Parallel Cross Entropy using online softmax #1456
Merged
+527
−0
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
Signed-off-by: Selvaraj Anandaraj <[email protected]>
for more information, see https://pre-commit.ci
timmoon10
reviewed
Feb 4, 2025
Signed-off-by: Selvaraj Anandaraj <[email protected]>
Signed-off-by: Selvaraj Anandaraj <[email protected]>
|
@timmoon10 Added tests. |
for more information, see https://pre-commit.ci
ksivaman
reviewed
Feb 19, 2025
ksivaman
reviewed
Feb 19, 2025
Signed-off-by: Selvaraj Anandaraj <[email protected]>
…j5597/TransformerEngine into parallel_cross_entropy
for more information, see https://pre-commit.ci
Signed-off-by: Selvaraj Anandaraj <[email protected]>
…j5597/TransformerEngine into parallel_cross_entropy
Signed-off-by: Selvaraj Anandaraj <[email protected]>
Signed-off-by: Selvaraj Anandaraj <[email protected]>
ksivaman
reviewed
Feb 19, 2025
There was a problem hiding this comment.
@sanandaraj5597 Could you fix the linting errors? You could run it locally using bash qa/L0_pytorch_lint/test.sh
Signed-off-by: Selvaraj Anandaraj <[email protected]>
|
Fixed lint errors. |
for more information, see https://pre-commit.ci
ksivaman
reviewed
Feb 19, 2025
Co-authored-by: Kirthi Shankar Sivamani <[email protected]> Signed-off-by: Selvaraj Anandaraj <[email protected]>
|
/te-ci pytorch |
added 2 commits
February 19, 2025 19:07
Signed-off-by: Selvaraj Anandaraj <[email protected]>
…j5597/TransformerEngine into parallel_cross_entropy
|
/te-ci pytorch |
Signed-off-by: Selvaraj Anandaraj <[email protected]>
…j5597/TransformerEngine into parallel_cross_entropy
|
/te-ci pytorch |
Signed-off-by: Selvaraj Anandaraj <[email protected]>
…j5597/TransformerEngine into parallel_cross_entropy
for more information, see https://pre-commit.ci
|
/te-ci pytorch |
Signed-off-by: Selvaraj Anandaraj <[email protected]>
Signed-off-by: Selvaraj Anandaraj <[email protected]>
for more information, see https://pre-commit.ci
|
/te-ci pytorch |
Signed-off-by: Tim Moon <[email protected]>
Signed-off-by: Tim Moon <[email protected]>
|
It seems that the Triton dependency is messing up other tests. The latest upstream Triton (3.2.0) doesn't support Blackwell, so the NVIDIA PyTorch container is using a custom internal build. I've removed Triton as a formal dependency, but we should put it back once Blackwell support is upstreamed. |
|
/te-ci pytorch |
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.
Description
This PR implements a parallel cross entropy function using the online technique to calculate softmax. This feature has multiple aspects:
[Thanks to Liger kernel implementation for providing the idea about online softmax and in-place gradient calculation.]