Allow cublas an cudnn to be in different nvidia folders #92122
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/92122
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 4ef41de: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
|
torch/__init__.py
Outdated
| # Should only be called on Linux if default path resolution have failed | ||
| assert platform.system() == 'Linux', 'Should only be called on Linux' |
There was a problem hiding this comment.
Please keep this check, as .so checks make sense only on Linux
| candidate_cublas_path = os.path.join(nvidia_path, 'cublas', 'lib', 'libcublas.so.11') | ||
| if os.path.exists(candidate_cublas_path) and not cublas_path: | ||
| cublas_path = candidate_cublas_path |
There was a problem hiding this comment.
| candidate_cublas_path = os.path.join(nvidia_path, 'cublas', 'lib', 'libcublas.so.11') | |
| if os.path.exists(candidate_cublas_path) and not cublas_path: | |
| cublas_path = candidate_cublas_path | |
| candidate_cublas_path = os.path.join(nvidia_path, 'cublas', 'lib', 'libcublas.so.11') | |
| if os.path.exists(candidate_cublas_path) and not cublas_path: | |
| if cublas_path is not None: | |
| import warnings | |
| warnings.warn(f"Replacing previosly found cublas {cublas_path} with new candidate {candidate_cublas_path}") | |
| cublas_path = candidate_cublas_path |
There was a problem hiding this comment.
I'm confused by your suggestion. As written the warning will never happen. Did you mean the following?
if os.path.exists(candidate_cublas_path):
if cublas path is not None:
import warnings
warnings.warn(f"Replacing previosly found cublas {cublas_path
8000
} with new candidate {candidate_cublas_path}")
cublas_path = candidate_cublas_pathThere was a problem hiding this comment.
But also I thought the general pattern is to take libraries from earlier in the path if possible, so shouldn't we never do the replacement?
There was a problem hiding this comment.
@malfet I've pushed just the assert. LMK if you definitely want these warnings and modified logic.
| cublas_path = candidate_cublas_path | ||
| candidate_cudnn_path = os.path.join(nvidia_path, 'cudnn', 'lib', 'libcudnn.so.8') | ||
| if os.path.exists(candidate_cudnn_path) and not cudnn_path: | ||
| cudnn_path = candidate_cudnn_path |
There was a problem hiding this comment.
Similar to the above, please add a warning there
|
I've signed the CLA document. I'm not sure why the check is failing. |
|
/easycla |
d3345ea to
97176fc
Compare
|
/easycla |
97176fc to
4ef41de
Compare
|
CLA fixed |
|
@malfet I think this is good to merge, though it looks like I can't merge it myself. |
|
@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: This PR is too stale; the last push date was more than 3 days ago. Please rebase and try again. You can rebase by leaving the following comment on this PR: Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot rebase |
|
You don't have permissions to rebase this PR since you are a first time contributor. If you think this is a mistake, please contact PyTorch Dev Infra. |
|
@pytorchbot merge -f "Lint is green and this code is not really tested by CI" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Fixes #92096