86becaef140b74d4923d2145f0a5c831ac56182778f83554ba6efdfd2be6f5c9", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841 --seed 1338 --profile dev --omit-checkpoint --tier 500m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 16 --sparse-top-k 5120 --sparse-training-steps 19561 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 2563899392, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 1280, "heads": 20, "layers": 16, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 5120, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "500m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.080546021461487, "perplexity": 21.770286194982404, "scored_tokens": 1048576, "seconds": 1.743782521924004}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 1280, "heads": 20, "layers": 16, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 5120, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "500m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 1956, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 1.3333333333333333, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 3.3333333333333335}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 19561, "target_tokens_per_parameter": 2.389537337594556, "tokens_per_parameter": 2.389537337594556, "train_tokens": 2563899392}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "a92329ced70fcc26568dda1ad8faba9bb13c24face0ed61fa737b3912cb67def", "stdout": "stdout.log", "stdout_sha256": "536cd368749643ea024aba484d7070802709be0658a92da87a6f89d9d995191c"}, "metrics": {"achieved_tflops": 1163.9397576841584, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 94.47339812200516, "diagnostic_point_count": 1958, "diagnostics_every": 10, "estimated_total_flops": 21023633708112936960, "eval_compile_seconds": 17.084043910959736, "final_validation_seconds": 1.743782521924004, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 41231686041600, "dot_general": 9448949022720, "tpu_flash_causal_attention_bwd_dkv": 5497558138880, "tpu_flash_causal_attention_bwd_dq": 5497558138880, "tpu_flash_causal_attention_fwd": 5497558138880}, "elementwise_per_sequence": 12057100530, "matmul_per_sequence": 67173309480960, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 8199866880, "mfu_estimate": 0.264531763110036, "model_tier": "500m", "parameter_count": 1072968960, "parameters": 1072968960, "schedule_steps": 19561, "sparsity_diagnostic_compile_seconds": 19.11918603302911, "sparsity_diagnostic_point_count": 1958, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.389537337594556, "tokens_per_parameter": 2.389537337594556, "tokens_per_second": 141946.1772633263, "tokens_processed": 2563899392, "total_compile_seconds": 203.83854955085553, "train_compile_seconds": 73.16192148486152, "train_loss": 3.1012890338897705, "train_seconds": 18062.475801963126, "training_data_epochs": 0.32454505767079067, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 19561, "training_token_budget": 2563899392, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.080546021461487, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338", "plan": {"achieved_tokens_per_parameter": 2.389537337594556, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 1072968960, "document_masking": true, "expected_tokens": 2563899392, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 2563899392, "profile": "dev", "run_kind": "full", "schedule_steps": 19561, "schema_version": 3, "sequence_length": 8192, "sha256": "2cbdd2d61dc2ea127447e7938c6fd686a20a0b6a63c42d22b6b50f34d085ff2e", "stop_after_step": null, "target_tokens_per_parameter": 2.389537337594556, "tier": "500m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": false, "head": "170d1f93d7b33a410ed1dc930d62c1bc8dce2d0d", "status_porcelain": "", "status_porcelain_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 35400, "path": "uv.lock", "sha256": "394baf4835b9a61f240aca357ed00b2a407f17ab2e07084fe46a6391b6032b3c"}}, "qualified": true, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841", "run_kind": "full", "seed": 1338, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T16:33:17.554417+00:00", "started_at": "2026-08-28T11:28:22.096494+00:00"}, "timing": {"observed_wall_seconds": 