highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-04-04 01:50:26 +02:00
Code Issues Packages Projects Releases Wiki Activity

Prevent underflow when FreeMemory < overhead (#8014)

Browse Source 
Co-authored-by: Richard Lyons <frob@cloudstaff.com>
This commit is contained in:
frob 2024-12-10 18:10:40 +01:00 committed by GitHub
parent 900f64e6be
commit 63269668c0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
llm/memory.go
View File

@ -182,7 +182,7 @@ func EstimateGPULayers(gpus []discover.GpuInfo, ggml *GGML, projectors []string,
gzo = gpuZeroOverhead
}
// Only include GPUs that can fit the graph, gpu minimum, the layer buffer and at least more layer
if (gpus[i].FreeMemory - overhead) < gzo+max(graphPartialOffload, graphFullOffload)+gpus[i].MinimumMemory+2*layerSize {
if gpus[i].FreeMemory < overhead+gzo+max(graphPartialOffload, graphFullOffload)+gpus[i].MinimumMemory+2*layerSize {
slog.Debug("gpu has too little memory to allocate any layers",
"id", gpus[i].ID,
"library", gpus[i].Library,
@ -228,7 +228,7 @@ func EstimateGPULayers(gpus []discover.GpuInfo, ggml *GGML, projectors []string,
for j := len(gpusWithSpace); j > 0; j-- {
g := gpusWithSpace[i%j]
used := gpuAllocations[g.i] + max(graphPartialOffload, graphFullOffload)
if (g.g.FreeMemory - overhead) > used+layerSize {
if g.g.FreeMemory > overhead+used+layerSize {
gpuAllocations[g.i] += layerSize
layerCounts[g.i]++
layerCount++
@ -251,7 +251,7 @@ func EstimateGPULayers(gpus []discover.GpuInfo, ggml *GGML, projectors []string,
for j := len(gpusWithSpace); j > 0; j-- {
g := gpusWithSpace[layerCount%j]
used := gpuAllocations[g.i] + max(graphPartialOffload, graphFullOffload)
if (g.g.FreeMemory - overhead) > used+memoryLayerOutput {
if g.g.FreeMemory > overhead+used+memoryLayerOutput {
gpuAllocations[g.i] += memoryLayerOutput
layerCounts[g.i]++
layerCount++