[3/N][Attention] Move AttentionMetadata-related code from utils.py to backend.py (#32054)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
@@ -23,10 +23,9 @@ from vllm.utils.torch_utils import (
|
||||
is_torch_equal_or_newer,
|
||||
set_random_seed,
|
||||
)
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
from vllm.v1.attention.backend import AttentionType, CommonAttentionMetadata
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.v1.attention.backends.utils import (
|
||||
CommonAttentionMetadata,
|
||||
set_kv_cache_layout,
|
||||
)
|
||||
from vllm.v1.kv_cache_interface import FullAttentionSpec
|
||||
|
||||
@@ -22,10 +22,10 @@ from vllm.config.vllm import set_current_vllm_config
|
||||
from vllm.model_executor.layers.attention_layer_base import AttentionLayerBase
|
||||
from vllm.utils.math_utils import cdiv
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
from vllm.v1.attention.backend import CommonAttentionMetadata
|
||||
from vllm.v1.attention.backends.fa_utils import flash_attn_supports_mla
|
||||
from vllm.v1.attention.backends.mla.common import QueryLenSupport
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.v1.attention.backends.utils import CommonAttentionMetadata
|
||||
from vllm.v1.attention.ops.flashmla import is_flashmla_dense_supported
|
||||
from vllm.v1.kv_cache_interface import FullAttentionSpec
|
||||
|
||||
|
||||
@@ -18,12 +18,12 @@ from vllm.config import (
|
||||
VllmConfig,
|
||||
)
|
||||
from vllm.config.model import ModelDType
|
||||
from vllm.v1.attention.backend import AttentionImpl
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.v1.attention.backends.utils import (
|
||||
from vllm.v1.attention.backend import (
|
||||
AttentionImpl,
|
||||
AttentionMetadataBuilder,
|
||||
CommonAttentionMetadata,
|
||||
)
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.v1.kv_cache_interface import FullAttentionSpec
|
||||
|
||||
|
||||
|
||||
@@ -19,7 +19,7 @@ def sync_tracker():
|
||||
Fixture that patches CommonAttentionMetadata.seq_lens_cpu to detect
|
||||
lazy init syncs. Prints stack traces immediately when syncs occur.
|
||||
"""
|
||||
from vllm.v1.attention.backends.utils import CommonAttentionMetadata
|
||||
from vllm.v1.attention.backend import CommonAttentionMetadata
|
||||
|
||||
# Shared counter for cross-process communication (inherited by fork)
|
||||
sync_count = multiprocessing.Value("i", 0)
|
||||
|
||||
@@ -12,9 +12,9 @@ from tests.v1.attention.utils import (
|
||||
try_get_attention_backend,
|
||||
)
|
||||
from vllm.config import ParallelConfig, SpeculativeConfig
|
||||
from vllm.v1.attention.backend import CommonAttentionMetadata
|
||||
from vllm.v1.attention.backends.fa_utils import is_flash_attn_varlen_func_available
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.v1.attention.backends.utils import CommonAttentionMetadata
|
||||
|
||||
if not is_flash_attn_varlen_func_available():
|
||||
pytest.skip(
|
||||
|
||||
Reference in New Issue
Block a user