Skip to content

fix: relax tolerance in test_flash_attention_block_masked_soft_cap#4439

Draft
hsuan-lun-chiang wants to merge 1 commit into
AI-Hypercomputer:mainfrom
CIeNET-International:fix/test-flash-attn-soft-cap
Draft

fix: relax tolerance in test_flash_attention_block_masked_soft_cap#4439
hsuan-lun-chiang wants to merge 1 commit into
AI-Hypercomputer:mainfrom
CIeNET-International:fix/test-flash-attn-soft-cap

Conversation

@hsuan-lun-chiang

Copy link
Copy Markdown
Collaborator

Relax tolerance in test_flash_attention_block_masked_soft_cap to accommodate TPU precision differences (bfloat16 matmul and hardware approximations). Confirmed that the failure is not NNX-specific.

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