CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

gguf-llama.cpp : fix n_head_kv

Browse Source
This commit is contained in:
klosax
2023-08-14 23:18:41 +02:00
committed by GitHub
parent ab2cbd03ca
commit ca4758290c
1 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
gguf-llama.cpp
View File

@@ -630,8 +630,9 @@ struct gguf_file_loader {
hparams.n_rot = read_u32("llama.rope.dimension_count");
hparams.f_rms_norm_eps = read_f32("llama.attention.layer_norm_rms_epsilon");
// LLaMAv2
// hparams.n_head_kv = read_u32("llama.attention.head_count_kv");
// n_head_kv default to n_head
hparams.n_head_kv = gguf_find_key(gguf_ctx, "llama.attention.head_count_kv") == -1 ? hparams.n_head : read_u32("llama.attention.head_count_kv");
}
void read_vocab() {