A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 11, 17:30 UTC
[CI] Reduce Flakiness For test_spec_decode.py::test_suffix_decoding_acceptance (#30367)
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
Micah Williamson committed
Dec 9, 2025 at 20:35 UTC
7d80c73d4277187d0468f15a22bba959ce853261
1 file changed
+2
-2
tests/v1/e2e/test_spec_decode.py
+2
-2
@@ -191,8 +191,8 @@ def test_suffix_decoding_acceptance(
191
# Expect the acceptance rate to improve.
192
assert first_accept_rate < last_accept_rate
193
194
- # Heuristic: expect at least 82.5% acceptance rate at the end.
195
- assert last_accept_rate > 0.825
194
+ # Heuristic: expect at least 80.0% acceptance rate at the end.
195
+ assert last_accept_rate > 0.80
196
197
del spec_llm
198
torch.cuda.empty_cache()