18295.457908394048}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T112822.095886Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1338-f4b3e841", "--seed", "1338", "--profile", "dev", "--omit-checkpoint", "--tier", "500m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "16", "--sparse-top-k", "5120", "--sparse-training-steps", "19561", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1338"} +{"artifacts": {"diagnostics": {"bytes": 2836024, "path": "diagnostics.riglog", "sha256": "af9d34d62f7f2931698edeeb3dd78253380c282409c2fccf583b1fb810a528d2"}, "fuzzy_sparsity": {"bytes": 10265566928, "path": "fuzzy_sparsity.rigvec", "sha256": "82f3cfc8d254e6eebb77330c07b394fe8385db90aec5252d4b362177639db7e0"}, "training_curve": {"bytes": 313080, "path": "training.riglog", "sha256": "006b65cd1635eafe32fa277c75b11080b855c61a382cf6d166cf355b17eadebd"}, "validation_curve": {"bytes": 210, "path": "validation.csv", "sha256": "2adb2a6d2c368d672238afb282010a05d658d0b62148cf3ebfc6e1c87aa7a046"}}, "checkpoint": null, "cohort": {"cohort_id": "86becaef140b74d4923d2145f0a5c831ac56182778f83554ba6efdfd2be6f5c9", "data": {"dataset_id": "fineweb-8b-gpt2", "manifest_canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "tokenizer_id": "gpt2", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_files_sha256": "70ac972b59e84b35e28a32db049a1c9c8a6665cf664ed8731aaa2668f0ce0def", "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760}, "declared_parameters": 1072968960, "hardware": {"accelerator": "TPU v4", "chips_per_host": 4, "total_chips": 16, "tpu_vm_count": 4}, "horizon": {"rounding": "nearest_complete_global_step", "target_tokens_per_parameter": "2.38953733759456"}, "profile": "dev", "qualification": {"target_loss": "3.28"}, "schema_version": 1, "tier": "500m"}, "cohort_id": "86becaef140b74d4923d2145f0a5c831ac56182778f83554ba6efdfd2be6f5c9", "command": ["pdsh", "-S", "-R", "ssh", "-f", "4", "-w", "t1v-n-a09f5679-w-[0-3]", "-N", "cd /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder && env JAX_COMPILATION_CACHE_DIR=/tmp/rig-jax-cache-20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5 JAX_PERSISTENT_CACHE_MIN_COMPILE_TIME_SECS=0 PYTHONUNBUFFERED=1 RIG_CLUSTER_WORKER=1 RIG_CONTROLLER_HOSTNAME=t1v-n-a09f5679-w-0 RIG_DISTRIBUTED=1 RIG_OUTPUT_DIR=/home/cubic27/GPT-speedrun-TPU/runs/20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5 RIG_PROCESS_COUNT=4 RIG_RUN_ID=20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5 /home/cubic27/GPT-speedrun-TPU/.venv/bin/python /home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py --output-dir /home/cubic27/GPT-speedrun-TPU/runs/20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5 --seed 1339 --profile dev --omit-checkpoint --tier 500m --context 8k --base-learning-rate 0.00390625 --batch-size 16 --sparse-layers 16 --sparse-top-k 5120 --sparse-training-steps 19561 --sparsity-diagnostics-every 10 --data-format llmc --dataset-id fineweb-8b-gpt2 --tokenizer-id gpt2 --color never --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin --train-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin --val-data /dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "constraints": {"training_tokens": 2563899392, "validation_tokens": null}, "contract": {"context_preset": "8k", "dataset_id": "fineweb-8b-gpt2", "model": {"d_model": 1280, "heads": 20, "layers": 16, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 5120, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "500m", "vocab_size": 50304}, "model_id": "fuzzy-topk-autoencoder-gpt-v1-family", "sequence_length": 8192, "tokenizer_id": "gpt2"}, "evaluations": {"fineweb": {"canonical": true, "loss": 3.1003285944461823, "perplexity": 22.20524660348485, "scored_tokens": 1048576, "seconds": 1.7348054798785597}}, "implementation": {"attention_backend": "tpu_flash", "attention_tuning": {"key_digest": "84fe3fd56290695de85e933b91350cf7559555228c45b8400b40d51e9e691c20", "resolution_source": "heuristic", "tiles": {"block_kv": 512, "block_kv_compute": 256, "block_kv_dkv": 512, "block_kv_dkv_compute": 256, "block_kv_dq": 