[Observability] Add startup, memory, and hybrid SWA diagnostics (#33375)
This commit is contained in:
@@ -77,12 +77,12 @@ class TestPrefillCudaGraphRunnerChunkedPrefix(CustomTestCase):
|
||||
"check_cuda_graph_backend",
|
||||
return_value=False,
|
||||
):
|
||||
runner = capture_prefill_graph(
|
||||
capture = capture_prefill_graph(
|
||||
model_runner=model_runner,
|
||||
eager_runner=eager_runner,
|
||||
)
|
||||
|
||||
self.assertIs(runner, eager_runner)
|
||||
self.assertIs(capture.runner, eager_runner)
|
||||
|
||||
def test_prefix_chunk_capacity_is_aggregate_and_can_be_overridden(self):
|
||||
model_runner = SimpleNamespace(
|
||||
|
||||
Reference in New Issue
Block a user