Add support for torch>=2, torchvision>=0.15.2, onnx>=1.14 #483
Conversation
This file contains 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
Looks like torch script compile was generating an OnnxPad operation with None passed to constant_value argument. This was causing a crash in the execution engine. This was uncovered after fixing a bug of handling prim::Constant Op correctly.
Little hack to pin ONNX ir_version to 8 if onnxruntime version is less than 1.15. This should get fixed soon with update of onnx package. See microsoft/onnxruntime#15874
It works but its slow.
Version update
Seems like ONNX has switched to using / characters in node names. This was causing issues with expression parsing because output port ids are used in expressions.
…nto fix_torch_v2
Keras model serialized in examples/TensorFlow/Keras/kr_N_model.h5 is failing to load on new TF version. Getting: TypeError: weight_decay is not a valid argument, kwargs should be empty for `optimizer_experimental.Optimizer`.
Keras model serialized in examples/TensorFlow/Keras/kr_N_model.h5 is failing to load. Getting: TypeError: weight_decay is not a valid argument, kwargs should be empty for `optimizer_experimental.Optimizer`. Can't reproduce locally, trying this suggested fix: https://stackoverflow.com/questions/74667876/typeerror-weight-decay-is-not-a-valid-argument-kwargs-should-be-empty-for-opt
Re-enable full testing of package on 3.10
Looks like scikit-learn wheels aren't available for 3.11. Getting compilation errors.
|
Thanks @davidt0x, will try it out. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Miscellaneous changes to support latest versions of torch, torchvsion, and ONNX. This adds tests for more torchvision models. I also paired down the torchvision tests to speed things up. There are a couple marked as xfail that I still need to work on.
@pgleeson, I cleaned up some stuff in the CI as well.