vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

metal : initial implementation

Browse source
This commit is contained in:
Georgi Gerganov 2024-01-20 17:32:28 +02:00
parent a9681febd6
commit 1173f49c3b
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
4 changed files with 180 additions and 33 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml.c
View file

@ -13419,8 +13419,8 @@ static void ggml_compute_forward_flash_attn_ext_f16(
const int ik2 = iq2 / rk2;
// v indices
const int iv2 = iq2 / rv2;
const int iv3 = iq3 / rv3;
const int iv2 = iq2 / rv2;
// online softmax / attention
// loop over n_kv and n_head_kv