apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

support dsv3 use flashmla (#6593)

Browse Source
This commit is contained in:
周周周
2026-03-03 11:09:43 +08:00
committed by GitHub
parent 0f718baaf2
commit 3cc09418f1
5 changed files with 266 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/gpu_ops/append_attn/get_block_shape_and_split_kv_block.cu
-1
View File
@@ -313,7 +313,6 @@ void GetBlockShapeAndSplitKVBlock(
// decoder
if (max_dec_len_this_time > 0) {
if (mla_backend) {
PADDLE_ENFORCE(group_size <= 64, "now only group_size <= 64");
const int set_chunk_size = get_mla_dec_chunk_size(bsz);
CUDA_CHECK(cudaMemsetAsync(