happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Add Euclidean-Cosine score to identify important tensors

This commit is contained in:
Ed Addario 2025-10-30 22:11:40 +00:00
parent 6e32244a06
commit c59bb6d49d
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 19 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
src/llama-quant.cpp
View File

@ -1571,12 +1571,25 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
// Certain tensors have a higher impact on model quality, so we apply a lower penalty to them
auto is_important = [&](const std::string & tensor_name) -> bool {
const auto important = tensor_name == "output.weight" ||
tensor_name.find(".ffn_down.weight") != std::string::npos ||
tensor_name.find(".ffn_down_exps.weight") != std::string::npos ||
tensor_name.find(".attn_output.weight") != std::string::npos ||
tensor_name.find(".time_mix_output.weight") != std::string::npos ||
tensor_name.find(".attn_o.weight") != std::string::npos;
bool important = false;
if (statistics_data) {
float ecs = 0.0f; // Euclidean-Cosine score
const std::string key = remap_imatrix(tensor_name, mapped);
const auto tstats = statistics_data->find(key);
if (tstats != statistics_data->end() && !tstats->second.empty()) {
ecs = tstats->second.front();
important = ecs == 100.0f; // mark as important if ecs is 100%
}
} else {
important = tensor_name == "output.weight" ||
tensor_name.find(".ffn_down.weight") != std::string::npos ||
tensor_name.find(".ffn_down_exps.weight") != std::string::npos ||
tensor_name.find(".attn_output.weight") != std::string::npos ||
tensor_name.find(".time_mix_output.weight") != std::string::npos ||
tensor_name.find(".attn_o.weight") != std::string::npos;
}
return important;
};