vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 00:54:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into concedo_experimental

Browse source 
# Conflicts:
#	.github/workflows/build.yml
#	.gitignore
#	CMakeLists.txt
#	Makefile
#	Package.swift
#	README.md
#	ggml-cuda.cu
#	llama.cpp
#	llama.h
#	scripts/sync-ggml.sh
#	tests/CMakeLists.txt
This commit is contained in:
Concedo 2023-12-08 17:42:26 +08:00
commit ec21fa7712
34 changed files with 5887 additions and 1435 deletions
Download patch file Download diff file Expand all files Collapse all files

7
gpttype_adapter.cpp
View file

@ -679,7 +679,6 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
{
blasbatchsize = 8;
}
params.memory_f16 = inputs.f16_kv;
auto clamped_max_context_length = inputs.max_context_length;
@ -768,7 +767,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
llama_ctx_params_v2.n_ctx = clamped_max_context_length;
//llama_ctx_params.n_parts = -1;
llama_ctx_params_v2.seed = -1;
llama_ctx_params_v2.f16_kv = inputs.f16_kv;
llama_ctx_params_v2.f16_kv = true;
llama_ctx_params_v2.logits_all = false;
llama_ctx_params_v2.use_mmap = inputs.use_mmap;
llama_ctx_params_v2.use_mlock = inputs.use_mlock;
@ -818,7 +817,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
llama_ctx_params.n_ctx = clamped_max_context_length;
//llama_ctx_paran_parts = -1;
llama_ctx_params.seed = -1;
llama_ctx_params.f16_kv = inputs.f16_kv;
llama_ctx_params.f16_kv = true;
llama_ctx_params.low_vram = inputs.low_vram;
llama_ctx_params.mul_mat_q = inputs.use_mmq;
llama_ctx_params.logits_all = false;
@ -895,7 +894,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
//llama_ctx_paran_parts = -1;
llama_ctx_params.seed = -1;
llama_ctx_params.f16_kv = inputs.f16_kv;
//llama_ctx_params.f16_kv = true;
//llama_ctx_params.low_vram = inputs.low_vram;
llama_ctx_params.mul_mat_q = inputs.use_mmq;
llama_ctx_params.logits_all = false;