[Attention] Update attention imports (#29540)

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

vllm/forward_context.py

@@ -5,19 +5,17 @@ import time
 from collections import defaultdict
 from contextlib import contextmanager
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Any, NamedTuple
+from typing import Any, NamedTuple
 
 import torch
 
 import vllm.envs as envs
+from vllm.attention.backends.abstract import AttentionMetadata
 from vllm.config import CUDAGraphMode, ParallelConfig, VllmConfig
 from vllm.logger import init_logger
 from vllm.v1.worker.dp_utils import coordinate_batch_across_dp
 from vllm.v1.worker.ubatch_utils import UBatchSlices
 
-if TYPE_CHECKING:
-    from vllm.attention.backends.abstract import AttentionMetadata
-
 logger = init_logger(__name__)
 
 track_batchsize: bool = envs.VLLM_LOG_BATCHSIZE_INTERVAL >= 0
@@ -195,7 +193,7 @@ class ForwardContext:
     for each microbatch.
     Set dynamically for each forward pass
     """
-    attn_metadata: dict[str, "AttentionMetadata"] | list[dict[str, "AttentionMetadata"]]
+    attn_metadata: dict[str, AttentionMetadata] | list[dict[str, AttentionMetadata]]
     # TODO: remove after making all virtual_engines share the same kv cache
     virtual_engine: int  # set dynamically for each forward pass
     # set dynamically for each forward pass