[Misc] Remove flashinfer warning, add flashinfer tests to CI (#6351)

.buildkite/test-pipeline.yaml

@@ -19,8 +19,10 @@ steps:
 - label: Basic Correctness Test
   mirror_hardwares: [amd]
   commands:
+  - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.8/flashinfer-0.0.8+cu121torch2.3-cp310-cp310-linux_x86_64.whl
   - VLLM_ATTENTION_BACKEND=XFORMERS pytest -v -s basic_correctness/test_basic_correctness.py
   - VLLM_ATTENTION_BACKEND=FLASH_ATTN pytest -v -s basic_correctness/test_basic_correctness.py
+  - VLLM_ATTENTION_BACKEND=FLASHINFER pytest -v -s basic_correctness/test_basic_correctness.py
   - VLLM_ATTENTION_BACKEND=XFORMERS pytest -v -s basic_correctness/test_chunked_prefill.py
   - VLLM_ATTENTION_BACKEND=FLASH_ATTN pytest -v -s basic_correctness/test_chunked_prefill.py
   - VLLM_TEST_ENABLE_ARTIFICIAL_PREEMPT=1 pytest -v -s basic_correctness/test_preemption.py
@@ -122,14 +124,14 @@ steps:
 - label: Kernels Test %N
   #mirror_hardwares: [amd]
   commands:
-    - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.7/flashinfer-0.0.7+cu121torch2.3-cp310-cp310-linux_x86_64.whl
+    - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.8/flashinfer-0.0.8+cu121torch2.3-cp310-cp310-linux_x86_64.whl
     - pytest -v -s kernels --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
   parallelism: 4
 
 - label: Models Test
   #mirror_hardwares: [amd]
   commands:
-    - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.7/flashinfer-0.0.7+cu121torch2.3-cp310-cp310-linux_x86_64.whl
+    - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.8/flashinfer-0.0.8+cu121torch2.3-cp310-cp310-linux_x86_64.whl
     - pytest -v -s models -m \"not vlm\"
 
 - label: Vision Language Models Test
@@ -240,7 +242,7 @@ steps:
   - pytest -v -s distributed/test_custom_all_reduce.py
   - 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
-  - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.7/flashinfer-0.0.7+cu121torch2.3-cp310-cp310-linux_x86_64.whl
+  - pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.8/flashinfer-0.0.8+cu121torch2.3-cp310-cp310-linux_x86_64.whl
   - VLLM_ATTENTION_BACKEND=FLASHINFER TEST_DIST_MODEL=facebook/opt-125m DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_basic_distributed_correctness.py
   - VLLM_ATTENTION_BACKEND=FLASHINFER TEST_DIST_MODEL=meta-llama/Meta-Llama-3-8B DISTRIBUTED_EXECUTOR_BACKEND=ray pytest -v -s distributed/test_basic_distributed_correctness.py
   - pytest -v -s -x lora/test_mixtral.py

tests/basic_correctness/test_basic_correctness.py

@@ -2,11 +2,13 @@
 
 Run `pytest tests/basic_correctness/test_basic_correctness.py`.
 """
+import os
 import weakref
 
 import pytest
 
 from vllm import LLM
+from vllm.utils import is_hip
 
 from ..models.utils import check_outputs_equal
 
@@ -26,6 +28,9 @@ def test_vllm_gc_ed():
     assert weak_llm() is None
 
 
+@pytest.mark.skipif(is_hip()
+                    and os.getenv("VLLM_ATTENTION_BACKEND") == "FLASHINFER",
+                    reason="Flashinfer does not support ROCm/HIP.")
 @pytest.mark.parametrize("model", MODELS)
 @pytest.mark.parametrize("dtype", ["half"])
 @pytest.mark.parametrize("max_tokens", [5])

vllm/attention/selector.py

@@ -77,9 +77,6 @@ def get_attn_backend(
         return IpexAttnBackend
     elif backend == _Backend.FLASHINFER:
         logger.info("Using Flashinfer backend.")
-        logger.warning(("Flashinfer will be stuck on llama-2-7b,"
-                        " please avoid using Flashinfer as the "
-                        "backend when running on llama-2-7b."))
         from vllm.attention.backends.flashinfer import FlashInferBackend
         return FlashInferBackend
     elif backend == _Backend.PALLAS: