Results from neuronxcc-2.28.7631.0+8f7b0cf5

README.md

@@ -0,0 +1,108 @@
+---
+license: apache-2.0
+pretty_name: torch-neuronx model support
+tags:
+- neuron
+- trainium
+- transformers
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/*/*.parquet
+---
+
+# torch-neuronx model support
+
+Which Hugging Face Hub models run unmodified on AWS Trainium, through the
+native PyTorch backend
+[`torch-neuronx`](https://github.com/aws-neuron/torch-neuronx), for each
+`neuronx-cc` and library version tested. The only library tested so far is
+[`transformers`](https://github.com/huggingface/transformers).
+
+To look up a model, use the
+[Space](https://huggingface.co/spaces/aws-neuron/torch-neuronx-model-support).
+
+## What is tested
+
+### transformers
+
+Each run loads a model on a `trn2` instance and generates 32 tokens greedily
+from the prompt `"The future of AI is"`, sent as a user message for chat
+models. The model is loaded with `tp_plan="auto"`, or expert-parallel for
+mixture-of-experts models without a tensor-parallel plan, at the smallest
+power-of-two degree that fits its bf16 weights. Generation uses
+`torch.compile(backend="neuron")` with a static cache and chunked prefill.
+
+| `result` | Meaning |
+|---|---|
+| `PASS` | The model loaded, `generate()` returned, and the output check passed or could not run. |
+| `FAIL(load)` | `from_pretrained` raised. |
+| `FAIL(gen)` | `generate()` raised. |
+| `FAIL(output)` | `generate()` returned, but the output check failed. |
+| `FAIL(other)` | The run crashed, or the model's extra Python packages could not be installed. |
+
+The output check compares a TP=1 run with a CPU run of the same checkpoint
+and prompt: the tokens must match, or differ first at a near-tie. At higher
+TP, or when the CPU run cannot complete, a small instruct model judges whether
+the text is fluent English. `PASS` says nothing about throughput or about
+longer generations.
+
+The models are the top `text-generation` and `image-text-to-text` models on
+the Hub by downloads and trending, reselected before each sweep. Quantized
+checkpoints and models that need `trust_remote_code` are left out. A model
+that leaves the list keeps the results it already has.
+
+## Layout
+
+There is one file per tested `(neuronx-cc, library version)` pair:
+
+```
+data/neuronxcc-<version>/<library>-<version>.parquet
+```
+
+Each sweep tests the current Neuron SDK with the library version of the day,
+so most pairs are never tested. A model missing from a pair was not tested
+there, which does not mean it fails.
+
+| Column | Content |
+|---|---|
+| `model_id`, `library`, `tp`, `runner` | The model, the library that loads it, its tensor- or expert-parallel degree, and the load class (`run_dense`: `AutoModelForCausalLM`, `run_multimodal`: `AutoModelForImageTextToText`, `run_dense_ep`: expert-parallel). |
+| `result`, `notes` | The outcome, and the start of the generated text or of the error. |
+| `error` | The full exception message of a `FAIL(load)` or `FAIL(gen)` run. |
+| `cpu_match` | TP=1 only: `Y` if the tokens match a CPU run, `~ near-tie …` if they differ only at a near-tie, `N …` if not. |
+| `english` | The judge's verdict, `Y` or `N`, when the CPU comparison did not run. |
+| `wall_s`, `hbm_peak_gb`, `hbm_capacity_gb` | Wall time of the run, including compilation, and peak device memory against what the TP degree provides. |
+| `neuronx_cc`, `library_version`, `transformers`, `accelerate`, `torch_neuronx`, `torch`, `nki`, `python`, `neuron_driver` | The stack the run used. `transformers` is also set for the libraries built on it. |
+| `tested_at` | When the run started (UTC). |
+| `run_py` | The standalone script that ran. |
+
+## Querying
+
+```python
+import pandas as pd
+
+df = pd.read_parquet("hf://datasets/aws-neuron/torch-neuronx-model-support/data")
+df[df.model_id == "Qwen/Qwen3-0.6B"][["neuronx_cc", "library", "library_version", "tp", "result"]]
+```
+
+## Reproducing a run
+
+Each `run_py` is a standalone script. Its header lists the Neuron stack it ran
+on and a `pip install` line for the packages it adds on top. On a Trainium2
+instance with that stack, for example the AWS Neuron Deep Learning Container,
+install those packages, then:
+
+```bash
+python3 run.py
+```
+
+On a `trn2` instance, the run can skip compilation by first copying the
+compiled graphs of its compiler, from the public
+[NEFF cache](https://huggingface.co/buckets/aws-neuron/torch-neuronx-neff-cache):
+
+```bash
+hf buckets sync hf://buckets/aws-neuron/torch-neuronx-neff-cache/neuronxcc-<neuronx_cc> /tmp/neff_cache
+```
+
+The script launches its own `torch.distributed` ranks for TP > 1.

data/neuronxcc-2.28.7631.0+8f7b0cf5/transformers-5.20.1.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d96da55efdece57ce13e0deaf3a8fe4ada0d21a6bb3b943a2c4069d789aa56d
+size 14186

data/neuronxcc-2.28.7631.0+8f7b0cf5/transformers-5.20.2.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eabc5e0c157025a8a975e5a18043aff6fc3f5f1d9aa1c50880532a48a8466313
+size 177518