from typing import Optional, Tuple, List import math import torch from torch import Tensor from torch.nn import Linear, Module from torch.nn import functional as F from torch.nn.init import constant_, xavier_normal_, xavier_uniform_ from torch.nn.modules.linear import NonDynamicallyQuantizableLinear from torch.nn.parameter import Parameter def _in_projection_packed( q: Tensor, k: Tensor, v: Tensor, w: Tensor, b: Optional[Tensor] = None, ) -> List[Tensor]: r""" Performs the in-projection step of the attention operation, using packed weights. Output is a triple containing projection tensors for query, key and value. Args: q, k, v: query, key and value tensors to be projected. For self-attention, these are typically the same tensor; for encoder-decoder attention, k and v are typically the same tensor. (We take advantage of these identities for performance if they are present.) Regardless, q, k and v must share a common embedding dimension; otherwise their shapes may vary. w: projection weights for q, k and v, packed into a single tensor. Weights are packed along dimension 0, in q, k, v order. b: optional projection biases for q, k and v, packed into a single tensor in q, k, v order. Shape: Inputs: - q: :math:`(..., E)` where E is the embedding dimension - k: :math:`(..., E)` where E is the embedding dimension - v: :math:`(..., E)` where E is the embedding dimension - w: :math:`(E * 3, E)` where E is the embedding dimension - b: :math:`E * 3` where E is the embedding dimension Output: - in output list :math:`[q', k', v']`, each output tensor will have the same shape as the corresponding input tensor. """ E = q.size(-1) if k is v: if q is k: # self-attention return F.linear(q, w, b).chunk(3, dim=-1) else: # encoder-decoder attention w_q, w_kv = w.split([E, E * 2]) if b is None: b_q = b_kv = None else: b_q, b_kv = b.split([E, E * 2]) return (F.linear(q, w_q, b_q),) + F.linear(k, w_kv, b_kv).chunk(2, dim=-1) else: w_q, w_k, w_v = w.chunk(3) if b is None: b_q = b_k = b_v = None else: b_q, b_k, b_v = b.chunk(3) return F.linear(q, w_q, b_q), F.linear(k, w_k, b_k), F.linear(v, w_v, b_v) def _scaled_dot_product_attention( q: Tensor, k: Tensor, v: Tensor, attn_mask: Optional[Tensor] = None, dropout_p: float = 0.0, ) -> Tuple[Tensor, Tensor]: r""" Computes scaled dot product attention on query, key and value tensors, using an optional attention mask if passed, and applying dropout if a probability greater than 0.0 is specified. Returns a tensor pair containing attended values and attention weights. Args: q, k, v: query, key and value tensors. See Shape section for shape details. attn_mask: optional tensor containing mask values to be added to calculated attention. May be 2D or 3D; see Shape section for details. dropout_p: dropout probability. If greater than 0.0, dropout is applied. Shape: - q: :math:`(B, Nt, E)` where B is batch size, Nt is the target sequence length, and E is embedding dimension. - key: :math:`(B, Ns, E)` where B is batch size, Ns is the source sequence length, and E is embedding dimension. - value: :math:`(B, Ns, E)` where B is batch size, Ns is the source sequence length, and E is embedding dimension. - attn_mask: either a 3D tensor of shape :math:`(B, Nt, Ns)` or a 2D tensor of shape :math:`(Nt, Ns)`. - Output: attention values have shape :math:`(B, Nt, E)`; attention weights have shape :math:`(B, Nt, Ns)` """ B, Nt, E = q.shape q = q / math.sqrt(E) # (B, Nt, E) x (B, E, Ns) -> (B, Nt, Ns) if attn_mask is not None: attn = torch.baddbmm(attn_mask, q, k.transpose(-2, -1)) else: attn = torch.bmm(q, k.transpose(-2, -1)) attn = F.softmax(attn, dim=-1) if dropout_p > 0.0: attn = F.dropout(attn, p=dropout_p) # (B, Nt, Ns) x (B, Ns, E) -> (B, Nt, E) output = torch.bmm(attn, v) return output, attn def multi_head_attention_forward( x, ipw, ipb, opw, opb, n_head, attn_mask, past_kv=None, use_cache=False, ): # x = x.transpose(1, 0) # tgt_len, bsz, embed_dim = x.shape # head_dim = embed_dim // n_head # q, k, v = _in_projection_packed(x, x, x, ipw, ipb) # q = q.contiguous().view(tgt_len, bsz * n_head, head_dim).transpose(0, 1) # k = k.contiguous().view(k.shape[0], bsz * n_head, head_dim).transpose(0, 1) # v = v.contiguous().view(v.shape[0], bsz * n_head, head_dim).transpose(0, 1) # new_attn_mask = torch.zeros_like(attn_mask, dtype=q.dtype) # new_attn_mask.masked_fill_(attn_mask, float("-inf")) # attn_mask = new_attn_mask # # attn_output, attn_output_weights = _scaled_dot_product_attention(q, k, v, attn_mask, 0.0) # attn_output = attn_output.transpose(0, 1).contiguous().view(tgt_len * bsz, embed_dim) # attn_output = torch._C._nn.linear(attn_output, opw, opb) # attn_output = attn_output.view(tgt_len, bsz, attn_output.size(1)) B, T, C = x.size() q, k, v = torch._C._nn.linear(x, ipw, ipb).chunk(3, dim=-1) k = k.view(B, T, n_head, C // n_head).transpose(1, 2) # (B, nh, T, hs) q = q.view(B, T, n_head, C // n_head).transpose(1, 2) # (B, nh, T, hs) v = v.view(B, T, n_head, C // n_head).transpose(1, 2) # (B, nh, T, hs) if past_kv is not None: past_key = past_kv[0] past_value = past_kv[1] k = torch.cat((past_key, k), dim=-2) v = torch.cat((past_value, v), dim=-2) FULL_T = k.shape[-2] if use_cache is True: present = (k, v) else: present = None att = (q @ k.transpose(-2, -1)) * (1.0 / math.sqrt(k.size(-1))) att = att.masked_fill(attn_mask[FULL_T - T:FULL_T, :FULL_T], float('-inf')) att = F.softmax(att, dim=-1) y = att @ v # (B, nh, T, T) x (B, nh, T, hs) -> (B, nh, T, hs) y = y.transpose(1, 2).contiguous().view(B, T, C) # re-assemble all head outputs side by side y = torch._C._nn.linear(y, opw, opb) return (y, present) class MultiheadAttention(Module): r"""Allows the model to jointly attend to information from different representation subspaces as described in the paper: `Attention Is All You Need `_. Multi-Head Attention is defined as: .. math:: \text{MultiHead}(Q, K, V) = \text{Concat}(head_1,\dots,head_h)W^O where :math:`head_i = \text{Attention}(QW_i^Q, KW_i^K, VW_i^V)`. ``forward()`` will use a special optimized implementation if all of the following conditions are met: - self attention is being computed (i.e., ``query``, ``key``, and ``value`` are the same tensor. This restriction will be loosened in the future.) - Either autograd is disabled (using ``torch.inference_mode`` or ``torch.no_grad``) or no tensor argument ``requires_grad`` - training is disabled (using ``.eval()``) - dropout is 0 - ``add_bias_kv`` is ``False`` - ``add_zero_attn`` is ``False`` - ``batch_first`` is ``True`` and the input is batched - ``kdim`` and ``vdim`` are equal to ``embed_dim`` - at most one of ``key_padding_mask`` or ``attn_mask`` is passed - if a `NestedTensor `_ is passed, neither ``key_padding_mask`` nor ``attn_mask`` is passed If the optimized implementation is in use, a `NestedTensor `_ can be passed for ``query``/``key``/``value`` to represent padding more efficiently than using a padding mask. In this case, a `NestedTensor `_ will be returned, and an additional speedup proportional to the fraction of the input that is padding can be expected. Args: embed_dim: Total dimension of the model. num_heads: Number of parallel attention heads. Note that ``embed_dim`` will be split across ``num_heads`` (i.e. each head will have dimension ``embed_dim // num_heads``). dropout: Dropout probability on ``attn_output_weights``. Default: ``0.0`` (no dropout). bias: If specified, adds bias to input / output projection layers. Default: ``True``. add_bias_kv: If specified, adds bias to the key and value sequences at dim=0. Default: ``False``. add_zero_attn: If specified, adds a new batch of zeros to the key and value sequences at dim=1. Default: ``False``. kdim: Total number of features for keys. Default: ``None`` (uses ``kdim=embed_dim``). vdim: Total number of features for values. Default: ``None`` (uses ``vdim=embed_dim``). batch_first: If ``True``, then the input and output tensors are provided as (batch, seq, feature). Default: ``False`` (seq, batch, feature). Examples:: >>> # xdoctest: +SKIP >>> multihead_attn = nn.MultiheadAttention(embed_dim, num_heads) >>> attn_output, attn_output_weights = multihead_attn(query, key, value) """ __constants__ = ["batch_first"] bias_k: Optional[torch.Tensor] bias_v: Optional[torch.Tensor] def __init__( self, embed_dim, num_heads, dropout=0.0, bias=True, add_bias_kv=False, add_zero_attn=False, kdim=None, vdim=None, batch_first=False, linear1_cls=Linear, linear2_cls=Linear, device=None, dtype=None, ) -> None: factory_kwargs = {"device": device, "dtype": dtype} super(MultiheadAttention, self).