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

Move is_iq() into a lambda and remove unused variables

This commit is contained in:
Ed Addario 2025-09-25 19:49:47 +01:00
parent 8eedcf74bc
commit a74b410f5f
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 18 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
src/llama-quant.cpp
View File

@ -727,11 +727,28 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
return (double)bytes * 8.0 / (double)ggml_nelements(t); return (double)bytes * 8.0 / (double)ggml_nelements(t);
}; };
auto is_compatible = [&](const ggml_tensor * t, const ggml_type typ) -> bool { auto is_compatible = [](const ggml_tensor * t, const ggml_type typ) -> bool {
const int64_t blck = ggml_blck_size(typ); const int64_t blck = ggml_blck_size(typ);
return blck <= 1 || (t->ne[0] % blck) == 0; return blck <= 1 || (t->ne[0] % blck) == 0;
}; };
auto is_iq = [](const enum ggml_type t) {
switch (t) {
case GGML_TYPE_IQ1_S:
case GGML_TYPE_IQ1_M:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ2_S:
case GGML_TYPE_IQ3_XXS:
case GGML_TYPE_IQ3_S:
case GGML_TYPE_IQ4_NL:
case GGML_TYPE_IQ4_XS:
return true;
default:
return false;
}
};
auto make_compatible = [&](const ggml_tensor * t, const ggml_type typ) -> ggml_type { auto make_compatible = [&](const ggml_tensor * t, const ggml_type typ) -> ggml_type {
if (is_compatible(t, typ)) return typ; if (is_compatible(t, typ)) return typ;
ggml_type fb = fallback_type(typ); ggml_type fb = fallback_type(typ);
@ -995,8 +1012,6 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
std::vector<tensor_info> all; std::vector<tensor_info> all;
all.reserve(tensors.size()); all.reserve(tensors.size());
for (const auto * tw : tensors) { for (const auto * tw : tensors) {
std::vector<std::thread> workers;
workers.reserve(std::max(1, nthread));
ggml_tensor * tensor = tw->tensor; ggml_tensor * tensor = tw->tensor;
const std::string name = ggml_get_name(tensor); const std::string name = ggml_get_name(tensor);
if (!can_quantize(tensor)) { continue; } if (!can_quantize(tensor)) { continue; }