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

test without g3n fix

Browse source
This commit is contained in:
Concedo 2025-07-13 23:42:59 +08:00
parent 8cebec5128
commit bc2877d2fe
1 changed files with 5 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

10
gpttype_adapter.cpp
View file

@ -2292,11 +2292,11 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
} }
//handle override tensor //handle override tensor
std::string tensoroverrides = inputs.override_tensors; std::string tensoroverrides = inputs.override_tensors;
if(file_format_meta.model_architecture==GGUFArch::ARCH_GEMMA3N) // if(file_format_meta.model_architecture==GGUFArch::ARCH_GEMMA3N)
{ // {
std::string forced = "per_layer_token_embd.weight=CPU"; //this tensor on gpu is problematic on unsloth q4_0 // std::string forced = "per_layer_token_embd.weight=CPU"; //this tensor on gpu is problematic on unsloth q4_0
tensoroverrides = (tensoroverrides=="" ? forced: (forced+","+tensoroverrides)); // tensoroverrides = (tensoroverrides=="" ? forced: (forced+","+tensoroverrides));
} // }
if(tensoroverrides!="" && ggml_backend_dev_count()>1) if(tensoroverrides!="" && ggml_backend_dev_count()>1)
{ {
printf("Handling Override Tensors for backends: "); printf("Handling Override Tensors for backends: ");