[RUNTIME] Simplify caching API #3433
Closed
Conversation
This file contains 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
andrewjcg
force-pushed
the
cache_api
branch
7 times, most recently
from
deb8d4a
to
e6bf836
Compare
This simplifies the caching API to require callers pass the entire list of files to cache in a "group" in the `put_group` method. This allows caching backends to e.g. serialize these files into a single cache blob, which can avoid issues with caching atomicity.
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Mar 22, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Apr 3, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Apr 3, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Apr 3, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Apr 4, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Apr 4, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
andrewjcg
added a commit
to andrewjcg/pytorch
that referenced
this pull request
Apr 4, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Context: https://fb.workplace.com/groups/420659799592399/posts/778155640509478/ Scuba query for these changes: https://fburl.com/scuba/triton_remote_cache/lb3t1cw4 Test Plan: With D55206078: ``` $ TORCHBENCH_TOL='1e-3' TORCHINDUCTOR_PERMUTE_FUSION='1' TORCHINDUCTOR_SHAPE_PADDING='1' buck2 run mode/opt //pytorch/benchmark:run -- cmf_10x -d cuda -t train --torchdynamo inductor ``` Reviewed By: bertmaher Differential Revision: D55206000
pytorchmergebot
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Apr 4, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Reviewed By: bertmaher Differential Revision: D55206000 Pull Request resolved: #122470 Approved by: https://github.com/bertmaher
sanketpurandare
pushed a commit
to sanketpurandare/pytorch
that referenced
this pull request
Apr 22, 2024
Summary: Pull cache API changes from triton-lang/triton#3433. Among other simplifications, this allows us the cache all files in a "group" atomically, in a single memcache blob, and avoid needing to use other approaches to handle these files coming from different runs. Reviewed By: bertmaher Differential Revision: D55206000 Pull Request resolved: pytorch#122470 Approved by: https://github.com/bertmaher
|
I assume this PR is dead at this point? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This simplifies the caching API to require callers pass the entire list
of files to cache in a "group" in the
put_groupmethod. This allowscaching backends to e.g. serialize these files into a single cache blob,
which can avoid issues with caching atomicity.