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

space after commas; Keep indentation multiple of 4 spaces

Browse source
This commit is contained in:
Ashish 2024-04-14 16:34:40 -07:00
parent d2ab693066
commit b7f984a0df
1 changed files with 12 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

24
llama.cpp
View file

@ -5054,8 +5054,8 @@ static bool llm_load_tensors(
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}, false);
// optional q and k layernorms, present in StableLM 2 12B
layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM,"weight", i), {hparams.n_embd_head_k, hparams.n_head}, false);
layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM,"weight", i), {hparams.n_embd_head_k, hparams.n_head_kv}, false);
layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head}, false);
layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head_kv}, false);
// optional FFN norm, not present in StableLM 2 12B which uses parallel residual
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, false);
@ -8118,18 +8118,18 @@ struct llm_build_context {
cb(Kcur, "Kcur", il);
if (model.layers[il].attn_q_norm) {
Qcur = llm_build_norm(ctx0, Qcur, hparams,
model.layers[il].attn_q_norm,
NULL,
LLM_NORM, cb, il);
cb(Qcur, "Qcur", il);
Qcur = llm_build_norm(ctx0, Qcur, hparams,
model.layers[il].attn_q_norm,
NULL,
LLM_NORM, cb, il);
cb(Qcur, "Qcur", il);
}
if (model.layers[il].attn_k_norm) {
Kcur = llm_build_norm(ctx0, Kcur, hparams,
model.layers[il].attn_k_norm,
NULL,
LLM_NORM, cb, il);
cb(Kcur, "Kcur", il);
Kcur = llm_build_norm(ctx0, Kcur, hparams,
model.layers[il].attn_k_norm,
NULL,
LLM_NORM, cb, il);
cb(Kcur, "Kcur", il);
}