CI: fix torchaudio pinning +proper break in rnnt - #47422
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
tarekziade
changed the title
|
re-trigger an image build https://github.com/huggingface/transformers/actions/runs/29722390933 |
|
new run with the smoke tets https://github.com/huggingface/transformers/actions/runs/29722764900 |
|
new run with 2.11 https://github.com/huggingface/transformers/actions/runs/29723933424 |
|
great work ! should we merge ? |
It's green again, @vasqu I will let you review and merge if the change looks good to you now |
|
Just removed the todo from rnnt directly, the comment is more for the docker build tbh (and also already exists there). Merging now and thanks again @IlyasMoutawwakil it should unblock your quantization PR iirc |
CI recapDashboard: View test results in Grafana |
* fix torchaudio pinning +proper break in rnnt * added test coverage for the ABI mistmatch * added a smoke test * bump to 2.11 * set the torch version for fouroversix * tweaks * reduce comment * tweaks * simplify the comment, docker should have the todo already --------- Co-authored-by: vasqu <antonprogamer@gmail.com>
* fix torchaudio pinning +proper break in rnnt * added test coverage for the ABI mistmatch * added a smoke test * bump to 2.11 * set the torch version for fouroversix * tweaks * reduce comment * tweaks * simplify the comment, docker should have the todo already --------- Co-authored-by: vasqu <antonprogamer@gmail.com>
5 participants
What does this PR do?
The quantization image was breaking because torchaudio was not pinned and the new version is not compatible
with the pinned torch.
The fix is similar to how the main docker does it.
This patch also harden the import in
src/transformers/loss/loss_rnnt.pyso the lib does not crash on a torch import and emits the correct error