【训练营】FlashAttention接入 @simon_chou#128

Closed

Simon-CHOU wants to merge 6 commits intoInfiniTensor:masterfrom

Simon-CHOU:simon_chou-20260316

Commits on Mar 23, 2026

feat(flash-attn): add flash attention forward and backward kernels
Simon Chou
committed
feat(examples): add LLaMA-3 reproduction scripts and update GPT-2
Simon Chou
committed
fix(runtime): improve CUDA error handling and model initialization stability
Simon Chou
committed
docs: update audit logs, performance reports, and build scripts
Simon Chou
committed
fix(cmake): align test executable definitions with master branch
Simon Chou
committed
chore: clean up non-code project files and normalize test directory structure
Simon Chou
committed