baalajimaestro/ollama
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

fix MB VRAM log output (#824)

Browse source
This commit is contained in:
Bruce MacDonald 2023-10-17 15:35:16 -04:00 committed by GitHub
parent f55bdb6f10
commit bd93a94abd
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
llm/llama.go
View file

@ -249,7 +249,7 @@ func NumGPU(numLayer, fileSizeBytes int64, opts api.Options) int {
// max number of layers we can fit in VRAM, subtract 8% to prevent consuming all available VRAM and running out of memory // max number of layers we can fit in VRAM, subtract 8% to prevent consuming all available VRAM and running out of memory
layers := int(freeBytes/bytesPerLayer) * 92 / 100 layers := int(freeBytes/bytesPerLayer) * 92 / 100
log.Printf("%d MiB VRAM available, loading up to %d GPU layers", freeBytes, layers) log.Printf("%d MB VRAM available, loading up to %d GPU layers", freeBytes/(1024*1024), layers)
return layers return layers
} }