[Attention] Move Backend enum into registry (#25893)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Matthew Bonanni
2025-10-02 20:32:24 -07:00
committed by GitHub
parent ad2d788016
commit 2aaa423842
31 changed files with 99 additions and 66 deletions
+1 -1
View File
@@ -9,6 +9,7 @@ import torch.nn.functional as F
from torch.nn import LayerNorm
from transformers.models.qwen2_vl import Qwen2VLProcessor
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.config import VllmConfig
from vllm.distributed import utils as dist_utils
@@ -38,7 +39,6 @@ from vllm.model_executor.models.utils import (AutoWeightsLoader, WeightsMapper,
from vllm.model_executor.models.vision import get_vit_attn_backend
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.inputs import MultiModalDataDict
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors
from vllm.transformers_utils.configs.dotsocr import (DotsOCRConfig,
DotsVisionConfig)
+2 -1
View File
@@ -34,6 +34,7 @@ import torch.nn.functional as F
from einops import rearrange, repeat
from transformers import BatchFeature
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.config import VllmConfig
from vllm.distributed import parallel_state
@@ -54,7 +55,7 @@ from vllm.multimodal.processing import (BaseMultiModalProcessor,
BaseProcessingInfo, PromptReplacement,
PromptUpdate)
from vllm.multimodal.profiling import BaseDummyInputsBuilder
from vllm.platforms import _Backend, current_platform
from vllm.platforms import current_platform
from vllm.sequence import IntermediateTensors
from vllm.utils.tensor_schema import TensorSchema, TensorShape
+1 -1
View File
@@ -46,6 +46,7 @@ from transformers.models.glm4v.video_processing_glm4v import (
Glm4vVideoProcessor)
from transformers.video_utils import VideoMetadata
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.config import VllmConfig
from vllm.distributed import (get_tensor_model_parallel_world_size,
@@ -69,7 +70,6 @@ from vllm.multimodal.processing import (BaseMultiModalProcessor,
BaseProcessingInfo, PromptReplacement,
PromptUpdate, PromptUpdateDetails)
from vllm.multimodal.profiling import BaseDummyInputsBuilder
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors
from vllm.utils.tensor_schema import TensorSchema, TensorShape
+1 -1
View File
@@ -17,6 +17,7 @@ from transformers.modeling_outputs import (BaseModelOutput,
BaseModelOutputWithPooling)
from transformers.utils import torch_int
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.config import VllmConfig
from vllm.distributed import get_tensor_model_parallel_world_size
@@ -39,7 +40,6 @@ from vllm.multimodal.processing import (BaseMultiModalProcessor,
BaseProcessingInfo, PromptReplacement,
PromptUpdate)
from vllm.multimodal.profiling import BaseDummyInputsBuilder
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors
from vllm.utils import is_list_of
from vllm.utils.tensor_schema import TensorSchema, TensorShape
+1 -1
View File
@@ -38,6 +38,7 @@ from transformers.models.qwen2_5_vl import Qwen2_5_VLProcessor
from transformers.models.qwen2_5_vl.configuration_qwen2_5_vl import (
Qwen2_5_VLConfig, Qwen2_5_VLVisionConfig)
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.config import VllmConfig
from vllm.distributed import parallel_state
@@ -62,7 +63,6 @@ from vllm.multimodal.evs import (compute_mrope_for_media,
from vllm.multimodal.inputs import MultiModalFieldConfig, MultiModalKwargs
from vllm.multimodal.parse import MultiModalDataItems
from vllm.multimodal.processing import PromptReplacement, PromptUpdate
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors
from vllm.utils import is_pin_memory_available
from vllm.utils.tensor_schema import TensorSchema, TensorShape
+1 -1
View File
@@ -41,6 +41,7 @@ from transformers.models.qwen2_vl.image_processing_qwen2_vl import smart_resize
from transformers.models.qwen2_vl.video_processing_qwen2_vl import (
Qwen2VLVideoProcessor)
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.config import VllmConfig
from vllm.distributed import parallel_state, tensor_model_parallel_all_gather
@@ -65,7 +66,6 @@ from vllm.multimodal.processing import (BaseMultiModalProcessor,
BaseProcessingInfo, PromptReplacement,
PromptUpdate)
from vllm.multimodal.profiling import BaseDummyInputsBuilder
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors
from vllm.transformers_utils.tokenizer import AnyTokenizer
from vllm.utils.tensor_schema import TensorSchema, TensorShape
+1 -1
View File
@@ -43,6 +43,7 @@ from transformers.models.qwen3_vl.video_processing_qwen3_vl import (
smart_resize as video_smart_resize)
from transformers.video_utils import VideoMetadata
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.compilation.decorators import support_torch_compile
from vllm.config import VllmConfig
@@ -66,7 +67,6 @@ from vllm.multimodal.processing import (BaseMultiModalProcessor,
PromptReplacement, PromptUpdate,
PromptUpdateDetails)
from vllm.multimodal.profiling import BaseDummyInputsBuilder
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors
from vllm.utils import is_list_of
+1 -1
View File
@@ -13,6 +13,7 @@ from torch.nn import functional as F
from transformers import Siglip2VisionConfig
from transformers.configuration_utils import PretrainedConfig
from vllm.attention.backends.registry import _Backend
from vllm.attention.layer import check_upstream_fa_availability
from vllm.distributed import divide, get_tensor_model_parallel_world_size
from vllm.model_executor.layers.activation import get_act_fn
@@ -22,7 +23,6 @@ from vllm.model_executor.layers.linear import (ColumnParallelLinear,
RowParallelLinear)
from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.platforms import _Backend
from .vision import get_vit_attn_backend
+2 -1
View File
@@ -10,11 +10,12 @@ from typing import (Callable, Final, Generic, Literal, Optional, Protocol,
import torch
from transformers import PretrainedConfig
from vllm.attention.backends.registry import _Backend
from vllm.distributed import (get_tensor_model_parallel_rank,
get_tensor_model_parallel_world_size,
tensor_model_parallel_all_gather)
from vllm.logger import init_logger
from vllm.platforms import _Backend, current_platform
from vllm.platforms import current_platform
logger = init_logger(__name__)