Does Your Model Run on AWS Trainium? Look It Up

Hugging Face and AWS Neuron published a torch-neuronx model support Space and public dataset that records which Transformers models run unmodified on AWS Trainium, tested on trn2 instances with up to 16 cores. Each entry reports the number of Trainium cores a model is sharded over, wall time including compilation, and peak device memory, and links a run.py script using from_pretrained, a static cache and torch.compile(backend="neuron"). Testing is currently limited to text-generation and image-text-to-text Transformers models that fit on 4 Trainium2 devices, and a PASS says nothing about throughput or longer generations.

Updated • 60 https://huggingface.co/datasets/aws-neuron/torch-neuronx-model-support Does Your Model Run on AWS Trainium? Look It Up Update on GitHub https://github.com/huggingface/blog/blob/main/torch-neuronx-model-support.md In a previous post https://huggingface.co/blog/hf-torchneuron , we introduced TorchNeuron, the native PyTorch backend for AWS Trainium the torch-neuronx package . With it, Transformers runs its own modeling code on Trainium, as it does on any other device, so any Transformers model on the Hub may run. Which ones actually do? The torch-neuronx model support https://huggingface.co/spaces/aws-neuron/torch-neuronx-model-support Space answers that question for the most popular models on the Hub. It tells you whether a model runs unmodified on Trainium, and gives you the script to run it yourself. Look up a model The Space opens on the Selected view: our own selection of models picked among the top models on the Hub by downloads and by trending score. Each row gives the result, the number of Trainium cores the model is sharded over, the wall time of the run compilation included and the peak device memory. To look up a given model, switch to the Trending or Downloads view and type part of its name. These views follow the Hub's own rankings, one task at a time, whether a model was tested or not. When a model was left out on purpose, the Space tells you why most of the time it was not prioritized, sometimes it is truly incompatible . You can also pick the versions of the Neuron compiler neuronx-cc and of Transformers to look at. At the time of writing, only Transformers models are tested, and our selection is limited to text-generation and image-text-to-text models. We plan to expand to diffusers, sentence transformers, and other transformers tasks. Run it yourself Select a model to see how to run it. The Space lists the Neuron stack the model was tested on, the packages to install on top of it, and the run.py script that ran. The script follows the recipe of the TorchNeuron announcement https://huggingface.co/blog/hf-torchneuron : it loads the model with from pretrained , sharded when it needs to be, then generates with a static cache and torch.compile backend="neuron" . On a Trainium2 trn2 instance with that stack, for example in the AWS Neuron Deep Learning Container https://github.com/aws-neuron/deep-learning-containers torch-neuronx , install the packages, then run: python3 run.py No need for torchrun : the script launches its own torch.distributed ranks when the model is sharded. A first run starts by compiling the model's graphs. To skip that step, copy the graphs the same compiler built during the test, from the public NEFF cache https://huggingface.co/buckets/aws-neuron/torch-neuronx-neff-cache bucket. The Space gives you the command, with the right compiler version: hf buckets sync hf://buckets/aws-neuron/torch-neuronx-neff-cache/neuronxcc-<version /tmp/neff cache What "runs" means Each test loads the model on a trn2 instance, sharded over the smallest number of Trainium cores that can hold its weights. Most models are sharded with tensor parallelism. Mixture-of-experts models without a tensor-parallel plan use expert parallelism only. The test then generates 32 tokens and checks them. On a single core, the tokens are compared with those of a CPU run of the same checkpoint. When the model is sharded, a small judge model checks that the text is fluent English. At the time of writing, only models that fit on 4 Trainium2 devices are tested, which means 16 cores at most. A PASS says nothing about throughput, or about longer generations. The results are a dataset Everything the Space shows comes from the public aws-neuron/torch-neuronx-model-support https://huggingface.co/datasets/aws-neuron/torch-neuronx-model-support dataset, with one Parquet file per pair of neuronx-cc and library versions. Each row holds a model's result, the error when it failed, the versions of the whole stack and the run.py that ran. You can query it directly: 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", "neuron cores", "result" The dataset card https://huggingface.co/datasets/aws-neuron/torch-neuronx-model-support gives the exact meaning of each result. Next time you wonder whether a model runs on Trainium, look it up https://huggingface.co/spaces/aws-neuron/torch-neuronx-model-support .