8000 [XLA:GPU]: Bump supported codegened all-reduce size up to 4MB. by copybara-service[bot] · Pull Request #112146 · tensorflow/tensorflow · GitHub
[go: up one dir, main page]

Skip to content

[XLA:GPU]: Bump supported codegened all-reduce size up to 4MB.#112146

Draft
copybara-service[bot] wants to merge 1 commit intomasterfrom
exported_pr_882068582
Draft

[XLA:GPU]: Bump supported codegened all-reduce size up to 4MB.#112146
copybara-service[bot] wants to merge 1 commit intomasterfrom
exported_pr_882068582

Conversation

@copybara-service
Copy link

[XLA:GPU]: Bump supported codegened all-reduce size up to 4MB.

Benchmarks show that 2shot is better than nccl based implementation up until
4MB.

Benchmarks show that 2shot is better than nccl based implementation up until
4MB.

PiperOrigin-RevId: 882068582
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labe 494A ls

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

0