tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-13 20:07:41 -04:00
Code Issues Packages Projects Releases Wiki Activity

quantize : skip tensor override when in fallback mode (#14995)

Browse Source
This commit is contained in:
Ed Addario
2025-07-31 20:32:18 +01:00
committed by GitHub
parent a06ed5feae
commit daf2dd7880
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llama-quant.cpp
View File

@@ -875,9 +875,10 @@ static void llama_model_quantize_impl(const std::string & fname_inp, const std::
// get more optimal quantization type based on the tensor shape, layer, etc.
if (!params->pure && ggml_is_quantized(default_type)) {
int fallback = qs.n_fallback;
new_type = llama_tensor_get_type(qs, new_type, tensor, ftype);
// unless the user specifies a type
if (params->tensor_types) {
// unless the user specifies a type, and the tensor geometry will not require fallback quantisation
if (params->tensor_types && qs.n_fallback - fallback == 0) {
const std::vector<tensor_quantization> & tensor_types = *static_cast<const std::vector<tensor_quantization> *>(params->tensor_types);
const std::string tensor_name(tensor->name);
for (const auto & [tname, qtype] : tensor_types) {
@@ -890,7 +891,6 @@ static void llama_model_quantize_impl(const std::string & fname_inp, const std::
}
}
}
if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) {
new_type = params->token_embedding_type;
}