[V1] TPU CI - Add basic perf regression test (#15414)

Signed-off-by: Alexander Matveev <amatveev@redhat.com>
This commit is contained in:
Alexander Matveev
2025-03-31 13:25:20 -04:00
committed by GitHub
parent 2de4118243
commit 9a2160fa55
5 changed files with 192 additions and 20 deletions

View File

@@ -32,7 +32,7 @@ TENSOR_PARALLEL_SIZES = [1]
@pytest.mark.parametrize("model", MODELS)
@pytest.mark.parametrize("max_tokens", [5])
@pytest.mark.parametrize("tensor_parallel_size", TENSOR_PARALLEL_SIZES)
def test_models(
def test_basic(
vllm_runner: type[VllmRunner],
monkeypatch: pytest.MonkeyPatch,
model: str,
@@ -58,4 +58,5 @@ def test_models(
vllm_outputs = vllm_model.generate_greedy(example_prompts,
max_tokens)
output = vllm_outputs[0][1]
assert "1024" in output
assert "1024" in output or "0, 1" in output