Skip to content

Conversation

@Colwzq
Copy link

@Colwzq Colwzq commented Dec 8, 2025

Code01: Add flash_attn lib comparsion in Code01. Fix Code01 typo.
Code02: Change hyper parameters to get clear comparsion and add new introduction about grad check.
Code03: Using official torch FP8 ops instead of numerical simulation, original simulation can make mistake. Add new content about acclerating in newer devices using transformer_engine.
Code04: Add outputs.

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