# Copyright (c) Facebook, Inc. and its affiliates. All rights reserved. # # This source code is licensed under the BSD license found in the # LICENSE file in the root directory of this source tree. from dataclasses import replace from enum import Enum from typing import Any, Iterable, List, Mapping, Optional, Set, Tuple, Union import torch from ..common import get_operator, register_operator from . import attn_bias from .attn_bias import ( AttentionBias, AttentionBiasSubTensor, BlockDiagonalCausalLocalAttentionFromBottomRightMask, BlockDiagonalCausalLocalAttentionMask, BlockDiagonalCausalMask, BlockDiagonalCausalWithOffsetGappyKeysMask, BlockDiagonalCausalWithOffsetPaddedKeysMask, BlockDiagonalGappyKeysMask, BlockDiagonalMask, BlockDiagonalPaddedKeysMask, LowerTriangularFromBottomRightLocalAttentionMask, LowerTriangularFromBottomRightMask, LowerTriangularMask, LowerTriangularMaskWithTensorBias, PagedBlockDiagonalCausalWithOffsetPaddedKeysMask, PagedBlockDiagonalGappyKeysMask, PagedBlockDiagonalPaddedKeysMask, ) from .common import ( AttentionBwOpBase, AttentionFwOpBase, Context, Gradients, Inputs, check_lastdim_alignment_stride1, ) def _minimum_gemm_alignment(inp: Inputs) -> int: return 1 def _get_seqlen_info( inp: Inputs, ) -> Tuple[Optional[torch.Tensor], Optional[torch.Tensor], int, int]: attn_bias = inp.attn_bias if isinstance( attn_bias, ( BlockDiagonalMask, BlockDiagonalPaddedKeysMask, BlockDiagonalGappyKeysMask, PagedBlockDiagonalPaddedKeysMask, PagedBlockDiagonalGappyKeysMask, ), ): attn_bias.k_seqinfo.to(inp.query.device) attn_bias.q_seqinfo.to(inp.query.device) seqstart_k = attn_bias.k_seqinfo.seqstart seqstart_q = attn_bias.q_seqinfo.seqstart max_seqlen_q = attn_bias.q_seqinfo.max_seqlen max_seqlen_k = attn_bias.k_seqinfo.max_seqlen else: seqstart_k = None seqstart_q = None max_seqlen_q = -1 max_seqlen_k = -1 return seqstart_k, seqstart_q, max_seqlen_q, max_seqlen_k def _get_tensor_bias( attn_bias: Optional[Union[torch.Tensor, AttentionBias]] ) -> Optional[torch.Tensor]: if isinstance(attn_bias, AttentionBiasSubTensor): if isinstance(attn_bias, LowerTriangularMaskWithTensorBias): return attn_bias._subtensor elif isinstance(attn_bias, torch.Tensor): return attn_bias return None def _check_bias_alignment( reasons: List[str], attn_bias: Optional[Union[torch.Tensor, AttentionBias]] ) -> None: attn_bias_tensor = _get_tensor_bias(attn_bias) if attn_bias_tensor is not None: alignment = 128 // torch.finfo(attn_bias_tensor.dtype).bits show_padding_hint = False for d in range(attn_bias_tensor.ndim - 1): if attn_bias_tensor.stride(d) % alignment != 0: reasons.append( f"attn_bias.stride(-2) % {alignment} != 0 (attn_bias.stride() = {attn_bias_tensor.stride()})" ) show_padding_hint = True if show_padding_hint: reasons.append( """\ HINT: To use an `attn_bias` with a sequence length that is not a multiple of 8, \ you need to ensure memory is aligned by slicing a bigger tensor. \ Example: use `attn_bias = torch.zeros([1, 1, 5, 8])[:,:,:,:5]` instead of `torch.zeros([1, 1, 5, 5])`""" ) # We can have stride=0 sometimes if dimension=1 if attn_bias_tensor.stride(-1) > 1: reasons.append( f"attn_bias.stride(-1) > 1 (attn_bias.stride() = {attn_bias_tensor.stride()}) - " "you should call `.contiguous()` on the bias" ) class _CustomMaskType(int, Enum): """ (Matches CustomMaskType in C++.) """ NoCustomMask = 0 CausalFromTopLeft = 1 CausalFromBottomRight = 2 def _custom_mask_type(bias: Optional[Union[torch.Tensor, AttentionBias]]) -> int: if isinstance( bias, ( LowerTriangularMask, BlockDiagonalCausalMask, BlockDiagonalCausalLocalAttentionMask, ), ): return int(_CustomMaskType.CausalFromTopLeft) if isinstance( bias, ( LowerTriangularFromBottomRightMask, LowerTriangularFromBottomRightLocalAttentionMask, attn_bias.BlockDiagonalCausalFromBottomRightMask, BlockDiagonalCausalWithOffsetPaddedKeysMask, BlockDiagonalCausalLocalAttentionFromBottomRightMask, PagedBlockDiagonalCausalWithOffsetPaddedKeysMask, ), ): return int(_CustomMaskType.CausalFromBottomRight) return int(_CustomMaskType.NoCustomMask) @register_operator class FwOp(AttentionFwOpBase): """xFormers' MHA kernel based on Composable Kernel.""" OPERATOR = get_operator("xformers", "efficient_attention_forward_ck") SUPPORTED_DEVICES: Set[str] = {"cuda"} SUPPORTED_DTYPES: Set[torch.dtype] = {torch.half, torch.bfloat16} SUPPORTED_MAX_K = 512 SUPPORTED_ATTN_BIAS_TYPES: Iterable[Any] = ( type(None), torch.Tensor, LowerTriangularMask, LowerTriangularFromBottomRightMask, LowerTriangularFromBottomRightLocalAttentionMask, LowerTriangularMaskWithTensorBias, BlockDiagonalMask, BlockDiagonalCausalMask, BlockDiagonalCausalWithOffsetGappyKeysMask, BlockDiagonalCausalWithOffsetPaddedKeysMask, BlockDiagonalGappyKeysMask, BlockDiagonalPaddedKeysMask, attn_bias.BlockDiagonalCausalFromBottomRightMask, attn_bias.BlockDiagonalCausalLocalAttentionMask, BlockDiagonalCausalLocalAttentionFromBottomRightMask, PagedBlockDiagonalPaddedKeysMask, PagedBlockDiagonalCausalWithOffsetPaddedKeysMask, PagedBlockDiagonalGappyKeysMask, ) SUPPORTS_DROPOUT = True SUPPORTS_CUSTOM_SCALE = True SUPPORTS_DIFFERENT_VALUE_EMBED = True SUPPORTS_PARTIAL = True SUPPORTS_BMGHK = True NAME = "ckF" ERROR_ATOL: Mapping[torch.dtype, float] = { torch.float: 3e-4, torch.half: 6e-3, torch.bfloat16: 2.8e-2, } ERROR_RTOL: Mapping[torch.dtype, float] = { torch.float: 2e-5, torch.half: 3e-3, torch.bfloat16: 2e-2, } _TEST_K: List[int] = [ 32, # 64x64 kernel 96, 128, # 64x128 kernel 256, # 64x128 with accumulation in gmem 512, ] @classmethod def apply( cls, inp: Inputs, needs_gradient: bool ) -> Tuple[torch.Tensor, Optional[Context]]: if type(inp.attn_bias) not in FwOp.SUPPORTED_ATTN_BIAS_TYPES: raise NotImplementedError("Unsupported attn_bias type") if inp.query.ndim in [1, 2, 3]: raise NotImplementedError("Unsupported number of dimensions") if inp.query.ndim in [4]: return cls.apply_bmhk(inp, needs_gradient=needs_gradient) assert inp.query.ndim == 5, f"query has shape {inp.query.shape}" ctx: Optional[Context] = None # when the input is expanded 5-D, the group dimension has zero stride if inp.key.stride()[3] == 0: assert ( inp.value.stride()[3] == 0 ), "key and value should be expanded in the same way" k_shape = inp.key.size() k_stride = inp.key.stride() key = inp.key.as_strided( (k_shape[0], k_shape[1], k_shape[2], k_shape[4]), (k_stride[0], k_stride[1], k_stride[2], k_stride[4]), ) v_shape = inp.value.size() v_stride = inp.value.stride() value = inp.value.as_strided( (v_shape[0], v_shape[1], v_shape[2], v_shape[4]), (v_stride[0], v_stride[1], v_stride[2], v_stride[4]), ) else: key = inp.key.flatten(2, 3) value = inp.value.flatten(2, 3) [_, _, G, Hq, _] = inp.query.shape attn_bias_replace = inp.attn_bias if isinstance(inp.attn_bias, LowerTriangularMaskWithTensorBias): bias_tensor = _get_tensor_bias(inp.attn_bias) if bias_tensor is not None and bias_tensor.ndim == 5: attn_bias_replace = LowerTriangularMaskWithTensorBias( bias_tensor.flatten(1, 2) ) elif isinstance(inp.attn_bias, torch.Tensor) and inp.attn_bias.ndim == 5: attn_bias_replace = inp.attn_bias.flatten(1, 2) inp = replace( inp, query=inp.query.flatten(2, 3), key=key, value=value, attn_bias=attn_bias_replace, ) out, ctx = cls.apply_bmhk(inp, needs_gradient=needs_gradient) out = out.unflatten(2, (G, Hq)) if ctx is not None: lse = ctx.lse.unflatten(1, (G, Hq)) ctx = replace(ctx, lse=lse, out=out) return out, ctx @classmethod def apply_bmhk( cls, inp: Inputs, needs_gradient: bool ) -> Tuple[torch.Tensor, Optional[Context]]: if type(inp.attn_bias) not in FwOp.SUPPORTED_ATTN_BIAS_TYPES: raise NotImplementedError("Unsupported attn_bias type") seqstart_k, seqstart_q, max_seqlen_q, _ = _get_seqlen_info(inp) out, lse, rng_seed, rng_offset = cls.OPERATOR( query=inp.query, key=inp.key, value=inp.value, attn_bias=_get_tensor_bias(inp.attn_bias), seqstart_q=seqstart_q, seqstart_k=seqstart_k, max_seqlen_q=max_seqlen_q, dropout_p=inp.p, compute_logsumexp=needs_gradient, custom_mask_type=_custom_mask_type(inp.attn_bias), scale=inp.scale, seqlen_k=( inp.attn_bias.k_seqinfo.seqlen if isinstance( inp.attn_bias, ( BlockDiagonalGappyKeysMask, BlockDiagonalPaddedKeysMask, PagedBlockDiagonalPaddedKeysMask, PagedBlockDiagonalGappyKeysMask, ), ) else None ), window_size=( inp.attn_bias._window_size if isinstance( inp.attn_bias, ( BlockDiagonalCausalLocalAttentionMask, BlockDiagonalCausalLocalAttentionFromBottomRightMask, LowerTriangularFromBottomRightLocalAttentionMask, ), ) else None ), block_tables=( inp.attn_bias.block_tables if isinstance( inp.attn_bias, ( PagedBlockDiagonalPaddedKeysMask, PagedBlockDiagonalGappyKeysMask, ), ) else None ), page_size=( inp.attn_bias.page_size if isinstance( inp.attn_bias, ( PagedBlockDiagonalPaddedKeysMask, PagedBlockDiagonalGappyKeysMask, ), ) else None ), ) ctx: Optional[Context] = None if needs_gradient: ctx = Context( out=out, lse=lse, # cutlass forward is only compatible with cutlass backward if # dropout is used (because of the way RNG states are passed and the # way random numbers are generated during backward) op_bw=BwOp if inp.p != 0 else None, ) if inp.p != 0: ctx.rng_state = torch.tensor( [rng_seed, rng_offset], dtype=torch.int64, device="cpu" ) return out, ctx @classmethod def not_supported_reasons(cls, d: Inputs) -> List[str]: reasons = super(FwOp, cls).not_supported_reasons(d) matmul_alignment_mn = _minimum_gemm_alignment(d) check_lastdim_alignment_stride1(reasons, "query", d.query, matmul_alignment_mn) check_lastdim_alignment_stride1(reasons, "value", d.value, matmul_alignment_mn) _check_bias_alignment(reasons, d.attn_bias) return reasons @register_operator class BwOp(AttentionBwOpBase): __doc__ = FwOp.__doc__ OPERATOR = get_operator("xformers", "efficient_attention_backward_ck") SUPPORTED_DEVICES = FwOp.SUPPORTED_DEVICES SUPPORTED_DTYPES = FwOp.SUPPORTED_DTYPES SUPPORTED_MAX_K = 256 SUPPORTED_ATTN_BIAS_TYPES: Iterable[Any] = ( type(None), torch.Tensor, LowerTriangularMask, LowerTriangularFromBottomRightMask, LowerTriangularFromBottomRightLocalAttentionMask, # TODO: Fix handling of gradient through the fMHA autograd function # LowerTriangularMaskWithTensorBias, BlockDiagonalMask, BlockDiagonalCausalMask, attn_bias.BlockDiagonalCausalFromBottomRightMask, attn_bias.BlockDiagonalCausalLocalAttentionMask, ) SUPPORTS_ATTN_BIAS_GRAD = True SUPPORTS_DROPOUT = FwOp.SUPPORTS_DROPOUT SUPPORTS_CUSTOM_SCALE = FwOp.SUPPORTS_CUSTOM_SCALE SUPPORTS_DIFFERENT_VALUE_EMBED = FwOp.SUPPORTS_DIFFERENT_VALUE_EMBED SUPPORTS_UNPADDED_LSE = True NAME = "ckB" _TEST_K: List[int] = [ 32, # 64x64 kernel 64, 96, 128, # 64x128/128x128 kernel 256, ] @classmethod def not_supported_reasons(cls, d: Inputs) -> List[str]: reasons = super(BwOp, cls).not_supported_reasons(d) matmul_alignment_mn = _minimum_gemm_alignment(d) check_lastdim_alignment_stride1(reasons, "query", d.query, matmul_alignment_mn) check_lastdim_alignment_stride1(reasons, "key", d.key, matmul_alignment_mn) check_lastdim_alignment_stride1(reasons, "value", d.value, matmul_alignment_mn) _check_bias_alignment(reasons, d.attn_bias) attn_bias_tensor = _get_tensor_bias(d.attn_bias) # Backprop of gradient through broadcasted bias is not supported if attn_bias_tensor is not None and attn_bias_tensor.requires_grad: # Don't forget that inputs are either in BMK or BMHK! if d.query.ndim == 3 and attn_bias_tensor.ndim == 3: expected_bias_shape = (*d.query.shape[:2], d.key.shape[1]) else: # bias is B H Mq Mk expected_bias_shape = ( d.query.shape[0], d.query.shape[2] if d.query.ndim == 4 else 1, d.query.shape[1], d.key.shape[1], ) if tuple(attn_bias_tensor.shape) != expected_bias_shape: reasons.append( "Broadcasting the `attn_bias` tensor is not supported " f"(shape: {tuple(attn_bias_tensor.shape)}" f"/ expected: {expected_bias_shape})" ) return reasons @classmethod def apply(cls, ctx: Context, inp: Inputs, grad: torch.Tensor) -> Gradients: if type(inp.attn_bias) not in BwOp.SUPPORTED_ATTN_BIAS_TYPES: raise NotImplementedError("Unsupported attn_bias type") seqstart_k, seqstart_q, max_seqlen_q, max_seqlen_k = _get_seqlen_info(inp) dtype = inp.query.dtype rng_seed = rng_offset = 0 if inp.p != 0.0: if ( ctx.rng_state is None or ctx.rng_state.dtype != torch.int64 or ctx.rng_state.device.type != "cpu" or ctx.rng_state.shape != (2,) ): raise NotImplementedError(f"Invalid rng_state: {ctx.rng_state}") rng_seed, rng_offset = ctx.rng_state.tolist() (grad_q, grad_k, grad_v, grad_bias) = cls.OPERATOR( grad.to(dtype), inp.query, inp.key, inp.value, attn_bias=_get_tensor_bias(inp.attn_bias), seqstart_q=seqstart_q, seqstart_k=seqstart_k, max_seqlen_q=max_seqlen_q, max_seqlen_k=max_seqlen_k, seqlen_k=( inp.attn_bias.k_seqinfo.seqlen if isinstance( inp.attn_bias, ( BlockDiagonalGappyKeysMask, BlockDiagonalPaddedKeysMask, ), ) else None ), logsumexp=ctx.lse, output=ctx.out.to(dtype), dropout_p=inp.p, # if not using dropout, seed and offset are irrelevant but still expected # in function signature so just pass 0 # seed and offset could be None if a different FW op other than cutlass # was used. rng_seed=rng_seed, rng_offset=rng_offset, custom_mask_type=_custom_mask_type(inp.attn_bias), scale=inp.scale, window_size=( inp.attn_bias._window_size if isinstance( inp.attn_bias, ( BlockDiagonalCausalLocalAttentionMask, BlockDiagonalCausalLocalAttentionFromBottomRightMask, LowerTriangularFromBottomRightLocalAttentionMask, ), ) else None ), ) # c++/CUDA implementation returns an uninitialized tensor if bias doesn't # require grad if not ( isinstance(inp.attn_bias, torch.Tensor) and inp.attn_bias.requires_grad ): grad_bias = None return Gradients(dq=grad_q, dk=grad_k, dv=grad_v, db=grad_bias)