OSSForks/ollama
1
0
Fork 0
mirror of https://github.com/zebrajr/ollama.git synced 2025-12-06 00:19:51 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

llm: Avoid underflow in free memory logging

Browse Source 
If a GPU's free memory is less than the reserved amount, we might get
an underflow. Since it is an unsigned uint64, we print this as a large
number rather than the more correct 0. This only affects logging, the
actual layout code already handles this correctly.

Bug #12138
This commit is contained in:
Jesse Gross 2025-09-02 10:47:33 -07:00 committed by Jesse Gross
parent 0cc90a8186
commit 8149a3c86e
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
llm/server.go
View File

@ -678,8 +678,12 @@ func (s *ollamaServer) Load(ctx context.Context, gpus discover.GpuInfoList, requ
if !(len(gpus) == 1 && gpus[0].Library == "cpu") { if !(len(gpus) == 1 && gpus[0].Library == "cpu") {
for _, gpu := range gpus { for _, gpu := range gpus {
available := gpu.FreeMemory - envconfig.GpuOverhead() - gpu.MinimumMemory
if gpu.FreeMemory < envconfig.GpuOverhead()+gpu.MinimumMemory {
available = 0
}
slog.Info("gpu memory", "id", gpu.ID, slog.Info("gpu memory", "id", gpu.ID,
"available", format.HumanBytes2(gpu.FreeMemory-envconfig.GpuOverhead()-gpu.MinimumMemory), "available", format.HumanBytes2(available),
"free", format.HumanBytes2(gpu.FreeMemory), "free", format.HumanBytes2(gpu.FreeMemory),
"minimum", format.HumanBytes2(gpu.MinimumMemory), "minimum", format.HumanBytes2(gpu.MinimumMemory),
"overhead", format.HumanBytes2(envconfig.GpuOverhead())) "overhead", format.HumanBytes2(envconfig.GpuOverhead()))