[dynamo][guards] Recursive dict tag optimization #159183
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/159183
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 Cancelled Job, 2 Unrelated FailuresAs of commit 453077c with merge base 8d37073 ( CANCELLED JOB - The following job was cancelled. Please retry:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
anijain2305
added
topic: not user facing
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames Lucaskabela [ghstack-poisoned]
| # tags to avoid full guard execution. | ||
| use_recursive_dict_tags_for_guards = False | ||
|
|
||
| max_saved_pointers_for_recursive_dict_tags_check = 256 |
There was a problem hiding this comment.
Will add in the next one, where I turn the flag on, and let this PR go into fbcode.
There was a problem hiding this comment.
One case I wonder about is if people use __slots__ which can cause object members to not come from __dict__. Maybe nn.Module doesn't supprt this though...
There was a problem hiding this comment.
In this case, we will not have any guards with __dict_ and we will not apply a recursive dict tag optimization.
It will be safe.
|
@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: slow / linux-jammy-rocm-py3.10 / test (slow, 1, 2, linux.rocm.gpu.2, module:rocm) Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge -i |
Merge startedYour change will be merged while ignoring the following 3 checks: trunk / linux-jammy-rocm-py3.10 / test (distributed, 1, 1, linux.rocm.gpu.4), slow / linux-jammy-rocm-py3.10 / test (slow, 1, 2, linux.rocm.gpu.2, module:rocm), pull / linux-jammy-py3_9-clang9-xla / test (xla, 1, 1, lf.linux.12xlarge, unstable) Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Design doc here - https://docs.google.com/document/d/1W29DrWID5miGWlZXspsQVN5U0zydE3kjZpziOXrhuaY/edit?tab=t.0#bookmark=id.sba04iw9sp68 Pull Request resolved: #159183 Approved by: https://github.com/jansel
3 participants
Stack from ghstack (oldest at bottom):
Design doc here - https://docs.google.com/document/d/1W29DrWID5miGWlZXspsQVN5U0zydE3kjZpziOXrhuaY/edit?tab=t.0#bookmark=id.sba04iw9sp68
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang @amjames @Lucaskabela