tests: fix some typo and limitation on test cases (#4989)

Signed-off-by: Ivy Zhang <[email protected]>

Author: crazydemo

tests/integration/defs/examples/test_gemma.py

@@ -91,6 +91,7 @@ def test_llm_hf_gemma_quantization_1gpu_vswa(batch_size, data_type,
                                              gemma_example_root,
                                              llm_datasets_root, llm_rouge_root,
                                              qformat):
+    skip_fp8_pre_ada(use_fp8=qformat == "fp8")
     max_attention_window = VSWA_ATTENTION[Path(gemma_model_root).stem]
     hf_gemma_quantization_1gpu(batch_size, data_type, gemma_model_root,
                                llm_venv, cmodel_dir, engine_dir,

tests/integration/defs/test_e2e.py

@@ -829,7 +829,7 @@ def test_trtllm_bench_mgmn(llm_root, llm_venv):
                                      delete_on_close=True) as running_log:
         check_call(benchmark_cmd,
                    shell=True,
-                   running_log=running_log,
+                   stdout=running_log,
                    env=llm_venv._new_env)
         _check_mem_usage(running_log, [30, 0, 0, 0])