Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
fix concurrency test
Michael Yang7ed367419e8f
diff --git a/integration/concurrency_test.go b/integration/concurrency_test.goindex 81d0b5878..42e9d0749 100644--- a/integration/concurrency_test.go+++ b/integration/concurrency_test.go@@ -5,6 +5,7 @@ package integrationimport ("context""log/slog"+ "os""strconv""sync""testing"@@ -13,7 +14,6 @@ import ("github.com/stretchr/testify/require""github.com/ollama/ollama/api"- "github.com/ollama/ollama/envconfig""github.com/ollama/ollama/format")@@ -41,8 +41,8 @@ func TestMultiModelConcurrency(t *testing.T) {},}resp = [2][]string{- []string{"sunlight"},- []string{"england", "english", "massachusetts", "pilgrims", "british"},+ {"sunlight"},+ {"england", "english", "massachusetts", "pilgrims", "british"},})var wg sync.WaitGroup@@ -71,12 +71,11 @@ func TestIntegrationConcurrentPredictOrcaMini(t *testing.T) {reqLimit := len(req)iterLimit := 5- vram := os.Getenv("OLLAMA_MAX_VRAM") // TODO - discover actual VRAM- if vram != "" {- max, err := strconv.ParseUint(vram, 10, 64)+ if s := os.Getenv("OLLAMA_MAX_VRAM"); s != "" {+ maxVram, err := strconv.ParseUint(s, 10, 64)require.NoError(t, err)// Don't hammer on small VRAM cards...- if max < 4*1024*1024*1024 {+ if maxVram < 4*format.GibiByte {reqLimit = min(reqLimit, 2)iterLimit = 2}@@ -233,12 +232,12 @@ func TestMultiModelStress(t *testing.T) {consumed := uint64(256 * format.MebiByte) // Assume some baseline usagefor i := 0; i < len(req); i++ {// Always get at least 2 models, but dont' overshoot VRAM too much or we'll take too long- if i > 1 && consumed > vram {- slog.Info("achieved target vram exhaustion", "count", i, "vram", format.HumanBytes2(vram), "models", format.HumanBytes2(consumed))+ if i > 1 && consumed > maxVram {+ slog.Info("achieved target vram exhaustion", "count", i, "vram", format.HumanBytes2(maxVram), "models", format.HumanBytes2(consumed))break}consumed += chosenModels[i].size- slog.Info("target vram", "count", i, "vram", format.HumanBytes2(vram), "models", format.HumanBytes2(consumed))+ slog.Info("target vram", "count", i, "vram", format.HumanBytes2(maxVram), "models", format.HumanBytes2(consumed))wg.Add(1)go func(i int) {diff --git a/integration/llm_test.go b/integration/llm_test.goindex 4952b0726..398e0a03a 100644--- a/integration/llm_test.go+++ b/integration/llm_test.go@@ -35,8 +35,8 @@ var (},}resp = [2][]string{- []string{"sunlight"},- []string{"england", "english", "massachusetts", "pilgrims"},+ {"sunlight"},+ {"england", "english", "massachusetts", "pilgrims"},})diff --git a/integration/max_queue_test.go b/integration/max_queue_test.goindex b06197e1f..ec9e085a5 100644--- a/integration/max_queue_test.go+++ b/integration/max_queue_test.go@@ -29,7 +29,7 @@ func TestMaxQueue(t *testing.T) {// Also note that by default Darwin can't sustain > ~128 connections without adjusting limitsthreadCount := 32if maxQueue := envconfig.MaxQueue(); maxQueue != 0 {- threadCount = maxQueue+ threadCount = int(maxQueue)} else {t.Setenv("OLLAMA_MAX_QUEUE", strconv.Itoa(threadCount))}diff --git a/integration/utils_test.go b/integration/utils_test.goindex c2b27ee93..a60109958 100644--- a/integration/utils_test.go+++ b/integration/utils_test.go@@ -334,10 +334,10 @@ func GenerateRequests() ([]api.GenerateRequest, [][]string) {},},[][]string{- []string{"sunlight"},- []string{"soil", "organic", "earth", "black", "tan"},- []string{"england", "english", "massachusetts", "pilgrims", "british"},- []string{"fourth", "july", "declaration", "independence"},- []string{"nitrogen", "oxygen", "carbon", "dioxide"},+ {"sunlight"},+ {"soil", "organic", "earth", "black", "tan"},+ {"england", "english", "massachusetts", "pilgrims", "british"},+ {"fourth", "july", "declaration", "independence"},+ {"nitrogen", "oxygen", "carbon", "dioxide"},}}