nod-ai / SHARK-TestSuite

Temporary home of a test suite we are evaluating
Apache License 2.0
2 stars 23 forks source link

Teach download_remote_files.py to use huggingface_hub. #274

Closed ScottTodd closed 1 week ago

ScottTodd commented 1 week ago

As part of this, I added support for downloading models from Hugging Face (in addition to Azure Storage). The huggingface_hub package has its own caching strategy, which I adapted to work with our custom caching for our GitHub Actions runners.

ScottTodd commented 1 week ago

Oops, description from another PR:

As part of this, I added support for downloading models from Hugging Face (in addition to Azure Storage). The huggingface_hub package has its own caching strategy, which I adapted to work with our custom caching for our GitHub Actions runners.

  • Hugging Face limits access to certain files based on authentication (e.g. the HF_TOKEN env var), which we already handle in the e2eshark/ subproject. If we add tests for gated models then we can set that up here somehow too.
  • Generic file downloading is still unimplemented, and as a reminder we can add other hosts like Google Cloud Storage, Amazon S3, etc. as needed.