Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 09:15:18 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into concedo_experimental

Browse Source 
# Conflicts:
#	.github/workflows/build.yml
#	CMakeLists.txt
#	README.md
#	scripts/build-info.cmake
This commit is contained in:
Concedo
2023-11-28 20:57:56 +08:00
parent ba5c33319b 8406b0924b
commit 581021ab93
19 changed files with 845 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
llama.cpp
+1 -11
View File
@@ -5582,18 +5582,8 @@ static int llama_decode_internal(
n_threads = std::min(4, n_threads);
}
// If all tensors can be run on the GPU then using more than 1 thread is detrimental.
const bool full_offload_supported =
model.arch == LLM_ARCH_LLAMA ||
model.arch == LLM_ARCH_BAICHUAN ||
model.arch == LLM_ARCH_FALCON ||
model.arch == LLM_ARCH_REFACT ||
model.arch == LLM_ARCH_MPT ||
model.arch == LLM_ARCH_STARCODER ||
model.arch == LLM_ARCH_STABLELM;
const bool fully_offloaded = model.n_gpu_layers >= (int) hparams.n_layer + 3;
if (ggml_cpu_has_cublas() && full_offload_supported && fully_offloaded) {
if (ggml_cpu_has_cublas() && fully_offloaded) {
n_threads = 1;
}