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

Update llama.cpp - Case IQ1_XS

Browse source
This commit is contained in:
Nexesenex 2024-03-25 19:30:19 +01:00 • committed by GitHub
parent 3d88431113
commit 8eff402498
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama.cpp
View file

@ -3018,6 +3018,7 @@ struct llama_model_loader {
case GGML_TYPE_IQ2_S: ftype = LLAMA_FTYPE_MOSTLY_IQ2_S; break; case GGML_TYPE_IQ2_S: ftype = LLAMA_FTYPE_MOSTLY_IQ2_S; break;
case GGML_TYPE_IQ3_XXS: ftype = LLAMA_FTYPE_MOSTLY_IQ3_XXS; break; case GGML_TYPE_IQ3_XXS: ftype = LLAMA_FTYPE_MOSTLY_IQ3_XXS; break;
case GGML_TYPE_IQ1_S: ftype = LLAMA_FTYPE_MOSTLY_IQ1_S; break; case GGML_TYPE_IQ1_S: ftype = LLAMA_FTYPE_MOSTLY_IQ1_S; break;
case GGML_TYPE_IQ1_XS: ftype = LLAMA_FTYPE_MOSTLY_IQ1_S; break;
case GGML_TYPE_IQ4_NL: ftype = LLAMA_FTYPE_MOSTLY_IQ4_NL; break; case GGML_TYPE_IQ4_NL: ftype = LLAMA_FTYPE_MOSTLY_IQ4_NL; break;
case GGML_TYPE_IQ4_XS: ftype = LLAMA_FTYPE_MOSTLY_IQ4_XS; break; case GGML_TYPE_IQ4_XS: ftype = LLAMA_FTYPE_MOSTLY_IQ4_XS; break;
case GGML_TYPE_IQ3_S: ftype = LLAMA_FTYPE_MOSTLY_IQ3_S; break; case GGML_TYPE_IQ3_S: ftype = LLAMA_FTYPE_MOSTLY_IQ3_S; break;
@ -12754,6 +12755,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
case LLAMA_FTYPE_MOSTLY_IQ2_M: default_type = GGML_TYPE_IQ2_S; break; case LLAMA_FTYPE_MOSTLY_IQ2_M: default_type = GGML_TYPE_IQ2_S; break;
case LLAMA_FTYPE_MOSTLY_IQ3_XXS: default_type = GGML_TYPE_IQ3_XXS; break; case LLAMA_FTYPE_MOSTLY_IQ3_XXS: default_type = GGML_TYPE_IQ3_XXS; break;
case LLAMA_FTYPE_MOSTLY_IQ1_S: default_type = GGML_TYPE_IQ1_S; break; case LLAMA_FTYPE_MOSTLY_IQ1_S: default_type = GGML_TYPE_IQ1_S; break;
case LLAMA_FTYPE_MOSTLY_IQ1_XS: default_type = GGML_TYPE_IQ1_S; break;
case LLAMA_FTYPE_MOSTLY_IQ4_NL: default_type = GGML_TYPE_IQ4_NL; break; case LLAMA_FTYPE_MOSTLY_IQ4_NL: default_type = GGML_TYPE_IQ4_NL; break;
case LLAMA_FTYPE_MOSTLY_IQ4_XS: default_type = GGML_TYPE_IQ4_XS; break; case LLAMA_FTYPE_MOSTLY_IQ4_XS: default_type = GGML_TYPE_IQ4_XS; break;
case LLAMA_FTYPE_MOSTLY_IQ3_S: default_type = GGML_TYPE_IQ3_S; break; case LLAMA_FTYPE_MOSTLY_IQ3_S: default_type = GGML_TYPE_IQ3_S; break;