[CI] test chunked prefill more (#5798)
This commit is contained in:
@@ -15,13 +15,13 @@ from sglang.test.test_utils import (
|
||||
CustomTestCase,
|
||||
is_in_ci,
|
||||
popen_launch_server,
|
||||
run_bench_one_batch,
|
||||
run_bench_offline_throughput,
|
||||
)
|
||||
|
||||
|
||||
class TestTritonAttnBackend(CustomTestCase):
|
||||
def test_latency(self):
|
||||
output_throughput = run_bench_one_batch(
|
||||
output_throughput = run_bench_offline_throughput(
|
||||
DEFAULT_MODEL_NAME_FOR_TEST,
|
||||
[
|
||||
"--attention-backend",
|
||||
@@ -32,6 +32,8 @@ class TestTritonAttnBackend(CustomTestCase):
|
||||
],
|
||||
)
|
||||
|
||||
print(f"{output_throughput=}")
|
||||
|
||||
if is_in_ci():
|
||||
self.assertGreater(output_throughput, 153)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user