Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-20 01:31:42 +02:00
Code Issues Packages Projects Releases Wiki Activity

support q5_1 kv

Browse Source
This commit is contained in:
Concedo
2026-04-17 17:06:15 +08:00
parent e074939c17
commit 9a38091207
6 changed files with 52 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gpttype_adapter.cpp
+2 -2
View File
@@ -2528,8 +2528,8 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
//set some ctx params early so autofit can use them.
llama_ctx_params.flash_attn_type = (kcpp_data->flash_attn?LLAMA_FLASH_ATTN_TYPE_ENABLED:LLAMA_FLASH_ATTN_TYPE_DISABLED);
llama_ctx_params.swa_full = kcpp_data->swa_full;
llama_ctx_params.type_k = (inputs.quant_k==2?GGML_TYPE_Q4_0:(inputs.quant_k==1?GGML_TYPE_Q8_0:(inputs.quant_k==3?GGML_TYPE_BF16:GGML_TYPE_F16)));
llama_ctx_params.type_v = (inputs.quant_v==2?GGML_TYPE_Q4_0:(inputs.quant_v==1?GGML_TYPE_Q8_0:(inputs.quant_v==3?GGML_TYPE_BF16:GGML_TYPE_F16)));
llama_ctx_params.type_k = (inputs.quant_k==4?GGML_TYPE_Q4_0:(inputs.quant_k==3?GGML_TYPE_Q5_1:(inputs.quant_k==2?GGML_TYPE_Q8_0:(inputs.quant_k==1?GGML_TYPE_BF16:GGML_TYPE_F16))));
llama_ctx_params.type_v = (inputs.quant_v==4?GGML_TYPE_Q4_0:(inputs.quant_v==3?GGML_TYPE_Q5_1:(inputs.quant_v==2?GGML_TYPE_Q8_0:(inputs.quant_v==1?GGML_TYPE_BF16:GGML_TYPE_F16))));
//apply overrides from autofit