-
Notifications
You must be signed in to change notification settings - Fork 355
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
fix: Upgrade to PyTorch 2.0.1 Release Candidate + Other improvements #1857
fix: Upgrade to PyTorch 2.0.1 Release Candidate + Other improvements #1857
Conversation
d5bd60e
to
339675d
Compare
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.
Docker updates were adapted from PR #1805, and customized to CUDA version 11.8
# TODO: The following test has been disabled due to a bug in TRT 8.5.1.7 | ||
# with self.linear2. Issue : https://github.com/pytorch/TensorRT/issues/1444 | ||
@unittest.skip( | ||
reason="test_multi_fuse_permute_linear has been disabled due to a bug in TRT 8.5.1.7 https://github.com/pytorch/TensorRT/issues/1444" | ||
) |
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.
Removed test skip since TRT 8.5 is no longer in use
339675d
to
17398a9
Compare
- Add support for `--legacy` flag in `setup.py`, to enable installation with PyTorch 1.13.1 - Add in test cases previously disabled due to dependency issues - Improve documentation of `README`s across repository - Update cudnn version to 8.8, update TRT version to 8.6 across all files and `README`s - Update `docker` folder on release branch to use new build system with custom TRT, cudnn, CUDA, and Python versions - Update `WORKSPACE` files to use Torch 2.0.1 Release Candidate
17398a9
to
7673be2
Compare
@@ -380,7 +386,7 @@ def run(self): | |||
long_description=long_description, | |||
ext_modules=ext_modules, | |||
install_requires=[ | |||
"torch==2.0.0", | |||
"torch >=2.0.1,<2.1" if not LEGACY else "torch >=1.13.0,<2.0", |
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.
Updated torch requirement to coincide with necessary versions for Legacy/Regular installs.
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.
LGTM
Description
--legacy
flag insetup.py
, to enable installation with PyTorch 1.13.1README
s across repositoryREADME
sdocker
folder on release branch to use new build system with custom TRT, cudnn, CUDA, and Python versionsWORKSPACE
files to use Torch 2.0.1 Release CandidateStacked with PR #1759
Type of change
Checklist: