"""
Implementation of "Attention is All You Need" and of
subsequent transformer based architectures
"""
import torch
import torch.nn as nn
from onmt.decoders.decoder import DecoderBase
from onmt.modules import MultiHeadedAttention, AverageAttention
from onmt.modules.position_ffn import PositionwiseFeedForward
from onmt.modules.position_ffn import ActivationFunction
from onmt.utils.misc import sequence_mask
class TransformerDecoderLayerBase(nn.Module):
def __init__(
self,
d_model,
heads,
d_ff,
dropout,
attention_dropout,
self_attn_type="scaled-dot",
max_relative_positions=0,
aan_useffn=False,
full_context_alignment=False,
alignment_heads=0,
pos_ffn_activation_fn=ActivationFunction.relu,
):
"""
Args:
d_model (int): the dimension of keys/values/queries in
:class:`MultiHeadedAttention`, also the input size of
the first-layer of the :class:`PositionwiseFeedForward`.
heads (int): the number of heads for MultiHeadedAttention.
d_ff (int): the second-layer of the
:class:`PositionwiseFeedForward`.
dropout (float): dropout in residual, self-attn(dot) and
feed-forward
attention_dropout (float): dropout in context_attn (and
self-attn(avg))
self_attn_type (string): type of self-attention scaled-dot,
average
max_relative_positions (int):
Max distance between inputs in relative positions
representations
aan_useffn (bool): Turn on the FFN layer in the AAN decoder
full_context_alignment (bool):
whether enable an extra full context decoder forward for
alignment
alignment_heads (int):
N. of cross attention heads to use for alignment guiding
pos_ffn_activation_fn (ActivationFunction):
activation function choice for PositionwiseFeedForward layer
"""
super(TransformerDecoderLayerBase, self).__init__()
if self_attn_type == "scaled-dot":
self.self_attn = MultiHeadedAttention(
heads,
d_model,
dropout=attention_dropout,
max_relative_positions=max_relative_positions,
)
elif self_attn_type == "average":
self.self_attn = AverageAttention(
d_model, dropout=attention_dropout, aan_useffn=aan_useffn
)
self.feed_forward = PositionwiseFeedForward(d_model, d_ff, dropout,
pos_ffn_activation_fn
)
self.layer_norm_1 = nn.LayerNorm(d_model, eps=1e-6)
self.drop = nn.Dropout(dropout)
self.full_context_alignment = full_context_alignment
self.alignment_heads = alignment_heads
def forward(self, *args, **kwargs):
"""Extend `_forward` for (possibly) multiple decoder pass:
Always a default (future masked) decoder forward pass,
Possibly a second future aware decoder pass for joint learn
full context alignement, :cite:`garg2019jointly`.
Args:
* All arguments of _forward.
with_align (bool): whether return alignment attention.
Returns:
(FloatTensor, FloatTensor, FloatTensor or None):
* output ``(batch_size, T, model_dim)``
* top_attn ``(batch_size, T, src_len)``
* attn_align ``(batch_size, T, src_len)`` or None
"""
with_align = kwargs.pop("with_align", False)
output, attns = self._forward(*args, **kwargs)
top_attn = attns[:, 0, :, :].contiguous()
attn_align = None
if with_align:
if self.full_context_alignment:
# return _, (B, Q_len, K_len)
_, attns = self._forward(*args, **kwargs, future=True)
if self.alignment_heads > 0:
attns = attns[:, : self.alignment_heads, :, :].contiguous()
# layer average attention across heads, get ``(B, Q, K)``
# Case 1: no full_context, no align heads -> layer avg baseline
# Case 2: no full_context, 1 align heads -> guided align
# Case 3: full_context, 1 align heads -> full cte guided align
attn_align = attns.mean(dim=1)
return output, top_attn, attn_align
def update_dropout(self, dropout, attention_dropout):
self.self_attn.update_dropout(attention_dropout)
self.feed_forward.update_dropout(dropout)
self.drop.p = dropout
def _forward(self, *args, **kwargs):
raise NotImplementedError
def _compute_dec_mask(self, tgt_pad_mask, future):
tgt_len = tgt_pad_mask.size(-1)
if not future: # apply future_mask, result mask in (B, T, T)
future_mask = torch.ones(
[tgt_len, tgt_len],
device=tgt_pad_mask.device,
dtype=torch.uint8,
)
future_mask = future_mask.triu_(1).view(1, tgt_len, tgt_len)
# BoolTensor was introduced in pytorch 1.2
try:
future_mask = future_mask.bool()
except AttributeError:
pass
dec_mask = torch.gt(tgt_pad_mask + future_mask, 0)
else: # only mask padding, result mask in (B, 1, T)
dec_mask = tgt_pad_mask
return dec_mask
def _forward_self_attn(self, inputs_norm, dec_mask, layer_cache, step):
if isinstance(self.self_attn, MultiHeadedAttention):
return self.self_attn(
inputs_norm,
inputs_norm,
inputs_norm,
mask=dec_mask,
layer_cache=layer_cache,
attn_type="self",
)
elif isinstance(self.self_attn, AverageAttention):
return self.self_attn(
inputs_norm, mask=dec_mask, layer_cache=layer_cache, step=step
)
else:
raise ValueError(
f"self attention {type(self.self_attn)} not supported"
)
class TransformerDecoderLayer(TransformerDecoderLayerBase):
"""Transformer Decoder layer block in Pre-Norm style.
