mirrors/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-05-11 10:26:53 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Integration test improvements (#9654)

Browse source 
Add some new test coverage for various model architectures,
and switch from orca-mini to the small llama model.
This commit is contained in:
Daniel Hiltgen 2025-04-16 14:25:55 -07:00 • committed by GitHub
parent 56dc316a57
commit ed4e139314
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
9 changed files with 709 additions and 67 deletions
Download patch file Download diff file Expand all files Collapse all files

33
integration/utils_test.go
View file

@ -24,9 +24,14 @@ import (
"github.com/ollama/ollama/api"
"github.com/ollama/ollama/app/lifecycle"
"github.com/ollama/ollama/format"
"github.com/stretchr/testify/require"
)
const (
smol = "llama3.2:1b"
)
func Init() {
lifecycle.InitLogging()
}
@ -140,7 +145,7 @@ func PullIfMissing(ctx context.Context, client *api.Client, modelName string) er
showCtx, cancel := context.WithDeadlineCause(
ctx,
time.Now().Add(10*time.Second),
time.Now().Add(20*time.Second),
fmt.Errorf("show for existing model %s took too long", modelName),
)
defer cancel()
@ -157,7 +162,7 @@ func PullIfMissing(ctx context.Context, client *api.Client, modelName string) er
}
slog.Info("model missing", "model", modelName)
stallDuration := 30 * time.Second // This includes checksum verification, which can take a while on larger models
stallDuration := 60 * time.Second // This includes checksum verification, which can take a while on larger models, and slower systems
stallTimer := time.NewTimer(stallDuration)
fn := func(resp api.ProgressResponse) error {
// fmt.Print(".")
@ -283,11 +288,11 @@ func DoGenerate(ctx context.Context, t *testing.T, client *api.Client, genReq ap
}
// Generate a set of requests
// By default each request uses orca-mini as the model
// By default each request uses llama3.2 as the model
func GenerateRequests() ([]api.GenerateRequest, [][]string) {
return []api.GenerateRequest{
{
Model: "orca-mini",
Model: smol,
Prompt: "why is the ocean blue?",
Stream: &stream,
KeepAlive: &api.Duration{Duration: 10 * time.Second},
@ -296,7 +301,7 @@ func GenerateRequests() ([]api.GenerateRequest, [][]string) {
"temperature": 0.0,
},
}, {
Model: "orca-mini",
Model: smol,
Prompt: "why is the color of dirt brown?",
Stream: &stream,
KeepAlive: &api.Duration{Duration: 10 * time.Second},
@ -305,7 +310,7 @@ func GenerateRequests() ([]api.GenerateRequest, [][]string) {
"temperature": 0.0,
},
}, {
Model: "orca-mini",
Model: smol,
Prompt: "what is the origin of the us thanksgiving holiday?",
Stream: &stream,
KeepAlive: &api.Duration{Duration: 10 * time.Second},
@ -314,7 +319,7 @@ func GenerateRequests() ([]api.GenerateRequest, [][]string) {
"temperature": 0.0,
},
}, {
Model: "orca-mini",
Model: smol,
Prompt: "what is the origin of independence day?",
Stream: &stream,
KeepAlive: &api.Duration{Duration: 10 * time.Second},
@ -323,7 +328,7 @@ func GenerateRequests() ([]api.GenerateRequest, [][]string) {
"temperature": 0.0,
},
}, {
Model: "orca-mini",
Model: smol,
Prompt: "what is the composition of air?",
Stream: &stream,
KeepAlive: &api.Duration{Duration: 10 * time.Second},
@ -341,3 +346,15 @@ func GenerateRequests() ([]api.GenerateRequest, [][]string) {
{"nitrogen", "oxygen", "carbon", "dioxide"},
}
}
func skipUnderMinVRAM(t *testing.T, gb uint64) {
// TODO use info API in the future
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 maxVram < gb*format.GibiByte {
t.Skip("skipping with small VRAM to avoid timeouts")
}
}
}