diff --git a/tests/v1/e2e/test_spec_decode.py b/tests/v1/e2e/test_spec_decode.py index ffd9f3e0370f7..4a6b84ae4817c 100644 --- a/tests/v1/e2e/test_spec_decode.py +++ b/tests/v1/e2e/test_spec_decode.py @@ -433,9 +433,9 @@ def test_eagle_correctness( print(f"ref_output: {ref_output.outputs[0].text}") print(f"spec_output: {spec_output.outputs[0].text}") - # Heuristic: expect at least 66% of the prompts to match exactly + # Heuristic: expect at least 60% of the prompts to match exactly # Upon failure, inspect the outputs to check for inaccuracy. - assert matches > int(0.66 * len(ref_outputs)) + assert matches > int(0.6 * len(ref_outputs)) del spec_llm torch.cuda.empty_cache() cleanup_dist_env_and_memory()