-
Notifications
You must be signed in to change notification settings - Fork 115
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
support stateful model #493
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. |
@echarlaix @AlexKoff88 please take a look, while @slyalin on vacation, I finished his PR with stateful models support |
@echarlaix, we are ready to merge. |
Co-authored-by: Sergey Lyalin <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have a few typo fixes and small comments but none of them should block merging. We should also add this to documentation, but there is more documentation to be fixed before 2023.3 so that can be done in a separate PR. Thanks for the PR @eaidova !
# TODO: Implement stateful for ONNX path as well, not doing it right now because of lack of validation | ||
logger.warn( | ||
"[ WARNING ] Making stateful models is not supported when exporting to ONNX as an intermediate step. A stateless model will be exported instead. It may result in sub-optimal inference performance." | ||
"Provide a model that can be converted to OpenVINO without fallback to ONNX conversion path." |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
"Provide a model that can be converted to OpenVINO without fallback to ONNX conversion path." | |
"Provide a model that can be converted to OpenVINO without fallback to ONNX conversion path." |
I think most people will need to export the model they are exporting and "provide another model" is then not really helpful. In my experience, when export with PyTorch failed and ONNX was used instead, it was usually because of a technical issue, with optimum-intel or dependencies. So we could suggest to upgrade dependencies with pip install --upgrade --upgrade-strategy eager optimum[openvino]
and maybe suggest to create an issue if exporting with PyTorch still fails?
Co-authored-by: Helena Kloosterman <[email protected]>
Co-authored-by: Ella Charlaix <[email protected]>
@echarlaix, I think we are ready to merge |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good thanks for the addition, can you confirm all tests are passing with openvino-nightly
(or another version enabling stateful models) before we can merge ?
Co-authored-by: Ella Charlaix <[email protected]>
I can confirm, I test this locally on my working machine |
@eaidova Do the tests that fail with nightly here also pass for you locally with nightly/2023.3? https://github.com/huggingface/optimum-intel/actions/runs/7473712623/job/20526199806
|
@helena-intel I mean all relevant tests passed. I do not take it into account as it reproducible without stateful models and does not related to my changes. it is probably related to difference model graph representation for quantization |
OK perfect, thanks a lot for checking @helena-intel @eaidova |
What does this PR do?
continue work on stateful models support based on #486
TO DO