Jiangyun ZhuandSamit
98fe15762d
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com> Signed-off-by: samithuang <285365963@qq.com> Co-authored-by: Samit <285365963@qq.com>
79 lines
1.9 KiB
Python
79 lines
1.9 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
|
|
from abc import ABC, abstractmethod
|
|
from dataclasses import dataclass
|
|
from typing import Generic, TypeVar
|
|
|
|
import torch
|
|
|
|
|
|
class AttentionBackend(ABC):
|
|
"""Abstract class for diffusion attention backends."""
|
|
|
|
accept_output_buffer: bool = False
|
|
|
|
@staticmethod
|
|
@abstractmethod
|
|
def get_name() -> str:
|
|
raise NotImplementedError
|
|
|
|
@staticmethod
|
|
@abstractmethod
|
|
def get_impl_cls() -> type["AttentionImpl"]:
|
|
raise NotImplementedError
|
|
|
|
@staticmethod
|
|
@abstractmethod
|
|
def get_metadata_cls() -> type["AttentionMetadata"]:
|
|
raise NotImplementedError
|
|
|
|
@staticmethod
|
|
@abstractmethod
|
|
def get_builder_cls(): # -> Type["AttentionMetadataBuilder"]:
|
|
raise NotImplementedError
|
|
|
|
@staticmethod
|
|
@abstractmethod
|
|
def get_supported_head_sizes() -> list[int]:
|
|
"""Get the list of supported head sizes for this backend."""
|
|
raise NotImplementedError
|
|
|
|
@classmethod
|
|
def supports_head_size(cls, head_size: int) -> bool:
|
|
supported_head_sizes = cls.get_supported_head_sizes()
|
|
return (not supported_head_sizes) or head_size in supported_head_sizes
|
|
|
|
|
|
@dataclass
|
|
class AttentionMetadata:
|
|
attn_mask: torch.Tensor | None = None
|
|
|
|
|
|
T = TypeVar("T", bound=AttentionMetadata)
|
|
|
|
|
|
class AttentionImpl(ABC, Generic[T]):
|
|
@abstractmethod
|
|
def __init__(
|
|
self,
|
|
num_heads: int,
|
|
head_size: int,
|
|
softmax_scale: float,
|
|
causal: bool = False,
|
|
num_kv_heads: int | None = None,
|
|
prefix: str = "",
|
|
**extra_impl_args,
|
|
) -> None:
|
|
raise NotImplementedError
|
|
|
|
@abstractmethod
|
|
def forward(
|
|
self,
|
|
query: torch.Tensor,
|
|
key: torch.Tensor,
|
|
value: torch.Tensor,
|
|
attn_metadata: T,
|
|
) -> torch.Tensor:
|
|
raise NotImplementedError
|