512, "block_kv_dq_compute": 256, "block_q": 512, "block_q_dkv": 512, "block_q_dkv_compute": 256, "block_q_dq": 256}, "tune_seconds": 0.0}, "configuration": {"context_preset": "8k", "path": "dev.yaml", "profile": "dev", "resolved": {"evaluation": {"eval_batches": 8, "final_predictions": 1048576, "probe_predictions": 0, "val_every": 0, "val_probe_batches": 0}, "kernels": {"attention_backend": "tpu_flash", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048}, "logging": {"diagnostics_every": 10, "log_every": 10, "sparsity_diagnostics_every": 10}, "model": {"d_model": 1280, "heads": 20, "layers": 16, "mlp_activation": "fuzzy_topk_relu", "mlp_mult": 16, "mlp_top_k": 5120, "normalization": "rms_norm", "parameterization": "completep_fixed_tpp_v1", "position_encoding": "rope_base_10000", "semantic_vocab_size": 50304, "tied_embeddings": false, "tier": "500m", "vocab_size": 50304}, "optimizer": {"adam_epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "effective": {"adam_epsilon_horizon_multiplier": 1.0, "beta1": 0.9, "beta2": 0.95, "global_peak_learning_rate": 0.00390625, "weight_decay_horizon_multiplier": 1.0}, "grad_clip": 0.0, "learning_rate": 0.00390625, "min_lr_ratio": 0.1, "warmup_steps": 1956, "weight_decay": 0.1}, "parameterization": {"attention_scale": "inverse_head_dim", "base_depth": 12, "base_width": 384, "batch_ratio": 1.0, "depth_alpha": 1.0, "depth_multiplier": 1.3333333333333333, "embeddings": "untied", "init_std": 0.02, "ladder_data_multiplier": 1.0, "name": "completep_fixed_tpp_v1", "width_multiplier": 3.3333333333333335}, "training": {"batch_size": 16, "dtype": "bfloat16", "sampling": "shuffled_epochs", "seq_len": 8192, "steps": 19561, "target_tokens_per_parameter": 2.389537337594556, "tokens_per_parameter": 2.389537337594556, "train_tokens": 2563899392}}, "schema_version": 6, "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "context_preset": "8k", "document_masking": true, "loss_backend": "tiled", "sparse_mlp_backend": "choicewise", "vocab_tile_size": 2048, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "logs": {"stderr": "stderr.log", "stderr_sha256": "c59a22df39c0a2706fe9b9f5672b07b78a5a3df56d73247d7c0c2c0799975a41", "stdout": "stdout.log", "stdout_sha256": "e8936d5372f4cb79697ce225faa06bd8764734e0ede79da309ef4e42dfde0eda"}, "metrics": {"achieved_tflops": 1164.112342725145, "attention_tune_seconds": 0.0, "base_learning_rate": 0.00390625, "diagnostic_compile_seconds": 96.65750715904869, "diagnostic_point_count": 1958, "diagnostics_every": 10, "estimated_total_flops": 21023633708112936960, "eval_compile_seconds": 17.654935136903077, "final_validation_seconds": 1.7348054798785597, "flop_accounting": {"by_site": {"_choicewise_fuzzy_topk_mlp": 41231686041600, "dot_general": 9448949022720, "tpu_flash_causal_attention_bwd_dkv": 5497558138880, "tpu_flash_causal_attention_bwd_dq": 5497558138880, "tpu_flash_causal_attention_fwd": 5497558138880}, "elementwise_per_sequence": 12057100530, "matmul_per_sequence": 67173309480960, "method": "traced-jaxpr", "warnings": []}, "flops_per_token": 8199866880, "mfu_estimate": 0.26457098698298753, "model_tier": "500m", "parameter_count": 1072968960, "parameters": 1072968960, "schedule_steps": 19561, "sparsity_diagnostic_compile_seconds": 19.68290765210986, "sparsity_diagnostic_point_count": 1958, "sparsity_diagnostics_every": 10, "stop_after_step": null, "target_tokens_per_parameter": 2.389537337594556, "tokens_per_parameter": 2.389537337594556, "tokens_per_second": 141967.2245612291, "tokens_processed": 2563899392, "total_compile_seconds": 224.390300540952, "train_compile_seconds": 90.39495059289038, "train_loss": 2.885099411010742, "train_seconds": 18059.797956353053, "training_data_epochs": 0.32454505767079067, "training_data_sharding": "rank_disjoint_shuffled_windows", "training_sampling": "shuffled_epochs", "training_steps": 19561, "training_token_budget": 2563899392, "training_usable_tokens_per_epoch": 7899979776, "validation_loss": 3.1003285944461823, "validation_probe_count": 0, "validation_probe_seconds": 0.0, "validation_tokens": 1048576}, "name": "500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339", "plan": {"achieved_tokens_per_parameter": 