Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-10-29 22:20:15 +01:00
Code Issues Packages Projects Releases Wiki Activity

metal : fix fa kernel

Browse Source
This commit is contained in:
Georgi Gerganov 2024-08-26 13:02:36 +03:00
parent 7a3df798fc
commit aa931d0375
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 20 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
ggml/src/ggml-metal.metal
View File

@ -2144,19 +2144,26 @@ kernel void kernel_flash_attn_ext_f16(
const short tx = tiisg%4;
const short ty = tiisg/4;
if (logit_softcap == 0.0f) {
if (mask != q) {
// mqk = mqk*scale + mask*slope
ss[8*cc + ty*TF + 2*tx + 0] = scale*ss[8*cc + ty*TF + 2*tx + 0] + slope*mp[ic + 8*cc + ty*nb31/sizeof(half) + 2*tx + 0];
ss[8*cc + ty*TF + 2*tx + 1] = scale*ss[8*cc + ty*TF + 2*tx + 1] + slope*mp[ic + 8*cc + ty*nb31/sizeof(half) + 2*tx + 1];
} else {
// mqk = mqk*scale
ss[8*cc + ty*TF + 2*tx + 0] *= scale;
ss[8*cc + ty*TF + 2*tx + 1] *= scale;
if (logit_softcap != 0.0f) {
ss[8*cc + ty*TF + 2*tx + 0] = logit_softcap*precise::tanh(ss[8*cc + ty*TF + 2*tx + 0]);
ss[8*cc + ty*TF + 2*tx + 1] = logit_softcap*precise::tanh(ss[8*cc + ty*TF + 2*tx + 1]);
}
} else {
if (mask != q) {
// mqk = mqk + mask*slope
ss[8*cc + ty*TF + 2*tx + 0] += slope*mp[ic + 8*cc + ty*nb31/sizeof(half) + 2*tx + 0];
ss[8*cc + ty*TF + 2*tx + 1] += slope*mp[ic + 8*cc + ty*nb31/sizeof(half) + 2*tx + 1];
// mqk = ls*tanh(mqk*scale) + mask*slope
ss[8*cc + ty*TF + 2*tx + 0] = logit_softcap*precise::tanh(scale*ss[8*cc + ty*TF + 2*tx + 0]) + slope*mp[ic + 8*cc + ty*nb31/sizeof(half) + 2*tx + 0];
ss[8*cc + ty*TF + 2*tx + 1] = logit_softcap*precise::tanh(scale*ss[8*cc + ty*TF + 2*tx + 1]) + slope*mp[ic + 8*cc + ty*nb31/sizeof(half) + 2*tx + 1];
} else {
// mqk = ls*tanh(mqk*scale)
ss[8*cc + ty*TF + 2*tx + 0] = logit_softcap*precise::tanh(scale*ss[8*cc + ty*TF + 2*tx + 0]);
ss[8*cc + ty*TF + 2*tx + 1] = logit_softcap*precise::tanh(scale*ss[8*cc + ty*TF + 2*tx + 1]);
}
}
}
}