Pre-Norm style is an improvement w.r.t. Original paper's Post-Norm style,
providing better converge speed and performance. This is also the actual
implementation in tensor2tensor and also avalable in fairseq.
See https://tunz.kr/post/4 and :cite:`DeeperTransformer`.
.. mermaid::
graph LR
%% "*SubLayer" can be self-attn, src-attn or feed forward block
A(input) --> B[Norm]
B --> C["*SubLayer"]
C --> D[Drop]
D --> E((+))
A --> E
E --> F(out)
"""
def __init__(
self,
d_model,
heads,
d_ff,
dropout,
attention_dropout,
self_attn_type="scaled-dot",
max_relative_positions=0,
aan_useffn=False,
full_context_alignment=False,
alignment_heads=0,
pos_ffn_activation_fn=ActivationFunction.relu,
):
"""
Args:
See TransformerDecoderLayerBase
"""
super(TransformerDecoderLayer, self).__init__(
d_model,
heads,
d_ff,
dropout,
attention_dropout,
self_attn_type,
max_relative_positions,
aan_useffn,
full_context_alignment,
alignment_heads,
pos_ffn_activation_fn=pos_ffn_activation_fn,
)
self.context_attn = MultiHeadedAttention(
heads, d_model, dropout=attention_dropout
)
self.layer_norm_2 = nn.LayerNorm(d_model, eps=1e-6)
def update_dropout(self, dropout, attention_dropout):
super(TransformerDecoderLayer, self).update_dropout(
dropout, attention_dropout
)
self.context_attn.update_dropout(attention_dropout)
def _forward(
self,
inputs,
memory_bank,
src_pad_mask,
tgt_pad_mask,
layer_cache=None,
step=None,
future=False,
):
"""A naive forward pass for transformer decoder.
# T: could be 1 in the case of stepwise decoding or tgt_len
Args:
inputs (FloatTensor): ``(batch_size, T, model_dim)``
memory_bank (FloatTensor): ``(batch_size, src_len, model_dim)``
src_pad_mask (bool): ``(batch_size, 1, src_len)``
tgt_pad_mask (bool): ``(batch_size, 1, T)``
layer_cache (dict or None): cached layer info when stepwise decode
step (int or None): stepwise decoding counter
future (bool): If set True, do not apply future_mask.
Returns:
(FloatTensor, FloatTensor):
* output ``(batch_size, T, model_dim)``
* attns ``(batch_size, head, T, src_len)``
"""
dec_mask = None
if inputs.size(1) > 1:
# masking is necessary when sequence length is greater than one
dec_mask = self._compute_dec_mask(tgt_pad_mask, future)
inputs_norm = self.layer_norm_1(inputs)
query, _ = self._forward_self_attn(
inputs_norm, dec_mask, layer_cache, step
)
query = self.drop(query) + inputs
query_norm = self.layer_norm_2(query)
mid, attns = self.context_attn(
memory_bank,
memory_bank,
query_norm,
mask=src_pad_mask,
layer_cache=layer_cache,
attn_type="context",
)
output = self.feed_forward(self.drop(mid) + query)
return output, attns
class TransformerDecoderBase(DecoderBase):
def __init__(self, d_model, copy_attn, alignment_layer):
super(TransformerDecoderBase, self).__init__()
# Decoder State
self.state = {}
# previously, there was a GlobalAttention module here for copy
# attention. But it was never actually used -- the "copy" attention
# just reuses the context attention.
self._copy = copy_attn
self.layer_norm = nn.LayerNorm(d_model, eps=1e-6)
self.alignment_layer = alignment_layer
@classmethod
def from_opt(cls, opt, embeddings):
"""Alternate constructor."""