2.389537337594556, "base_learning_rate": 0.00390625, "batch_ratio": 1.0, "batch_size": 16, "config_schema_version": 6, "config_sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4", "context_preset": "8k", "declared_parameters": 1072968960, "document_masking": true, "expected_tokens": 2563899392, "ladder_data_multiplier": 1.0, "parameterization": "completep_fixed_tpp_v1", "planned_tokens": 2563899392, "profile": "dev", "run_kind": "full", "schedule_steps": 19561, "schema_version": 3, "sequence_length": 8192, "sha256": "2cbdd2d61dc2ea127447e7938c6fd686a20a0b6a63c42d22b6b50f34d085ff2e", "stop_after_step": null, "target_tokens_per_parameter": 2.389537337594556, "tier": "500m", "tokens_per_step": 131072, "validation_predictions": 1048576, "weight_decay_policy": "weights_and_embeddings_only_v2"}, "profile": "dev", "provenance": {"cluster": {"tpu_vm_count": 4, "tpu_vm_hosts": "t1v-n-a09f5679-w-[0-3]"}, "config_yaml": {"bytes": 2941, "path": "recipes/fuzzy_topk_autoencoder/dev.yaml", "sha256": "35ebd3ca279a100bf7c76299955bbcb664c4f1cc959e21f5c332270d82aad7a4"}, "dataset": {"integrity": "sha256", "manifest": {"canonical_sha256": "34b3a4a2098d31c2633b8065e3dad70a1f53ca8bff24c1061d344a01d68a6804", "path": "data/manifests/fineweb-scaled-gpt2/8B.json", "sha256": "d77eb50e311bd36b8a3fc3ec4da4c71c6b200c308167a3afcb3a27f75a430d52"}, "name": "fineweb-8b-gpt2", "profile": "official", "train_files": ["fineweb_train_000001.bin", "fineweb_train_000002.bin", "fineweb_train_000003.bin", "fineweb_train_000004.bin", "fineweb_train_000005.bin", "fineweb_train_000006.bin", "fineweb_train_000007.bin", "fineweb_train_000008.bin", "fineweb_train_000009.bin", "fineweb_train_000010.bin", "fineweb_train_000011.bin", "fineweb_train_000012.bin", "fineweb_train_000013.bin", "fineweb_train_000014.bin", "fineweb_train_000015.bin", "fineweb_train_000016.bin", "fineweb_train_000017.bin", "fineweb_train_000018.bin", "fineweb_train_000019.bin", "fineweb_train_000020.bin", "fineweb_train_000021.bin", "fineweb_train_000022.bin", "fineweb_train_000023.bin", "fineweb_train_000024.bin", "fineweb_train_000025.bin", "fineweb_train_000026.bin", "fineweb_train_000027.bin", "fineweb_train_000028.bin", "fineweb_train_000029.bin", "fineweb_train_000030.bin", "fineweb_train_000031.bin", "fineweb_train_000032.bin", "fineweb_train_000033.bin", "fineweb_train_000034.bin", "fineweb_train_000035.bin", "fineweb_train_000036.bin", "fineweb_train_000037.bin", "fineweb_train_000038.bin", "fineweb_train_000039.bin", "fineweb_train_000040.bin", "fineweb_train_000041.bin", "fineweb_train_000042.bin", "fineweb_train_000043.bin", "fineweb_train_000044.bin", "fineweb_train_000045.bin", "fineweb_train_000046.bin", "fineweb_train_000047.bin", "fineweb_train_000048.bin", "fineweb_train_000049.bin", "fineweb_train_000050.bin", "fineweb_train_000051.bin", "fineweb_train_000052.bin", "fineweb_train_000053.bin", "fineweb_train_000054.bin", "fineweb_train_000055.bin", "fineweb_train_000056.bin", "fineweb_train_000057.bin", "fineweb_train_000058.bin", "fineweb_train_000059.bin", "fineweb_train_000060.bin", "fineweb_train_000061.bin", "fineweb_train_000062.bin", "fineweb_train_000063.bin", "fineweb_train_000064.bin", "fineweb_train_000065.bin", "fineweb_train_000066.bin", "fineweb_train_000067.bin", "fineweb_train_000068.bin", "fineweb_train_000069.bin", "fineweb_train_000070.bin", "fineweb_train_000071.bin", "fineweb_train_000072.bin", "fineweb_train_000073.bin", "fineweb_train_000074.bin", "fineweb_train_000075.bin", "fineweb_train_000076.bin", "fineweb_train_000077.bin", "fineweb_train_000078.bin", "fineweb_train_000079.bin"], "train_tokens_available": 7900000000, "validation_files": ["fineweb_val_000000.bin"], "validation_prefix_tokens": 10485760, "validation_tokens_available": 100000000}, "git": {"dirty": true, "head": "6d3ad351837e5b0d2e5320105b4ee4d5a250e1cb", "status_porcelain": " M pyproject.toml\n M uv.lock\n", "status_porcelain_sha256": "6881d720b28fadbe8b982f6e7cabbac44d8be11c0b09d17e3c36337f26d1bdf5"}, "shared_python": {"bytes": 817194, "entries": [{"bytes": 