diff --git a/.buildkite/test-amd.yaml b/.buildkite/test-amd.yaml index 2d7542226a..c598a37545 100644 --- a/.buildkite/test-amd.yaml +++ b/.buildkite/test-amd.yaml @@ -1996,7 +1996,8 @@ steps: - label: Extract Hidden States Integration # TBD timeout_in_minutes: 180 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300] - agent_pool: mi300_1 + agent_pool: mi300_2 + num_gpus: 2 working_dir: "/vllm-workspace/tests" source_file_dependencies: - vllm/config/speculative.py diff --git a/tests/v1/kv_connector/extract_hidden_states_integration/test_extraction.py b/tests/v1/kv_connector/extract_hidden_states_integration/test_extraction.py index 390519fb55..4a84ed0314 100644 --- a/tests/v1/kv_connector/extract_hidden_states_integration/test_extraction.py +++ b/tests/v1/kv_connector/extract_hidden_states_integration/test_extraction.py @@ -12,6 +12,7 @@ from vllm import LLM, ModelRegistry, SamplingParams from vllm.distributed.kv_transfer.kv_connector.v1 import ( example_hidden_states_connector, ) +from vllm.platforms import current_platform def get_and_check_output(output, expected_shape): @@ -296,7 +297,7 @@ def test_extract_hidden_states_qwen35_hybrid_smoke(tmp_path): ) -@pytest.mark.timeout(60) +@pytest.mark.timeout(240 if current_platform.is_rocm() else 60) @multi_gpu_test(num_gpus=2) @create_new_process_for_each_test() def test_extract_hidden_states_tp2():