Refactor code according to keyedModelHandler changes #21819
Conversation
|
Run Python 3.7 PostCommit |
|
Run Python 3.9 PostCommit |
| PytorchModelHandler( | ||
| state_dict_path=known_args.model_state_dict_path, | ||
| model_class=model_class, | ||
| model_params=model_params)) |
There was a problem hiding this comment.
Do we still need the type hints on line 139?
There was a problem hiding this comment.
yes, still getting the error(expecting Union) if I remove the output hint.
There was a problem hiding this comment.
Do you think it's because RunInference is imported from api.py instead of base.py and the RunInference in api.py has Union[...] type?
There was a problem hiding this comment.
ok. I thought the splitting of model handlers would allow us getting the type deterministically instead of having to make a union. we can look into it separately.
cc: @robertwb @yeandy @rezarokni
| state_dict_path=known_args.model_state_dict_path, | ||
| model_class=model_class, | ||
| model_params=model_params) | ||
| # the input to RunInference transform is keyed. Wrap |
There was a problem hiding this comment.
This sounds like the input is always keyed. Isn't it user's choice? If so, would this be more appropriate instead:
In this example we pass keyed inputs to RunInference transform.
Therefore, we use KeyedModelHandler wrapper over PytorchModelHandler.
sdks/python/apache_beam/examples/inference/pytorch_image_classification.py
Outdated
Show resolved
Hide resolved
|
@AnandInguva can you fix the input and state_dict path names too? Or maybe in a separate PR |
|
Ran the IT test that uses the above example offline and it passes. Also all the unit tests for torch RunInference passes locally. |
will look into it in a separate PR. thanks |
* Refactor code according to keyedModelHandler changes * Add comments on why keyedModelHandler is used.
Resolve tests failing due to #21777.
Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:
R: @username).addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, commentfixes #<ISSUE NUMBER>instead.CHANGES.mdwith noteworthy changes.See the Contributor Guide for more tips on how to make review process smoother.
To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md
GitHub Actions Tests Status (on master branch)
See CI.md for more information about GitHub Actions CI.