72, "path": "rig/__init__.py", "sha256": "2d49b7dac21990cfa10df0c80b81edd9486966d5a4c995f1cfec2bfaabec3041"}, {"bytes": 80, "path": "rig/__main__.py", "sha256": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a"}, {"bytes": 13410, "path": "rig/arguments.py", "sha256": "1788ad9442edf5d2a783b4d97c15534072e1f68cf1181aa0596527138519b125"}, {"bytes": 7304, "path": "rig/attention.py", "sha256": "17bff4822d14b3a1dcd50400a272a2a6cce9a397234afc7ceba884c742b411a8"}, {"bytes": 75081, "path": "rig/cli.py", "sha256": "47790d0b9652f5e7868c9d34aacb06563aa01481f1fb43670ec673360d9b4c13"}, {"bytes": 12712, "path": "rig/cohort.py", "sha256": "79ce93d29d721fbd6dd14e8cebcfa71b76761437d3d06cbdb8794a95cb2e65f8"}, {"bytes": 15394, "path": "rig/config.py", "sha256": "42d57e18ddbd2087a63f0db1cd4ab63d232f41099780dd1a886b9a5876db8569"}, {"bytes": 4264, "path": "rig/configfile.py", "sha256": "58ba216edfb570a690ee7663a5a24b33016117e2f5cab848bcc4a65bc1f99511"}, {"bytes": 10278, "path": "rig/configschema.py", "sha256": "bf73be2dea4719912b48f60a25714d3b83c86b88e8f4ebabb1661fdb6655602a"}, {"bytes": 12247, "path": "rig/console.py", "sha256": "92637e59a389e2da302a9f351879fd9d5f6a7467726ea7d7970df97eaf5d8cb2"}, {"bytes": 46054, "path": "rig/data.py", "sha256": "bf53bcb172042518ab2735606d4fefb1c97388b96d6ac4a2a28e728e27df750c"}, {"bytes": 15917, "path": "rig/data_routing.py", "sha256": "11ddfa1fd929d2233f79402483bbaa2c33ea45aa32b34cbb91cf6c32b2d6e7ed"}, {"bytes": 9320, "path": "rig/diagnostics.py", "sha256": "f09643280a3e97899ba4f8d07401b7edd1df38442b35b1b3cebd5f383b105901"}, {"bytes": 15382, "path": "rig/doctor.py", "sha256": "d6456fdbbbc759ea53ee682e3baf10a2d4e5ae7c97c0b586110254a2233587a0"}, {"bytes": 13119, "path": "rig/evaluation.py", "sha256": "f5901c9c8502a494df0b710c28b97d19643c1951671e383c759c84fb4e451c72"}, {"bytes": 18768, "path": "rig/flops.py", "sha256": "285e97527379aa6ea698c15714d9e71459bf26d84e0ae0659c0643f7a409a323"}, {"bytes": 1241, "path": "rig/harness/__init__.py", "sha256": "6b86b47e9fcf50a2883f263f499b96a425098b500f98a5b0704ed98c0310d54b"}, {"bytes": 45418, "path": "rig/harness/cluster.py", "sha256": "9945259ffdfc9a8f910fd93533c2ac0ddd6f95dabfd4a559fe21ef823a66b964"}, {"bytes": 610, "path": "rig/harness/errors.py", "sha256": "90cc900967e2cd86d190deba60180cf541f1c5c6e3a639dcf34954ceb428598b"}, {"bytes": 3025, "path": "rig/harness/models.py", "sha256": "548fd91ed408949ea04c90cc4140ffe15a1ef8111d9c9064969c1318621afabc"}, {"bytes": 2485, "path": "rig/harness/records.py", "sha256": "d8c5392395fcb432fd746ab08a1fa85cf43dcb3fe3966d3bfa89b37f757e9ed6"}, {"bytes": 34486, "path": "rig/harness/runner.py", "sha256": "2e44ed85bbb56b00b544fd0c84358d689e01b52ff8ce7309f7350bbf4765ff64"}, {"bytes": 4538, "path": "rig/harness/scoring.py", "sha256": "1cab21bf26a80894e4c1629095e003bd4583ae503dbc1e55284d2d3237fea9bb"}, {"bytes": 16652, "path": "rig/harness/validation.py", "sha256": "dfc65d15cdfeeb0386dcf1c997d476020763b25465acaadff5d436b75586a4bf"}, {"bytes": 2442, "path": "rig/kernels/__init__.py", "sha256": "c7d41dfa29913eb59a374ed18fbbfcb975eaeb20348eae50675a97471b8b1e93"}, {"bytes": 45858, "path": "rig/kernels/autotune.py", "sha256": "7e430d68d8f1bd324fe7d29f7113229571058d25c9dcd291978f0325aa94a588"}, {"bytes": 31960, "path": "rig/kernels/double_fuzzy_topk.py", "sha256": "ef132d42381bfa099c3f2e69d851cea8fbd9660c75544f13c4bd20d5c441a13e"}, {"bytes": 22148, "path": "rig/kernels/fuzzy_topk.py", "sha256": "e7e7b705b1ebd932d34dc18af9e84349db597ebb45ffca49b6146806d15932d3"}, {"bytes": 13320, "path": "rig/kernels/linear_cross_entropy.py", "sha256": "e0cbab4cda97a8f5878e0deaf0a573dd28cf76952a1226e8b3c6612e9d3d79b9"}, {"bytes": 19834, "path": "rig/kernels/sparse_mlp.py", "sha256": "10f54ef4cb5ccd03e05d93caabdc6a27963eb142917768391ff68e98a5748c68"}, {"bytes": 44208, "path": "rig/kernels/tpu_flash_attention.py", "sha256": "f437acc9aff584974ee5c681427193839236e7d1df23b791ea2f5f14f0b500a5"}, {"bytes": 15883, "path": "rig/logpack.py", "sha256": "d64f42f4ee76809a071a464e419b0497abf5ee69167e84668e13dc880552e4f9"}, {"bytes": 8273, "path": "rig/mesh.py", "sha256": "20fff82a93b54abefd9eb66fdd3f34115e94755687da9a2ffb7f5a592088fdd1"}, {"bytes": 