fix: kimi k2 thinking accuracy threshold change (#16585)
This commit is contained in:
@@ -42,7 +42,7 @@ class TestKimiK2Unified(unittest.TestCase):
|
|||||||
run_combined_tests(
|
run_combined_tests(
|
||||||
models=variants,
|
models=variants,
|
||||||
test_name="Kimi-K2-Thinking Unified",
|
test_name="Kimi-K2-Thinking Unified",
|
||||||
accuracy_params=AccuracyTestParams(dataset="gsm8k", baseline_accuracy=0.95),
|
accuracy_params=AccuracyTestParams(dataset="gsm8k", baseline_accuracy=0.94),
|
||||||
performance_params=PerformanceTestParams(
|
performance_params=PerformanceTestParams(
|
||||||
profile_dir="performance_profiles_kimi_k2_thinking",
|
profile_dir="performance_profiles_kimi_k2_thinking",
|
||||||
),
|
),
|
||||||
|
|||||||
Reference in New Issue
Block a user