Fix inductor fallback_random for dropout/rand_like #89515
Conversation
- Avoid fx graph rewrite that replaces certain ops with ones using triton random - Keep track of replacement ops using triton random, so it is possible to not disable all replacements when using fallback_random [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/89515
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 18cee96: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
|
||
|
|
||
| def replace_fx(gm: torch.fx.GraphModule): | ||
| # Sometimes patch_functions() misses things already in the graph |
There was a problem hiding this comment.
What is AutogradMonkeypatch used for? do i also need to update it? not obvious what it means that 'replacements' "is" the function being called on torch_function
(see above, L27, not in this diff)
There was a problem hiding this comment.
That looks like a typo to me. I think it should be: func in replacements.
There was a problem hiding this comment.
ok makes sense. i'll PR that too but separately in case it breaks something :)
|
@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 |
- Avoid fx graph rewrite that replaces certain ops with ones using triton random - Keep track of replacement ops using triton random, so it is possible to not disable all replacements when using fallback_random Pull Request resolved: pytorch#89515 Approved by: https://github.com/ngimel
4 participants
Stack from ghstack (oldest at bottom):
triton random
to not disable all replacements when using fallback_random
cc @mlazos @soumith @voznesenskym @yanboliang @penguinwu @anijain2305 @EikanWang @jgong5 @Guobing-Chen @chunyuan-w @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @peterbell10 @desertfire