9450, "path": "rig/metrics.py", "sha256": "d8b349057e7dbee98fbac5d5d3388729b84a346e62ce5e5be63d2d16cf22979c"}, {"bytes": 2744, "path": "rig/nn.py", "sha256": "96802cae7d14d71c61fa5dc3862afedeadff979ed8b6ed3209b353ab066acb21"}, {"bytes": 10220, "path": "rig/plan.py", "sha256": "57aaac5ac391cd40d09ad2f297800195bb50cdea1365a4d03727ca027f749594"}, {"bytes": 8565, "path": "rig/recipe_args.py", "sha256": "2f5bf5e4114aaf9f9dbce39e2310098831494a71c35e7d260a4ed2a77c24014b"}, {"bytes": 151393, "path": "rig/report.py", "sha256": "3013dbfbef16137b124ab595f063f5c179906bcd863bc8c66a3a61d70996a11e"}, {"bytes": 492, "path": "rig/rules.py", "sha256": "6041b7cfdc41d0c2542006d321a8f955512d1dd633dbdbd71e0c70a66b77d29d"}, {"bytes": 15666, "path": "rig/runlog.py", "sha256": "6ca2c934286db32f6664a908af0bcf8c071a9a88738076d4bf4b2be2a4fc99d3"}, {"bytes": 26407, "path": "rig/tokens.py", "sha256": "2bfc782cc7cfbbaf1f5a4384b9b56ca337eebf3aca4d0b2c1a6655e383eeda49"}, {"bytes": 10474, "path": "rig/vectorlog.py", "sha256": "f38bda9e9589f3d35d0d203265601dfb6e774eb1ad2bb9430490527d2adde195"}], "files": 42, "sha256": "66d0a5f4366c2ef3ffedcbabff9588c0fe72897d669a26dd44326d4c01f327fc"}, "train_py": {"bytes": 107541, "path": "recipes/fuzzy_topk_autoencoder/train.py", "sha256": "8b13c59e90ba1c203aa3259b4c4b3f968755882ed4d045e8948fd79eeccc84e1"}, "uv_lock": {"bytes": 92151, "path": "uv.lock", "sha256": "dc531e95fdee1b4fc93802d3d2e5bfd984b9a1ea063832d28dc9b1ded1e305e0"}}, "qualified": true, "recipe": "fuzzy_topk_autoencoder", "record_version": 1, "run_id": "20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5", "run_kind": "full", "seed": 1339, "status": "ok", "system": {"controller_process_index": 1, "device_count": 16, "device_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], "device_kinds": ["TPU v4"], "jax_version": "0.11.0", "jaxlib_version": "0.11.0", "libtpu_version": "0.0.44.1", "local_device_count": 4, "platform": "tpu", "process_count": 4, "process_indices": [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3], "python_version": "3.12.13"}, "target_loss": 3.28, "timestamps": {"finished_at": "2026-08-28T21:38:54.673895+00:00", "started_at": "2026-08-28T16:33:41.488564+00:00"}, "timing": {"observed_wall_seconds": 18313.185316057876}, "track": "open", "trainer_command": ["/home/cubic27/GPT-speedrun-TPU/.venv/bin/python", "/home/cubic27/GPT-speedrun-TPU/recipes/fuzzy_topk_autoencoder/train.py", "--output-dir", "/home/cubic27/GPT-speedrun-TPU/runs/20260828T163341.487835Z-fuzzy_topk_autoencoder-500m-fuzzy-sparsity-c10-s1339-09d410d5", "--seed", "1339", "--profile", "dev", "--omit-checkpoint", "--tier", "500m", "--context", "8k", "--base-learning-rate", "0.00390625", "--batch-size", "16", "--sparse-layers", "16", "--sparse-top-k", "5120", "--sparse-training-steps", "19561", "--sparsity-diagnostics-every", "10", "--data-format", "llmc", "--dataset-id", "fineweb-8b-gpt2", "--tokenizer-id", "gpt2", "--color", "never", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000001.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000002.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000003.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000004.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000005.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000006.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000007.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000008.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000009.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000010.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000011.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000012.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000013.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000014.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000015.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000016.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000017.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000018.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000019.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000020.