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

Increase IQ options

This commit is contained in:
Ed Addario 2025-09-13 11:49:18 +01:00
parent 2b516068e2
commit 8503d59ee4
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 10 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
src/llama-quant.cpp
View File

@ -641,12 +641,21 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
constexpr ggml_type iq_quants[] = { constexpr ggml_type iq_quants[] = {
GGML_TYPE_IQ1_S, GGML_TYPE_IQ1_S,
GGML_TYPE_IQ2_XXS,
GGML_TYPE_IQ2_XS,
GGML_TYPE_IQ2_S, GGML_TYPE_IQ2_S,
GGML_TYPE_IQ3_S, GGML_TYPE_IQ3_S,
GGML_TYPE_IQ4_XS, GGML_TYPE_IQ4_XS,
GGML_TYPE_IQ4_NL,
GGML_TYPE_Q5_K, GGML_TYPE_Q5_K,
GGML_TYPE_Q6_K, GGML_TYPE_Q6_K,
GGML_TYPE_Q8_0 GGML_TYPE_Q8_0,
// TODO: find better way to handle F16/BF16
#ifdef GGML_USE_METAL
GGML_TYPE_F16
#else
GGML_TYPE_BF16
#endif
}; };
constexpr double epsilon = 1e-12; constexpr double epsilon = 1e-12;