[tests] use torch_device instead of auto for model testing
#29531
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
|
Hi @faaany - thanks for opening this PR! Using device_map="auto" is necessary for this test - it's checking that beam search works when the model is split across devices. If it doesn't work with XPU, then you can add a skip with unittest e.g. if "xpu" in torch_device:
return unittest.skip("device_map='auto' does not work with XPU devices") |
Contributor
Author
|
Hi @amyeroberts , thanks so much for reviewing this PR! I updated my patch and used |
amyeroberts
reviewed
Mar 8, 2024
Contributor
amyeroberts
left a comment
amyeroberts
left a comment
There was a problem hiding this comment.
Thanks for handling!
Just a small nit
Co-authored-by: amyeroberts <[email protected]>
Contributor
Author
done, thx! |
dvrogozh
added a commit
to dvrogozh/transformers
that referenced
this pull request
Jan 17, 2025
`return unittest.skip()` used in the `test_model_parallel_beam_search` in skip condition for xpu did not actually mark test to be skipped running under pytest: * 148 passed, 1 skipped Other tests use `self.skipTest()`. Reusing this approach and moving the condition outside the loop (since it does not depend on it) allows to skip for xpu correctly: * 148 skipped Secondly, `device_map="auto"` is now implemented for XPU for IPEX>=2.5 and torch>=2.6, so we can now enable these tests for XPU for new IPEX/torch versions. Fixes: 1ea3ad1 ("[tests] use `torch_device` instead of `auto` for model testing (huggingface#29531)") Signed-off-by: Dmitry Rogozhkin <[email protected]>
ydshieh
pushed a commit
that referenced
this pull request
Jan 17, 2025
`return unittest.skip()` used in the `test_model_parallel_beam_search` in skip condition for xpu did not actually mark test to be skipped running under pytest: * 148 passed, 1 skipped Other tests use `self.skipTest()`. Reusing this approach and moving the condition outside the loop (since it does not depend on it) allows to skip for xpu correctly: * 148 skipped Secondly, `device_map="auto"` is now implemented for XPU for IPEX>=2.5 and torch>=2.6, so we can now enable these tests for XPU for new IPEX/torch versions. Fixes: 1ea3ad1 ("[tests] use `torch_device` instead of `auto` for model testing (#29531)") Signed-off-by: Dmitry Rogozhkin <[email protected]>
bursteratom
pushed a commit
to bursteratom/transformers
that referenced
this pull request
Jan 31, 2025
…ngface#35742) `return unittest.skip()` used in the `test_model_parallel_beam_search` in skip condition for xpu did not actually mark test to be skipped running under pytest: * 148 passed, 1 skipped Other tests use `self.skipTest()`. Reusing this approach and moving the condition outside the loop (since it does not depend on it) allows to skip for xpu correctly: * 148 skipped Secondly, `device_map="auto"` is now implemented for XPU for IPEX>=2.5 and torch>=2.6, so we can now enable these tests for XPU for new IPEX/torch versions. Fixes: 1ea3ad1 ("[tests] use `torch_device` instead of `auto` for model testing (huggingface#29531)") Signed-off-by: Dmitry Rogozhkin <[email protected]>
elvircrn
pushed a commit
to elvircrn/transformers
that referenced
this pull request
Feb 13, 2025
…ngface#35742) `return unittest.skip()` used in the `test_model_parallel_beam_search` in skip condition for xpu did not actually mark test to be skipped running under pytest: * 148 passed, 1 skipped Other tests use `self.skipTest()`. Reusing this approach and moving the condition outside the loop (since it does not depend on it) allows to skip for xpu correctly: * 148 skipped Secondly, `device_map="auto"` is now implemented for XPU for IPEX>=2.5 and torch>=2.6, so we can now enable these tests for XPU for new IPEX/torch versions. Fixes: 1ea3ad1 ("[tests] use `torch_device` instead of `auto` for model testing (huggingface#29531)") Signed-off-by: Dmitry Rogozhkin <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
When running test cases under the models folder on XPU, I found that many model tests fail at the same test
test_model_parallel_beam_search, e.g.FAILED tests/models/bigbird_pegasus/test_modeling_bigbird_pegasus.py::BigBirdPegasusModelTest::test_model_parallel_beam_search - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and xpu:0! (when cheThis is because
device_map="auto"is used. As elaborated in this PR, the device_map="auto" mechanism is still not mature on XPU, causing model to be loaded on CPU, rather than on XPU.If there is no particular reason for using
auto, I would suggest usingtorch_deviceinsteadauto, becausetorch_deviceis more specific thanautoand we don't have the need to useauto, e.g. large model inference, in our tests anyway. WDYT? @ArthurZucker