highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-09-28 08:47:35 +02:00
Code Issues Packages Projects Releases Wiki Activity

use 10% vram overhead for cuda

Browse Source
This commit is contained in:
Jeffrey Morgan
2024-01-08 23:17:44 -05:00
parent 58ce2d8273
commit cb534e6ac2
2 changed files with 6 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
gpu/gpu.go
View File

@@ -131,10 +131,11 @@ func getCPUMem() (memInfo, error) {
func CheckVRAM() (int64, error) {
gpuInfo := GetGPUInfo()
if gpuInfo.FreeMemory > 0 && (gpuInfo.Library == "cuda" || gpuInfo.Library == "rocm") {
// allocate 384MiB for llama.cpp overhead (outside of model)
overhead := uint64(384 * 1024 * 1024)
if gpuInfo.FreeMemory <= overhead {
return 0, nil
// leave 10% or 400MiB of VRAM free for overhead
overhead := gpuInfo.FreeMemory / 10
minOverhead := 400 * 1024 * 1024
if overhead < minOverhead {
overhead = minOverhead
}
return int64(gpuInfo.FreeMemory - overhead), nil