kernel: support slightly faster merge_state_v2 cuda kernel (#5381)
This commit is contained in:
@@ -170,6 +170,7 @@ string(REPLACE "-D__CUDA_NO_HALF2_OPERATORS__" "" CMAKE_CUDA_FLAGS "${CMAKE
|
||||
set(SOURCES
|
||||
"csrc/allreduce/custom_all_reduce.cu"
|
||||
"csrc/attention/cascade.cu"
|
||||
"csrc/attention/merge_attn_states.cu"
|
||||
"csrc/attention/cutlass_mla_kernel.cu"
|
||||
"csrc/attention/lightning_attention_decode_kernel.cu"
|
||||
"csrc/elementwise/activation.cu"
|
||||
|
||||
Reference in New Issue
Block a user