-
Notifications
You must be signed in to change notification settings - Fork 25.7k
[Experimental][ONNX] Export with symbolic shapes in proto #112179
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112179
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 4459e9c with merge base b6a30bb ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
[ghstack-poisoned]
[ghstack-poisoned]
…ith symbolic shapes in proto" Experimental feature to store symbolic shapes produced by torch dynamo inside the exported onnx model. There is no official ONNX spec to support nodes within FunctionProto to have value info, onnx/onnx#5487. The names for value info are generated uniquely to be retrievable based on the call site and call stack. This requires onnxscript with https://github.com/microsoft/onnxscript/tree/bowbao/export_symbolic_shapes [ghstack-poisoned]
Experimental feature to store symbolic shapes produced by torch dynamo inside the exported onnx model. There is no official ONNX spec to support nodes within FunctionProto to have value info, onnx/onnx#5487. The names for value info are generated uniquely to be retrievable based on the call site and call stack. This requires onnxscript with https://github.com/microsoft/onnxscript/tree/bowbao/export_symbolic_shapes [ghstack-poisoned]
…Export with symbolic shapes in proto" Experimental feature to store symbolic shapes produced by torch dynamo inside the exported onnx model. There is no official ONNX spec to support nodes within FunctionProto to have value info, onnx/onnx#5487. The names for value info are generated uniquely to be retrievable based on the call site and call stack. This requires onnxscript with https://github.com/microsoft/onnxscript/tree/bowbao/export_symbolic_shapes [ghstack-poisoned]
The link doesn't work on my side. Is it a PR in onnxscript? |
Found it. It's merged. |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: 1 jobs have failed, first few of them are: docker-builds / docker-build (pytorch-linux-focal-rocm-n-1-py3) Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: 1 jobs have failed, first few of them are: docker-builds / docker-build (pytorch-linux-focal-rocm-n-1-py3) Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Pull Request resolved: #114767 Approved by: https://github.com/thiagocrepaldi ghstack dependencies: #112179
Increases perf accuracy especially for low iteration runs. Pull Request resolved: #114821 Approved by: https://github.com/thiagocrepaldi ghstack dependencies: #112179, #114767
ghstack-source-id: 81bd276 Pull Request resolved: pytorch/pytorch#112179
…2179) Experimental feature to store symbolic shapes produced by torch dynamo inside the exported onnx model. There is no official ONNX spec to support nodes within FunctionProto to have value info, onnx/onnx#5487. The names for value info are generated uniquely to be retrievable based on the call site and call stack. This requires onnxscript with https://github.com/microsoft/onnxscript/tree/bowbao/export_symbolic_shapes Pull Request resolved: pytorch#112179 Approved by: https://github.com/titaiwangms, https://github.com/thiagocrepaldi
Pull Request resolved: pytorch#114767 Approved by: https://github.com/thiagocrepaldi ghstack dependencies: pytorch#112179
Increases perf accuracy especially for low iteration runs. Pull Request resolved: pytorch#114821 Approved by: https://github.com/thiagocrepaldi ghstack dependencies: pytorch#112179, pytorch#114767
|
Was https://github.com/microsoft/onnxscript/tree/bowbao/export_symbolic_shapes ever merged into main? |
Stack from ghstack (oldest at bottom):
Experimental feature to store symbolic shapes produced by torch dynamo inside the exported onnx model.
There is no official ONNX spec to support nodes within FunctionProto to have value info, onnx/onnx#5487. The names for value info are generated uniquely to be retrievable based on the call site and call stack.
This requires onnxscript with https://github.com/microsoft/onnxscript/tree/bowbao/export_symbolic_shapes