Add basic correctness 2 GPU tests to 4 GPU pipeline (vllm-project#5518)

.buildkite/test-pipeline.yaml

@@ -32,7 +32,7 @@ steps:
   working_dir: "/vllm-workspace/tests"
   num_gpus: 2
 
-- label: Distributed Tests
+- label: Distributed Tests (2 GPUs)
   mirror_hardwares: [amd]
   working_dir: "/vllm-workspace/tests"
   num_gpus: 2
@@ -50,12 +50,16 @@ steps:
   - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s test_sharded_state_loader.py
   - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s distributed/test_utils.py
 
-- label: Distributed Tests (Multiple Groups)
+- label: Distributed Tests (4 GPUs)
   #mirror_hardwares: [amd]
   working_dir: "/vllm-workspace/tests"
   num_gpus: 4
   commands:
   - pytest -v -s distributed/test_pynccl.py
+  # We want to test that models which use 2 GPUs work with 4 GPUs, which is why we duplicate them here.
+  # See https://github.com/vllm-project/vllm/pull/5473#issuecomment-2166601837 for context.
+  - TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_basic_distributed_correctness.py
+  - TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=mp pytest -v -s distributed/test_basic_distributed_correctness.py
 
 - label: Engine Test
   mirror_hardwares: [amd]