[UX] Make vllm bench serve discover model by default and use --input-len (#30816)
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
@@ -19,21 +19,18 @@ def server():
|
||||
|
||||
@pytest.mark.benchmark
|
||||
def test_bench_serve(server):
|
||||
# Test default model detection and input/output len
|
||||
command = [
|
||||
"vllm",
|
||||
"bench",
|
||||
"serve",
|
||||
"--model",
|
||||
MODEL_NAME,
|
||||
"--host",
|
||||
server.host,
|
||||
"--port",
|
||||
str(server.port),
|
||||
"--dataset-name",
|
||||
"random",
|
||||
"--random-input-len",
|
||||
"--input-len",
|
||||
"32",
|
||||
"--random-output-len",
|
||||
"--output-len",
|
||||
"4",
|
||||
"--num-prompts",
|
||||
"5",
|
||||
|
||||
Reference in New Issue
Block a user