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

quantize: add iq3_xxs warning

Browse source 
iq3_xxs should not be used without imatrix.
This commit is contained in:
ymcui 2024-02-05 15:14:40 +08:00
parent 4833ac209d
commit a5aa793ed4
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/quantize/quantize.cpp
View file

@ -287,9 +287,9 @@ int main(int argc, char ** argv) {
}
}
if ((params.ftype == LLAMA_FTYPE_MOSTLY_IQ2_XS || params.ftype == LLAMA_FTYPE_MOSTLY_IQ2_XXS || params.ftype == LLAMA_FTYPE_MOSTLY_Q2_K_S) && imatrix_data.empty()) {
if ((params.ftype == LLAMA_FTYPE_MOSTLY_IQ2_XS || params.ftype == LLAMA_FTYPE_MOSTLY_IQ2_XXS || params.ftype == LLAMA_FTYPE_MOSTLY_Q2_K_S || params.ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) && imatrix_data.empty()) {
fprintf(stderr, "\n===============================================================================================\n");
fprintf(stderr, "Please do not use IQ2_XXS, IQ2_XS or Q2_K_S quantization without an importance matrix\n");
fprintf(stderr, "Please do not use IQ2_XXS, IQ2_XS, Q2_K_S or IQ3_XXS quantization without an importance matrix\n");
fprintf(stderr, "===============================================================================================\n\n\n");
return 1;
}