wassname/vllm
1
0
Fork 0
mirror of https://github.com/wassname/vllm.git synced 2026-06-27 18:27:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Model] Support MAP-NEO model (#5081)

Browse Source 
Co-authored-by: Zhuohan Li <zhuohan123@gmail.com>
This commit is contained in:
SnowDist
2024-05-31 10:24:41 +08:00
committed by GitHub
parent 533c217792
commit a22dea54d3
8 changed files with 18 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
csrc/cpu/attention.cpp
+6
View File
@@ -390,6 +390,9 @@ void paged_attention_v1_impl_launcher(
case 128:
LAUNCH_V1_ATTENTION_KERNEL(T, 128, BLOCK_SIZE);
break;
case 192:
LAUNCH_V1_ATTENTION_KERNEL(T, 192, BLOCK_SIZE);
break;
case 256:
LAUNCH_V1_ATTENTION_KERNEL(T, 256, BLOCK_SIZE);
break;
@@ -703,6 +706,9 @@ void paged_attention_v2_impl_launcher(
case 128:
LAUNCH_V2_ATTENTION_KERNEL(T, 128, BLOCK_SIZE);
break;
case 192:
LAUNCH_V2_ATTENTION_KERNEL(T, 192, BLOCK_SIZE);
break;
case 256:
LAUNCH_V2_ATTENTION_KERNEL(T, 256, BLOCK_SIZE);
break;