Implements dual-chunk-flash-attn backend for dual chunk attention with sparse attention support (#11844)

This commit is contained in:
Tao He
2025-05-12 19:52:47 -07:00
committed by GitHub
parent f6518b2b48
commit 60f7624334
17 changed files with 2444 additions and 32 deletions
+1
View File
@@ -230,6 +230,7 @@ set(VLLM_EXT_SRC
"csrc/attention/paged_attention_v1.cu"
"csrc/attention/paged_attention_v2.cu"
"csrc/attention/merge_attn_states.cu"
"csrc/attention/vertical_slash_index.cu"
"csrc/pos_encoding_kernels.cu"
"csrc/activation_kernels.cu"
"csrc/layernorm_kernels.cu"