vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add flash attention toggle

Browse source
This commit is contained in:
Concedo 2024-04-30 21:29:11 +08:00
parent 17a24d753c
commit c65448d17a
3 changed files with 10 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

3
gpttype_adapter.cpp
View file

@ -785,12 +785,12 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
{
kcpp_params->n_ubatch = (kcpp_params->n_batch>1024?1024:kcpp_params->n_batch);
}
kcpp_params->flash_attn = inputs.flash_attention;
modelname = kcpp_params->model = inputs.model_filename;
useSmartContext = inputs.use_smartcontext;
useContextShift = inputs.use_contextshift;
debugmode = inputs.debugmode;
auto clamped_max_context_length = inputs.max_context_length;
if(clamped_max_context_length>16384 &&
@ -1089,6 +1089,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
}
}
llama_ctx_params.flash_attn = kcpp_params->flash_attn;
llama_ctx_v4 = llama_new_context_with_model(llamamodel, llama_ctx_params);
if (llama_ctx_v4 == NULL)