[Attention] Update tests to remove deprecated env vars (#30563)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>

tests/v1/kv_connector/unit/test_nixl_connector.py

@@ -1132,7 +1132,7 @@ def _run_abort_timeout_test(llm: LLM, timeout: int):
         "TRITON_ATTN",
     ],
 )
-def test_register_kv_caches(dist_init, attn_backend, monkeypatch):
+def test_register_kv_caches(dist_init, attn_backend):
     """
     Test that register_kv_caches() properly calls nixl_wrapper methods with
     correct data.
@@ -1144,9 +1144,7 @@ def test_register_kv_caches(dist_init, attn_backend, monkeypatch):
        block layout info
     """
 
-    monkeypatch.setenv("VLLM_ATTENTION_BACKEND", attn_backend)
-
-    vllm_config = create_vllm_config()
+    vllm_config = create_vllm_config(attention_backend=attn_backend)
 
     # Import the appropriate backend based on the parameter
     if attn_backend == "FLASH_ATTN":

tests/v1/kv_connector/unit/utils.py

@@ -11,6 +11,7 @@ import torch
 
 from vllm import SamplingParams
 from vllm.config import (
+    AttentionConfig,
     CacheConfig,
     DeviceConfig,
     KVTransferConfig,
@@ -94,6 +95,7 @@ def create_vllm_config(
     dtype: str = "float16",
     cache_dtype: str = "auto",
     hf_overrides: dict[str, Any] | None = None,
+    attention_backend: str | None = None,
 ) -> VllmConfig:
     """Initialize VllmConfig For Testing."""
     model_config = ModelConfig(
@@ -124,12 +126,14 @@ def create_vllm_config(
         enable_permute_local_kv=enable_permute_local_kv,
         kv_connector_extra_config=kv_connector_extra_config or {},
     )
+    attention_config = AttentionConfig(backend=attention_backend)
     return VllmConfig(
         scheduler_config=scheduler_config,
         model_config=model_config,
         cache_config=cache_config,
         kv_transfer_config=kv_transfer_config,
         device_config=DeviceConfig("cpu"),
+        attention_config=attention_config,
     )