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

quantize : fix precedence of cli args

Browse source
This commit is contained in:
Georgi Gerganov 2024-04-08 14:24:36 +03:00
parent 87fb5b4234
commit f6da969be7
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
llama.cpp
View file

@ -13559,10 +13559,10 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
if (!params->pure && ggml_is_quantized(default_type)) {
new_type = llama_tensor_get_type(qs, new_type, tensor, ftype);
}
else if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) {
if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) {
new_type = params->token_embedding_type;
}
else if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) {
if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) {
new_type = params->output_tensor_type;
}