EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[CI] Fix disaggregation failure tolerance CI (#9378)

Browse Source 
Signed-off-by: Shangming Cai <csmthu@gmail.com>
This commit is contained in:
Shangming Cai
2025-08-20 14:31:08 +08:00
committed by GitHub
parent f1b0eda55c
commit d8ed60f254
1 changed files with 15 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
test/srt/test_disaggregation.py
View File

@@ -323,9 +323,22 @@ class TestDisaggregationMooncakeFailure(CustomTestCase):
host=f"http://{self.base_host}",
port=int(self.lb_port),
)
metrics = run_eval_few_shot_gsm8k(args)
print(f"Evaluation metrics: {metrics}")
# Expect lots of failure but the server cannot crash
try:
metrics = run_eval_few_shot_gsm8k(args)
print(f"Evaluation metrics: {metrics}")
except Exception as e:
print(f"Test encountered expected errors: {e}")
# Check if servers are still healthy
try:
response = requests.get(self.prefill_url + "/health_generate")
assert response.status_code == 200
response = requests.get(self.decode_url + "/health_generate")
assert response.status_code == 200
except Exception as health_check_error:
# If health check fails, re-raise the original exception
raise e from health_check_error
class TestDisaggregationMooncakeSpec(CustomTestCase):