[AOTAutograd] Use _set_grad_enabled instead of no_grad #128183
Closed
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/128183
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ You can merge normally! (3 Unrelated Failures)As of commit 6306c3f with merge base 9554300 ( FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
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 was referenced Jun 7, 2024
This saves ~1us of overhead from each inductor graph call. [ghstack-poisoned]
lezcano
approved these changes
Jun 8, 2024
| t._dynamo_weak_dynamic_indices = o.dynamic_dims.copy() | ||
| if runtime_metadata.grad_enabled_mutation is not None: | ||
| torch.set_grad_enabled(runtime_metadata.grad_enabled_mutation) | ||
| torch._C._set_grad_enabled(runtime_metadata.grad_enabled_mutation) |
There was a problem hiding this comment.
is there so much overhead calling torch.set_grad_enabled vs calling the C API directly?
There was a problem hiding this comment.
In [1]: import torch
...: %timeit torch.set_grad_enabled(False)
...: %timeit torch._C._set_grad_enabled(False)
536 ns ± 3.02 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)
217 ns ± 1.28 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)It's 2.5 x slower, mostly because torch.set_grad_enabled is a context manager object that pretends to be a normal function.
peterbell10
added a commit
to peterbell10/pytorch
that referenced
this pull request
Jun 9, 2024
This saves ~1us of overhead from each inductor graph call. ghstack-source-id: 217cfd7 Pull Request resolved: pytorch#128183
peterbell10
added a commit
to peterbell10/pytorch
that referenced
this pull request
Jun 9, 2024
This saves ~1us of overhead from each inductor graph call. ghstack-source-id: a6dd843 Pull Request resolved: pytorch#128183
pytorchmergebot
pushed a commit
that referenced
this pull request
Jun 9, 2024
`gen_alias_from_base` spends about ~0.5 us in this import statement, which is called for each view in the graph output. Pull Request resolved: #128184 Approved by: https://github.com/lezcano ghstack dependencies: #128183
pytorchmergebot
pushed a commit
that referenced
this pull request
Jun 9, 2024
Going through the dispatcher + pybind11 + torch.ops adds about 2 us overhead per call compared to `PyArgParser`. Note that views of inputs are reconstructed by AOTAutograd before being returned to the python code, so dispatching for autograd's sake shouldn't be required here. Pull Request resolved: #128185 Approved by: https://github.com/lezcano ghstack dependencies: #128183, #128184
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
This saves ~1us of overhead from each inductor graph call. Pull Request resolved: pytorch#128183 Approved by: https://github.com/lezcano
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
…h#128184) `gen_alias_from_base` spends about ~0.5 us in this import statement, which is called for each view in the graph output. Pull Request resolved: pytorch#128184 Approved by: https://github.com/lezcano ghstack dependencies: pytorch#128183
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
…8185) Going through the dispatcher + pybind11 + torch.ops adds about 2 us overhead per call compared to `PyArgParser`. Note that views of inputs are reconstructed by AOTAutograd before being returned to the python code, so dispatching for autograd's sake shouldn't be required here. Pull Request resolved: pytorch#128185 Approved by: https://github.com/lezcano ghstack dependencies: pytorch#128183, pytorch#128184
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
… from C++ (pytorch#128187) Marginal overhead reduction when calling through the `torch.ops` API. Pull Request resolved: pytorch#128187 Approved by: https://github.com/lezcano ghstack dependencies: pytorch#128183, pytorch#128184, pytorch#128185
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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):
This saves ~1us of overhead from each inductor graph call.