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

llama : use unused n_embd_k_gqa in k_shift

Browse source 
This also slightly reduces the diff from the master branch
This commit is contained in:
Francis Couture-Harpin 2024-09-01 21:59:24 -04:00
parent 5f62db790b
commit 375de5b1f8
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/llama.cpp
View file

@ -10806,7 +10806,7 @@ struct llm_build_context {
ggml_view_3d(ctx0, kv_self.k_l[il], ggml_view_3d(ctx0, kv_self.k_l[il],
n_embd_head_k, n_head_kv, n_ctx, n_embd_head_k, n_head_kv, n_ctx,
ggml_row_size(kv_self.k_l[il]->type, n_embd_head_k), ggml_row_size(kv_self.k_l[il]->type, n_embd_head_k),
ggml_row_size(kv_self.k_l[il]->type, hparams.n_embd_k_gqa(il)), ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa),
0), 0),
lctx.inp_K_shift, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, lctx.inp_K_shift, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow); ext_factor, attn_factor, beta_fast, beta_slow);