Skip to content

fix: Dont explicitly source AWS plugin on Polaris#95

Open
saforem2 wants to merge 1 commit intomainfrom
saforem2/polaris-fix
Open

fix: Dont explicitly source AWS plugin on Polaris#95
saforem2 wants to merge 1 commit intomainfrom
saforem2/polaris-fix

Conversation

@saforem2
Copy link
Member

Copilot Summary

This pull request makes minor adjustments to the ALCF/helpers.sh script, primarily focusing on changing the default data type and commenting out plugin setup for AWS NCCL OFI on Polaris.

Configuration updates:

  • Changed the default value of the DTYPE environment variable from fp16 to bf16 in the setParams() function, making bfloat16 the default data type.

Infrastructure setup:

  • Commented out the sourcing of the AWS NCCL OFI Plugin script for Polaris, so it is no longer set up by default.

Other:

  • Removed the loss_scale parameter from the bfloat16 section in the generated DeepSpeed config.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant