highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-07-09 18:20:13 +02:00
Code Issues Packages Projects Releases Wiki Activity

use benchmark loop

Browse Source
This commit is contained in:
Bruce MacDonald
2025-03-14 13:52:22 -07:00
parent ac6bc09fa3
commit bc46d0f2dd
1 changed files with 3 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
benchmark/server_benchmark_test.go
View File

@ -83,8 +83,7 @@ func BenchmarkColdStart(b *testing.B) {
// Set number of tokens as our throughput metric // Set number of tokens as our throughput metric
b.SetBytes(int64(tt.maxTokens)) b.SetBytes(int64(tt.maxTokens))
b.ResetTimer() for b.Loop() {
for range b.N {
b.StopTimer() b.StopTimer()
// Ensure model is unloaded before each iteration // Ensure model is unloaded before each iteration
unload(client, m, b) unload(client, m, b)
@ -122,12 +121,11 @@ func BenchmarkWarmStart(b *testing.B) {
// Set number of tokens as our throughput metric // Set number of tokens as our throughput metric
b.SetBytes(int64(tt.maxTokens)) b.SetBytes(int64(tt.maxTokens))
b.ResetTimer() for b.Loop() {
for range b.N {
req := &api.GenerateRequest{ req := &api.GenerateRequest{
Model: m, Model: m,
Prompt: tt.prompt, Prompt: tt.prompt,
Options: map[string]interface{}{"num_predict": tt.maxTokens, "temperature": 0.1}, Options: map[string]any{"num_predict": tt.maxTokens, "temperature": 0.1},
} }
runGenerateBenchmark(b, ctx, client, req) runGenerateBenchmark(b, ctx, client, req)