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

Modify mask_offset‘s format (#3525)

Browse Source 
* modify mask_offset in decode

* modify mask_offset unittest

---------

Co-authored-by: YuBaoku <49938469+EmmonsCurse@users.noreply.github.com>
This commit is contained in:
lzy
2025-09-02 18:05:35 +08:00
committed by GitHub
parent f296aff6cf
commit 7a521bbf62
5 changed files with 13 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/gpu_ops/append_attn/append_attention_func.cuh
+1 -1
View File
@@ -929,7 +929,7 @@ __device__ __forceinline__ void mask_s(const bool* attn_mask,
8 * (reg_id / 4) + reg_id % 2;
bool out_of_boundary;
if (mask_offset) {
out_of_boundary = q_idx < qo_len ? (kv_idx > mask_offset[q_idx]) : true;
out_of_boundary = q_idx < qo_len ? (kv_idx >= mask_offset[q_idx * 2 + 1] || kv_idx < mask_offset[q_idx * 2]) : true;
} else {
out_of_boundary =
(causal