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000021.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000022.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000023.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000024.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000025.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000026.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000027.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000028.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000029.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000030.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000031.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000032.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000033.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000034.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000035.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000036.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000037.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000038.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000039.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000040.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000041.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000042.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000043.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000044.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000045.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000046.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000047.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000048.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000049.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000050.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000051.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000052.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000053.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000054.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000055.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000056.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000057.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000058.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000059.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000060.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000061.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000062.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000063.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000064.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000065.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000066.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000067.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000068.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000069.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000070.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000071.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000072.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000073.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000074.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000075.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000076.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000077.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000078.bin", "--train-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_train_000079.bin", "--val-data", "/dev/shm/.speedrun-cache/fineweb-scaled/8B/fineweb_val_000000.bin"], "folder": "500m-fuzzy-h16d-k4d-l16-2tpp-bs16-lr2e-8-s1339"} diff --git a/fuzzy-topk-sparsity-diagnostics-ladder/snapshot.json.gz b/fuzzy-topk-sparsity-diagnostics-ladder/snapshot.json.gz new file mode 100644 index 0000000000000000000000000000000000000000..6e9e3e71dea776fb7288ef7da773eea545fa6183 --- /dev/null +++ b/fuzzy-topk-sparsity-diagnostics-ladder/snapshot.json.gz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6660ad3b5251fa2c1c54c72055c02ff9eac2fdd960a41203fafb1b82a6abdcaf +size 479201