[3/N][Attention] Move AttentionMetadata-related code from utils.py to backend.py (#32054)

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

vllm/model_executor/layers/attention/chunked_local_attention.py

@@ -8,11 +8,13 @@ from vllm.attention.layer import Attention
 from vllm.config import CacheConfig
 from vllm.config.vllm import VllmConfig
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.v1.attention.backend import AttentionBackend
-from vllm.v1.attention.backends.utils import (
+from vllm.v1.attention.backend import (
+    AttentionBackend,
     AttentionCGSupport,
     AttentionMetadataBuilder,
     CommonAttentionMetadata,
+)
+from vllm.v1.attention.backends.utils import (
     make_local_attention_virtual_batches,
     subclass_attention_backend,
 )

vllm/model_executor/layers/attention/cross_attention.py

@@ -14,9 +14,9 @@ from vllm.v1.attention.backend import (
     AttentionBackend,
     AttentionMetadata,
     AttentionType,
+    CommonAttentionMetadata,
 )
 from vllm.v1.attention.backends.utils import (
-    CommonAttentionMetadata,
     subclass_attention_backend,
 )
 from vllm.v1.attention.selector import get_attn_backend

vllm/model_executor/layers/attention/encoder_only_attention.py

@@ -12,9 +12,9 @@ from vllm.v1.attention.backend import (
     AttentionBackend,
     AttentionMetadata,
     AttentionType,
+    CommonAttentionMetadata,
 )
 from vllm.v1.attention.backends.utils import (
-    CommonAttentionMetadata,
     subclass_attention_backend,
 )
 from vllm.v1.attention.selector import get_attn_backend

vllm/model_executor/layers/attention/static_sink_attention.py

@@ -15,9 +15,9 @@ from vllm.v1.attention.backend import (
     AttentionBackend,
     AttentionMetadata,
     AttentionType,
+    CommonAttentionMetadata,
 )
 from vllm.v1.attention.backends.utils import (
-    CommonAttentionMetadata,
     subclass_attention_backend,
 )
 from vllm.v1.attention.ops.triton_reshape_and_cache_flash import (