[benchmark] add a benchmark file for comms used in TorchRec #3435
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
|
@TroyGarden has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83900855. |
meta-cla
bot
added
the
CLA Signed
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Oct 5, 2025
Summary: # context * add benchmark to `torch.distributed.all_to_all_single`. * in this very first edition, this all_to_all_single is running with sync on the same stream. * basic operations: pre-comms compute (GPU compute heavy) ==> all_to_all_single comms ==> irrelevant compute (GPU compute heavy, does not depend on comms data) ==> post-comms compute (GPU compute heavy, uses the comms data) # other changs * extend the cmd_conf decorator so that it can support multiple-program selection in command line ``` python -m torchrec.distributed.benchmark.benchmark_comms \ a2a_single --name=a2a_sync_base-$(git rev-parse --short HEAD || echo $USER) ``` * add a config (dataclass) class **`BenchmarkFunc`** for `benchmark_func`, which includes most common arguments used in `benchmark_func` Differential Revision: D83900855
TroyGarden
force-pushed
the
export-D83900855
branch
from
60277c1 to
c98fb11
Compare
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Oct 6, 2025
Summary: # context * add benchmark to `torch.distributed.all_to_all_single`. * in this very first edition, this all_to_all_single is running with sync on the same stream. * basic operations: pre-comms compute (GPU compute heavy) ==> all_to_all_single comms ==> irrelevant compute (GPU compute heavy, does not depend on comms data) ==> post-comms compute (GPU compute heavy, uses the comms data) # other changs * extend the cmd_conf decorator so that it can support multiple-program selection in command line ``` python -m torchrec.distributed.benchmark.benchmark_comms \ a2a_single --name=a2a_sync_base-$(git rev-parse --short HEAD || echo $USER) ``` * add a config (dataclass) class **`BenchmarkFunc`** for `benchmark_func`, which includes most common arguments used in `benchmark_func` Reviewed By: spmex Differential Revision: D83900855
TroyGarden
force-pushed
the
export-D83900855
branch
from
c98fb11 to
00272da
Compare
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Oct 6, 2025
Summary: # context * add benchmark to `torch.distributed.all_to_all_single`. * in this very first edition, this all_to_all_single is running with sync on the same stream. * basic operations: pre-comms compute (GPU compute heavy) ==> all_to_all_single comms ==> irrelevant compute (GPU compute heavy, does not depend on comms data) ==> post-comms compute (GPU compute heavy, uses the comms data) # other changs * extend the cmd_conf decorator so that it can support multiple-program selection in command line ``` python -m torchrec.distributed.benchmark.benchmark_comms \ a2a_single --name=a2a_sync_base-$(git rev-parse --short HEAD || echo $USER) ``` * add a config (dataclass) class **`BenchmarkFunc`** for `benchmark_func`, which includes most common arguments used in `benchmark_func` Reviewed By: spmex Differential Revision: D83900855
TroyGarden
force-pushed
the
export-D83900855
branch
from
00272da to
02b8ff7
Compare
TroyGarden
changed the title
Summary: # context * add benchmark to `torch.distributed.all_to_all_single`. * in this very first edition, this all_to_all_single is running with sync on the same stream. * basic operations: pre-comms compute (GPU compute heavy) ==> all_to_all_single comms ==> irrelevant compute (GPU compute heavy, does not depend on comms data) ==> post-comms compute (GPU compute heavy, uses the comms data) # other changs * extend the cmd_conf decorator so that it can support multiple-program selection in command line ``` python -m torchrec.distributed.benchmark.benchmark_comms \ a2a_single --name=a2a_sync_base-$(git rev-parse --short HEAD || echo $USER) ``` * add a config (dataclass) class **`BenchmarkFunc`** for `benchmark_func`, which includes most common arguments used in `benchmark_func` Reviewed By: spmex Differential Revision: D83900855
TroyGarden
force-pushed
the
export-D83900855
branch
from
02b8ff7 to
0aba8b3
Compare
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Oct 6, 2025
Summary: # context * add benchmark to `torch.distributed.all_to_all_single`. * in this very first edition, this all_to_all_single is running with sync on the same stream. * basic operations: pre-comms compute (GPU compute heavy) ==> all_to_all_single comms ==> irrelevant compute (GPU compute heavy, does not depend on comms data) ==> post-comms compute (GPU compute heavy, uses the comms data) # other changs * extend the cmd_conf decorator so that it can support multiple-program selection in command line ``` python -m torchrec.distributed.benchmark.benchmark_comms \ a2a_single --name=a2a_sync_base-$(git rev-parse --short HEAD || echo $USER) ``` * add a config (dataclass) class **`BenchmarkFunc`** for `benchmark_func`, which includes most common arguments used in `benchmark_func` Reviewed By: spmex Differential Revision: D83900855
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Summary:
context
torch.distributed.all_to_all_single.pre-comms compute (GPU compute heavy) ==> all_to_all_single comms ==> irrelevant compute (GPU compute heavy, does not depend on comms data) ==> post-comms compute (GPU compute heavy, uses the comms data)
other changs
add a config (dataclass) class
BenchmarkFuncforbenchmark_func, which includes most common arguments used inbenchmark_functrace shows the all_to_all_single (comms) is in the same stream as compute.
Differential Revision: D83900855