[dynamo] Store CompilationEvents in a buffer in torch._dynamo.utils #115788
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Motivation: for testing in dynamo, it would be nice to have an equivalent to torch._inductor.metrics. We already have log_compilation_event, but that only prints logs or dumps to a database (in fbcode); these aren't appropriate for unit tests. This change: adds torch._dynamo.metrics, and then changes convert_frame so that it calls torch._dynamo.metrics.record_compilation_metrics instead. To prevent a super large list from forming, use a deque with a max length; then if the max length is exceeded, the oldest records will be dropped. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/115788
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (2 Unrelated Failures)As of commit a3a2b1e with merge base 87ea6fb ( FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
…ics" Motivation: for testing in dynamo, it would be nice to have an equivalent to torch._inductor.metrics. We already have log_compilation_event, but that only prints logs or dumps to a database (in fbcode); these aren't appropriate for unit tests. This change: adds torch._dynamo.metrics, and then changes convert_frame so that it calls torch._dynamo.metrics.record_compilation_metrics instead. To prevent a super large list from forming, use a deque with a max length; then if the max length is exceeded, the oldest records will be dropped. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
…ics" Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests. This change: * always record the information in torch._dynamo.utils.record_compilation_metrics; here, log into a limited-size deque to prevent the list of metrics from getting too long * if config.log_compilation_metrics, then call back into the original log_compilation_event function cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
davidberard98
changed the title
|
@yanboliang should we test compile time on this to make sure nothing regresses or do you think it should be safe? |
yanboliang
approved these changes
Dec 14, 2023
…amo.utils" Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests. This change: * always record the information in torch._dynamo.utils.record_compilation_metrics; here, log into a limited-size deque to prevent the list of metrics from getting too long * if config.log_compilation_metrics, then call back into the original log_compilation_event function cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
davidberard98
added a commit
that referenced
this pull request
Dec 14, 2023
Motivation: for testing in dynamo, it would be nice to have an equivalent to torch._inductor.metrics. We already have log_compilation_event, but that only prints logs or dumps to a database (in fbcode); these aren't appropriate for unit tests. This change: adds torch._dynamo.metrics, and then changes convert_frame so that it calls torch._dynamo.metrics.record_compilation_metrics instead. To prevent a super large list from forming, use a deque with a max length; then if the max length is exceeded, the oldest records will be dropped. ghstack-source-id: ea76a5b Pull Request resolved: #115788
…amo.utils" Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests. This change: * always record the information in torch._dynamo.utils.record_compilation_metrics; here, log into a limited-size deque to prevent the list of metrics from getting too long * if config.log_compilation_metrics, then call back into the original log_compilation_event function cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
davidberard98
added a commit
that referenced
this pull request
Dec 16, 2023
Motivation: for testing in dynamo, it would be nice to have an equivalent to torch._inductor.metrics. We already have log_compilation_event, but that only prints logs or dumps to a database (in fbcode); these aren't appropriate for unit tests. This change: adds torch._dynamo.metrics, and then changes convert_frame so that it calls torch._dynamo.metrics.record_compilation_metrics instead. To prevent a super large list from forming, use a deque with a max length; then if the max length is exceeded, the oldest records will be dropped. ghstack-source-id: d124c19 Pull Request resolved: #115788
|
Compile times did not change except for when models didn't run. I'm assuming these models didn't run for some other reason, because they weren't even attempted (no logs), and the CI on the PR isn't failing. |
…amo.utils" Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests. This change: * always record the information in torch._dynamo.utils.record_compilation_metrics; here, log into a limited-size deque to prevent the list of metrics from getting too long * if config.log_compilation_metrics, then call back into the original log_compilation_event function cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
|
@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 |
facebook-github-bot
pushed a commit
to pytorch/benchmark
that referenced
this pull request
Dec 20, 2023
Summary: Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests. This change: * always record the information in torch._dynamo.utils.record_compilation_metrics; here, log into a limited-size deque to prevent the list of metrics from getting too long * if config.log_compilation_metrics, then call back into the original log_compilation_event function X-link: pytorch/pytorch#115788 Approved by: https://github.com/yanboliang Reviewed By: jeanschmidt Differential Revision: D52298053 Pulled By: davidberard98 fbshipit-source-id: ef291255d6148c0479f3000b4fb21a4ed72cadcb
dmenig
pushed a commit
to dmenig/pytorch
that referenced
this pull request
Dec 21, 2023
…ytorch#115788) Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests. This change: * always record the information in torch._dynamo.utils.record_compilation_metrics; here, log into a limited-size deque to prevent the list of metrics from getting too long * if config.log_compilation_metrics, then call back into the original log_compilation_event function Pull Request resolved: pytorch#115788 Approved by: https://github.com/yanboliang
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
Motivation: it would be nice to be able to test using the metrics in log_compilation_event; currently dumps logs (or logs to a database in fbcode) - these are hard to use in unit tests.
This change:
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @aakhundov @kadeng