Shard RegisterDispatchKey #144364
Shard RegisterDispatchKey #144364
Conversation
Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/144364
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 40388ae with merge base bc57635 ( UNSTABLE - The following job failed but was likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) ghstack-source-id: 260579275 Pull Request resolved: #144364
|
whoops, pushed 7df69db to the wrong PR (it goes in the bottom-of-stack PR), will fix shortly |
|
@swolchok has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
Pull Request resolved: #144364 Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) ghstack-source-id: 260637494
Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
Should fix #143952 . Testing: built PyTorch on Raspberry Pi 5; this seemed to alleviate high peak memory requirement. (I did increase shard counts for other generated files along the way, but I need to go back and figure out how much of that was strictly necessary vs. needing to use -j1 or -j2.) Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
| ":{}[{}]".format(aten_rule_name, "Register" + backend + ".cpp") | ||
| for backend in enabled_backends | ||
| ] | ||
| ":{}[{}]".format(aten_rule_name, "Register" + backend + "_0.cpp") |
There was a problem hiding this comment.
While adjusting this, f-string might be a bit cleaner
There was a problem hiding this comment.
Starlark doesn't have f-strings -- bazelbuild/starlark#91
| def write_sharded_with_template( | ||
| self, | ||
| filename: str, | ||
| template_fn: str, |
There was a problem hiding this comment.
nit: template_fn -> template_name? (it looks like this is the string name of the input template file, not a function)
There was a problem hiding this comment.
this matches the existing template_fn parameter to write_with_template and substitute_with_template. I agree that it is confusing to overload "fn" to abbreviate both "filename" and "function", and there is a lot of historical precedent for "function", so I will just do a file-wide find-replace.
There was a problem hiding this comment.
file-wide find-replace
BC-Lint has informed me that I can't do that, so I will keep it template_fn for consistency :\
Should fix #143952 . Testing: built PyTorch on Raspberry Pi 5; this seemed to alleviate high peak memory requirement. (I did increase shard counts for other generated files along the way, but I need to go back and figure out how much of that was strictly necessary vs. needing to use -j1 or -j2.) Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
Pull Request resolved: #144364 Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). ghstack-source-id: 260776837 Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/)
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
Should fix #143952 . Testing: built PyTorch on Raspberry Pi 5; this seemed to alleviate high peak memory requirement. (I did increase shard counts for other generated files along the way, but I need to go back and figure out how much of that was strictly necessary vs. needing to use -j1 or -j2.) Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
Should fix #143952 . Testing: built PyTorch on Raspberry Pi 5; this seemed to alleviate high peak memory requirement. (I did increase shard counts for other generated files along the way, but I need to go back and figure out how much of that was strictly necessary vs. needing to use -j1 or -j2.) Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/) [ghstack-poisoned]
Pull Request resolved: #144364 Should fix #143952 . Testing: build PyTorch on Raspberry Pi 5 to demonstrate this fixes high peak memory requirement (TODO). ghstack-source-id: 260820041 Differential Revision: [D67925496](https://our.internmc.facebook.com/intern/diff/D67925496/)
|
This pull request was exported from Phabricator. Differential Revision: D67925496 |
|
linux-jammy-py3-clang12-executorch job failed but is unstable and failing on main , so I'm ignoring the failure |
|
@pytorchbot merge (Initiating merge automatically since Phabricator Diff has merged) |
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 |
introduced by #144364. [ghstack-poisoned]
Stack from ghstack (oldest at bottom):
Should fix #143952 .
Testing: built PyTorch on Raspberry Pi 5; this seemed to alleviate high peak memory requirement. (I did increase shard counts for other generated files along the way, but I need to go back and figure out how much of that was strictly necessary vs. needing to use -j1 or -j2.)
Differential Revision: D67925496