starred/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-04-25 10:16:00 +02:00
Code Issues Packages Projects Releases Wiki Activity

gocritic

Browse Source
This commit is contained in:
Michael Yang
2025-11-18 14:07:58 -08:00
parent f01c83ed6d
commit 4d24d8a77d
22 changed files with 58 additions and 91 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
llm/server.go
View File

@@ -982,13 +982,11 @@ nextLayer:
slog.Warn("model request too large for system", "requested", format.HumanBytes2(cpuSize), "available", format.HumanBytes2(available), "total", format.HumanBytes2(systemInfo.TotalMemory), "free", format.HumanBytes2(systemInfo.FreeMemory), "swap", format.HumanBytes2(systemInfo.FreeSwap))
return fmt.Errorf("model requires more system memory (%s) than is available (%s)", format.HumanBytes2(cpuSize), format.HumanBytes2(available))
}
} else {
if vramSize > systemInfo.TotalMemory {
// disable partial offloading when model is greater than total system memory as this
// can lead to locking up the system
s.options.NumGPU = 0
gpuLayers = ml.GPULayersList{}
}
} else if vramSize > systemInfo.TotalMemory {
// disable partial offloading when model is greater than total system memory as this
// can lead to locking up the system
s.options.NumGPU = 0
gpuLayers = ml.GPULayersList{}
}
if gpuLayers.Sum() == 0 {