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

falcon : do not offload the embeddings layer

Browse source
This commit is contained in:
Georgi Gerganov 2023-08-25 11:54:57 +03:00
parent ac4bb6ba02
commit 333e27b31f
1 changed files with 0 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama.cpp
View file

@ -2756,8 +2756,6 @@ static struct ggml_cgraph * llm_build_falcon(
cur = ggml_add(ctx0, cur = ggml_add(ctx0,
ggml_mul(ctx0, cur, model.output_norm), ggml_mul(ctx0, cur, model.output_norm),
model.output_norm_b); model.output_norm_b);
offload_func_nr(cur->src[0]);
offload_func_nr(cur);
ggml_set_name(cur, "result_norm"); ggml_set_name(cur, "result_norm");
} }