vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-07 00:59:03 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

reverse 300MiB for Metal kernel

Browse source
This commit is contained in:
Lizonghang 2025-01-31 16:24:44 +04:00
parent dd632ee6df
commit 17cd8ba618
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
src/llama.cpp
View file

@ -3622,8 +3622,8 @@ void llama_profile_device(
dev_info->gpu_props.name = gpu_props.name;
dev_info->gpu_props.description = gpu_props.description;
// reserved/limit memory to avoid potential OOM, default to 200 MiB
dev_info->gpu_props.memory_free = round(gpu_props.memory_free / (double)(1 << 30) * 100) / 100 - 0.2;
// reserved/limit memory to avoid potential OOM, default to 300 MiB
dev_info->gpu_props.memory_free = round(gpu_props.memory_free / (double)(1 << 30) * 100) / 100 - 0.3;
dev_info->gpu_props.memory_free = std::min((float)gpu_mem, dev_info->gpu_props.memory_free);
dev_info->gpu_props.memory_total = round(gpu_props.memory_total / (double)(1 << 30) * 100) / 100;