diff --git a/models/demos/qwen/tests/test_qwen_perf.py b/models/demos/qwen/tests/test_qwen_perf.py index f5a9d2d0714..b67863b1889 100644 --- a/models/demos/qwen/tests/test_qwen_perf.py +++ b/models/demos/qwen/tests/test_qwen_perf.py @@ -51,7 +51,7 @@ def test_qwen_model_perf(mesh_device, kv_cache_len, expected_compile_time, use_p model_args = TtModelArgs(mesh_device) tokenizer = Tokenizer(model_args.tokenizer_path) - expected_inference_time = 0.07 + expected_inference_time = 0.08 # model_args.n_layers = 1 # Clear global profiler state before starting measurements