diff --git a/tests/integration/test_lists/waives.txt b/tests/integration/test_lists/waives.txt index 0cfeddceaab..e92a95f78de 100644 --- a/tests/integration/test_lists/waives.txt +++ b/tests/integration/test_lists/waives.txt @@ -328,7 +328,6 @@ accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus_python perf/test_perf_sanity.py::test_e2e[aggr_upload-deepseek_v32_fp4_blackwell-v32_fp4_tep8_mtp3_8k1k] SKIP (https://nvbugs/5997092) accuracy/test_llm_api_pytorch.py::TestQwen3_5_35B_A3B::test_fp8 SKIP (https://nvbugs/6004530) unittest/_torch/modules/moe/test_moe_module.py::test_configurable_moe_multi_gpu[parallel=DEP-comm=DEEPEP-e60_k4_h2048_i1408-seq=8-dtype=torch.bfloat16-backend=TRTLLM-quant=NVFP4-routing=Renormalize] SKIP (https://nvbugs/6007285) -disaggregated/test_disaggregated.py::test_disaggregated_gpt_oss_120b_harmony[gpt_oss/gpt-oss-120b] SKIP (https://nvbugs/6011317) accuracy/test_llm_api_pytorch.py::TestNemotronV3Super::test_bf16_4gpu_mtp_ar SKIP (https://nvbugs/5959992) accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_eagle3_vswa_reuse_4gpus[two_model] SKIP (https://nvbugs/6013562) accuracy/test_llm_api_pytorch.py::TestQwen3_8B::test_bf16[latency] SKIP (https://nvbugs/6012526) diff --git a/tests/unittest/conftest.py b/tests/unittest/conftest.py index 65cfa85a75c..b3df6951f2f 100644 --- a/tests/unittest/conftest.py +++ b/tests/unittest/conftest.py @@ -433,6 +433,8 @@ def process_gpu_memory_info_available(): @pytest.fixture(scope="function") def setup_ray_cluster() -> Generator[int, None, None]: + import time + runtime_env = { "env_vars": { "RAY_EXPERIMENTAL_NOSET_CUDA_VISIBLE_DEVICES": "1" @@ -448,6 +450,8 @@ def setup_ray_cluster() -> Generator[int, None, None]: ray.init(address="local", **ray_init_args) gcs_addr = ray.get_runtime_context().gcs_address port = int(gcs_addr.split(":")[1]) + # Allow raylet to complete GCS registration before tests create actors. + time.sleep(2) yield port finally: if ray.is_initialized():