root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 19:04:35 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

metal : another fix for the fa kernel

Browse Source
This commit is contained in:
Georgi Gerganov 2024-08-26 14:55:28 +03:00
parent 7a3df798fc
commit a95225cdfd
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 13 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-metal.metal
View File

@ -2144,6 +2144,7 @@ kernel void kernel_flash_attn_ext_f16(
const short tx = tiisg%4;
const short ty = tiisg/4;
if (iq1 + ty < ne01) {
// mqk = mqk*scale
ss[8*cc + ty*TF + 2*tx + 0] *= scale;
ss[8*cc + ty*TF + 2*tx + 1] *= scale;
@ -2160,6 +2161,7 @@ kernel void kernel_flash_attn_ext_f16(
}
}
}
}
// used to detect blocks full of -INF
float smax = -INFINITY;