__init__() self.embed_dim = embed_dim self.kdim = kdim if kdim is not None else embed_dim self.vdim = vdim if vdim is not None else embed_dim self._qkv_same_embed_dim = ( self.kdim == embed_dim and self.vdim == embed_dim ) self.num_heads = num_heads self.dropout = dropout self.batch_first = batch_first self.head_dim = embed_dim // num_heads assert ( self.head_dim * num_heads == self.embed_dim ), "embed_dim must be divisible by num_heads" if add_bias_kv: self.bias_k = Parameter( torch.empty((1, 1, embed_dim), **factory_kwargs) ) self.bias_v = Parameter( torch.empty((1, 1, embed_dim), **factory_kwargs) ) else: self.bias_k = self.bias_v = None if linear1_cls == Linear: if not self._qkv_same_embed_dim: self.q_proj_weight = Parameter( torch.empty((embed_dim, embed_dim), **factory_kwargs) ) self.k_proj_weight = Parameter( torch.empty((embed_dim, self.kdim), **factory_kwargs) ) self.v_proj_weight = Parameter( torch.empty((embed_dim, self.vdim), **factory_kwargs) ) self.register_parameter("in_proj_weight", None) else: self.in_proj_weight = Parameter( torch.empty((3 * embed_dim, embed_dim), **factory_kwargs) ) self.register_parameter("q_proj_weight", None) self.register_parameter("k_proj_weight", None) self.register_parameter("v_proj_weight", None) if bias: self.in_proj_bias = Parameter( torch.empty(3 * embed_dim, **factory_kwargs) ) else: self.register_parameter("in_proj_bias", None) self.out_proj = NonDynamicallyQuantizableLinear( embed_dim, embed_dim, bias=bias, **factory_kwargs ) self._reset_parameters() else: if not self._qkv_same_embed_dim: raise NotImplementedError else: self.in_proj_linear = linear1_cls( embed_dim, 3 * embed_dim, bias=bias, **factory_kwargs ) self.in_proj_weight = self.in_proj_linear.weight self.register_parameter("q_proj_weight", None) self.register_parameter("k_proj_weight", None) self.register_parameter("v_proj_weight", None) if bias: self.in_proj_bias = self.in_proj_linear.bias else: self.register_parameter("in_proj_bias", None) self.out_proj = linear2_cls( embed_dim, embed_dim, bias=bias, **factory_kwargs ) if self.bias_k is not None: xavier_normal_(self.bias_k) if self.bias_v is not None: xavier_normal_(self.bias_v) self.add_zero_attn = add_zero_attn def _reset_parameters(self): if self._qkv_same_embed_dim: xavier_uniform_(self.in_proj_weight) else: xavier_uniform_(self.q_proj_weight) xavier_uniform_(self.k_proj_weight) xavier_uniform_(self.v_proj_weight) if self.in_proj_bias is not None: constant_(self.in_proj_bias, 0.0) constant_(self.out_proj.bias, 0.0) if self.bias_k is not None: xavier_normal_(self.bias_k) if self.bias_v is not None: xavier_normal_(self.bias_v) def __setstate__(self, state): # Support loading old MultiheadAttention checkpoints generated by v1.1.0 if "_qkv_same_embed_dim" not in state: state["_qkv_same_embed_dim"] = True super(MultiheadAttention, self).__setstate__(state) def forward( self, query: Tensor, key: Tensor, value: Tensor, key_padding_mask: Optional[Tensor] = None, need_weights: bool = True, attn_mask: Optional[Tensor] = None, average_attn_weights: bool = True, ) -> Tuple[Tensor, Optional[Tensor]]: r""" Args: query: Query embeddings of shape :math:`(L, E_q)` for unbatched input, :math:`(L, N, E_q)` when ``batch_first=False`` or :math:`(N, L, E_q)` when ``batch_first=True``, where :math:`L` is the target sequence length, :math:`N` is the batch size, and :math:`E_q` is the query embedding dimension ``embed_dim``. Queries are compared against key-value pairs to produce the output. See "Attention Is All You Need" for more details. key: Key embeddings of shape :math:`(S, E_k)` for unbatched input, :math:`(S, N, E_k)` when ``batch_first=False`` or :math:`(N, S, E_k)` when ``batch_first=True``, where :math:`S` is the source sequence length, :math:`N` is the batch size, and :math:`E_k` is the key embedding dimension ``kdim``. See "Attention Is All You Need" for more details. value: Value embeddings of shape :math:`(S, E_v)` for unbatched input, :math:`(S, N, E_v)` when ``batch_first=False`` or :math:`(N, S, E_v)` when ``batch_first=True``, where :math:`S` is the source sequence length, :math:`N` is the batch size, and :math:`E_v` is the value embedding dimension ``vdim``. See "Attention Is All You Need" for more details. key_padding_mask: If specified, a mask of shape :math:`(N, S)` indicating which elements within ``key`` to ignore for the purpose of attention (i.e. treat as "padding"). For unbatched `query`, shape should be :math:`(S)`. Binary and byte masks are supported. For a binary mask, a ``True`` value indicates that the corresponding ``key`` value will be ignored for the purpose of attention. For a float mask, it will be directly added to the corresponding ``key`` value. need_weights: If specified, returns ``attn_output_weights`` in addition to ``attn_outputs``. Default: ``True``. attn_mask: If specified, a 2D or 3D mask preventing attention to certain positions. Must be of shape :math:`(L, S)` or :math:`(N\cdot\text{num\_heads}, L, S)`, where :math:`N` is the batch size, :math:`L` is the target sequence length, and :math:`S` is the source sequence length. A 2D mask will be broadcasted across the batch while a 3D mask allows for a different mask for each entry in the batch. Binary, byte, and float masks are supported. For a binary mask, a ``True`` value indicates that the corresponding position is not allowed to attend. For a byte mask, a non-zero value indicates that the corresponding position is not allowed to attend. For a float mask, the mask values will be added to the attention weight. average_attn_weights: If true, indicates that the returned ``attn_weights`` should be averaged across heads. Otherwise, ``attn_weights`` are provided separately per head. Note that this flag only has an effect when ``need_weights=True``. Default: ``True`` (i.e. average weights across heads) Outputs: - **attn_output** - Attention outputs of shape :math:`(L, E)` when input is unbatched, :math:`(L, N, E)` when ``batch_first=False`` or :math:`(N, L, E)` when ``batch_first=True``, where :math:`L` is the target sequence length, :math:`N` is the batch size, and :math:`E` is the embedding dimension ``embed_dim``. - **attn_output_weights** - Only returned when ``need_weights=True``. If ``average_attn_weights=True``, returns attention weights averaged across heads of shape :math:`(L, S)` when input is unbatched or :math:`(N, L, S)`, where :math:`N` is the batch size, :math:`L` is the target sequence length, and :math:`S` is the source sequence length. If ``average_attn_weights=False``, returns attention weights per head of shape :math:`(\text{num\_heads}, L, S)` when input is unbatched or :math:`(N, \text{num\_heads}, L, S)`. .. note:: `batch_first` argument is ignored for unbatched inputs. """ is_batched = query.dim() == 3 if key_padding_mask is not None: _kpm_dtype = key_padding_mask.dtype if _kpm_dtype != torch.bool and not torch.is_floating_point( key_padding_mask ): raise AssertionError( "only bool and floating types of key_padding_mask are supported" ) why_not_fast_path = "" if not is_batched: why_not_fast_path = f"input not batched; expected query.dim() of 3 but got {query.dim()}" elif query is not key or key is not value: # When lifting this restriction, don't forget to either # enforce that the dtypes all match or test cases where # they don't! why_not_fast_path = "non-self attention was used (query, key, and value are not the same Tensor)" elif ( self.in_proj_bias is not None and query.dtype != self.in_proj_bias.dtype ): why_not_fast_path = f"dtypes of query ({query.dtype}) and self.in_proj_bias ({self.in_proj_bias.dtype}) don't match" elif ( self.in_proj_weight is not None and query.dtype != self.in_proj_weight.dtype ): # this case will fail anyway, but at least they'll get a useful error message. why_not_fast_path = f"dtypes of query ({query.dtype}) and self.in_proj_weight ({self.in_proj_weight.dtype}) don't match" elif self.training: why_not_fast_path = "training is enabled" elif not self.batch_first: why_not_fast_path = "batch_first was not True" elif self.bias_k is not None: why_not_fast_path = "self.bias_k was not None" elif self.bias_v is not None: why_not_fast_path = "self.bias_v was not None" elif self.dropout: why_not_fast_path = f"dropout was {self.dropout}, required zero" elif self.add_zero_attn: why_not_fast_path = "add_zero_attn was enabled" elif not self._qkv_same_embed_dim: why_not_fast_path = "_qkv_same_embed_dim was not True" elif attn_mask is not None: why_not_fast_path = "attn_mask was not None" elif query.is_nested and key_padding_mask is not None: why_not_fast_path = ( "key_padding_mask is not supported with NestedTensor input" ) elif self.num_heads % 2 == 1: why_not_fast_path = "num_heads is odd" elif torch.is_autocast_enabled(): why_not_fast_path = "autocast is enabled" if not why_not_fast_path: tensor_args = ( query, key, value, self.in_proj_weight, self.in_proj_bias, self.out_proj.weight, self.out_proj.bias, ) # We have to use list comprehensions below because TorchScript does not support # generator expressions. if torch.overrides.has_torch_function(tensor_args): why_not_fast_path = "some Tensor argument has_torch_function" elif not all( [ (x is None or x.is_cuda or "cpu" in str(x.device)) for x in tensor_args ] ): why_not_fast_path = ( "some Tensor argument is neither CUDA nor CPU" ) elif torch.is_grad_enabled() and any( [x is not None and x.requires_grad for x in tensor_args] ): why_not_fast_path = ( "grad is enabled and at least one of query or the " "input/output projection weights or biases requires_grad" ) if not why_not_fast_path: return torch._native_multi_head_attention( query, key, value, self.embed_dim, self.num_heads, self.in_proj_weight, self.in_proj_bias, self.out_proj.weight, self.out_proj.bias, key_padding_mask if key_padding_mask is not None else attn_mask, need_weights, average_attn_weights, 1 if key_padding_mask is not None else 0 if attn_mask is not None else None, ) any_nested = query.is_nested or key.is_nested or value.is_nested assert not any_nested, ( "MultiheadAttention does not support NestedTensor outside of its fast path. " + f"The fast path was not hit because {why_not_fast_path}" ) if self.batch_first and is_batched: # make sure that the transpose op does not affect the "is" property if key is value: if query is key: query = key = value = query.transpose(1, 0) else: query, key = [x.transpose(1, 0) for x in (query, key)] value = key else: query, key, value = [ x.transpose(1, 0) for x in (query, key, value) ] if not self._qkv_same_embed_dim: attn_output, attn_output_weights = F.multi_head_attention_forward( query, key, value, self.embed_dim, self.num_heads, self.in_proj_weight, self.in_proj_bias, self.bias_k, self.bias_v, self.add_zero_attn, self.dropout, self.out_proj.weight, self.out_proj.bias, training=self.training, key_padding_mask=key_padding_mask, need_weights=need_weights, attn_mask=attn_mask, use_separate_proj_weight=True, q_proj_weight=self.q_proj_weight, k_proj_weight=self.k_proj_weight, v_proj_weight=self.v_proj_weight, average_attn_weights=average_attn_weights, ) else: attn_output, attn_output_weights = F.multi_head_attention_forward( query, key, value, self.embed_dim, self.num_heads, self.in_proj_weight, self.in_proj_bias, self.bias_k, self.bias_v, self.add_zero_attn, self.dropout, self.out_proj.weight, self.out_proj.bias, training=self.training, key_padding_mask=key_padding_mask, need_weights=need_weights, attn_mask=attn_mask, average_attn_weights=average_attn_weights, ) if self.batch_first and is_batched: return attn_output.transpose(1, 0), attn_output_weights else: return attn_output, attn_output_weights def infer(self, x: Tensor, key_padding_mask: Optional[Tensor] = None, need_weights: bool = True, attn_mask: Optional[Tensor] = None, average_attn_weights: bool = True, past_kv = None, use_cache = False ): # x = x.transpose(1, 0) y, kv = multi_head_attention_forward( x=x, ipw=self.in_proj_weight, ipb=self.in_proj_bias, opw=self.out_proj.weight, opb=self.out_proj.bias, n_head=self.num_heads, attn_mask=attn_mask, past_kv=past_kv, use_cache=use_cache, ) return (y, kv)