return cls(
opt.dec_layers,
opt.dec_rnn_size,
opt.heads,
opt.transformer_ff,
opt.copy_attn,
opt.self_attn_type,
opt.dropout[0] if type(opt.dropout) is list else opt.dropout,
opt.attention_dropout[0] if type(opt.attention_dropout) is list else opt.attention_dropout,
embeddings,
opt.max_relative_positions,
opt.aan_useffn,
opt.full_context_alignment,
opt.alignment_layer,
alignment_heads=opt.alignment_heads,
pos_ffn_activation_fn=opt.pos_ffn_activation_fn,
)
def init_state(self, src, memory_bank, enc_hidden):
"""Initialize decoder state."""
self.state["src"] = src
self.state["cache"] = None
def map_state(self, fn):
def _recursive_map(struct, batch_dim=0):
for k, v in struct.items():
if v is not None:
if isinstance(v, dict):
_recursive_map(v)
else:
struct[k] = fn(v, batch_dim)
if self.state["src"] is not None:
self.state["src"] = fn(self.state["src"], 1)
if self.state["cache"] is not None:
_recursive_map(self.state["cache"])
def detach_state(self):
raise NotImplementedError
def forward(self, *args, **kwargs):
raise NotImplementedError
def update_dropout(self, dropout, attention_dropout):
self.embeddings.update_dropout(dropout)
for layer in self.transformer_layers:
layer.update_dropout(dropout, attention_dropout)
class TransformerDecoder(TransformerDecoderBase):
"""The Transformer decoder from "Attention is All You Need".
:cite:`DBLP:journals/corr/VaswaniSPUJGKP17`
.. mermaid::
graph BT
A[input]
B[multi-head self-attn]
BB[multi-head src-attn]
C[feed forward]
O[output]
A --> B
B --> BB
BB --> C
C --> O
Args:
num_layers (int): number of decoder layers.
d_model (int): size of the model
heads (int): number of heads
d_ff (int): size of the inner FF layer
copy_attn (bool): if using a separate copy attention
self_attn_type (str): type of self-attention scaled-dot, average
dropout (float): dropout in residual, self-attn(dot) and feed-forward
attention_dropout (float): dropout in context_attn (and self-attn(avg))
embeddings (onmt.modules.Embeddings):
embeddings to use, should have positional encodings
max_relative_positions (int):
Max distance between inputs in relative positions representations
aan_useffn (bool): Turn on the FFN layer in the AAN decoder
full_context_alignment (bool):
whether enable an extra full context decoder forward for alignment
alignment_layer (int): N° Layer to supervise with for alignment guiding
alignment_heads (int):
N. of cross attention heads to use for alignment guiding
"""
def __init__(
self,
num_layers,
d_model,
heads,
d_ff,
copy_attn,
self_attn_type,
dropout,
attention_dropout,
max_relative_positions,
aan_useffn,
full_context_alignment,
alignment_layer,
alignment_heads,
pos_ffn_activation_fn=ActivationFunction.relu,
):
super(TransformerDecoder, self).__init__(
d_model, copy_attn, alignment_layer
)
self.transformer_layers = nn.ModuleList(
[
TransformerDecoderLayer(
d_model,
heads,
d_ff,
dropout,
attention_dropout,
self_attn_type=self_attn_type,
max_relative_positions=max_relative_positions,
aan_useffn=aan_useffn,
full_context_alignment=full_context_alignment,
alignment_heads=alignment_heads,
pos_ffn_activation_fn=pos_ffn_activation_fn,
)
for i in range(num_layers)
]
)
def detach_state(self):
self.state["src"] = self.state["src"].detach()
def forward(self, tgt_emb, memory_bank, src_pad_mask=None, tgt_pad_mask=None, step=None, **kwargs):
"""Decode, possibly stepwise."""
if step == 0:
self._init_cache(memory_bank)
batch_size, src_len, src_dim = memory_bank.size()
device = memory_bank.device
if src_pad_mask is None:
src_pad_mask = torch.zeros((batch_size, 1, src_len), dtype=torch.bool, device=device)
output = tgt_emb
batch_size, tgt_len, tgt_dim = tgt_emb.size()
if tgt_pad_mask is None:
tgt_pad_mask = torch.zeros((batch_size, 1, tgt_len), dtype=torch.bool, device=device)
future = kwargs.pop("future", False)
with_align = kwargs.pop("with_align", False)
attn_aligns = []
hiddens = []
for i, layer in enumerate(self.transformer_layers):
layer_cache = (
self.state["cache"]["layer_{}".format(i)]
if step is not None
else None
)
output, attn, attn_align = layer(
output,
memory_bank,
src_pad_mask,
tgt_pad_mask,
layer_cache=layer_cache,
step=step,
with_align=with_align,
future=future
)
hiddens.append(output)
if attn_align is not None:
attn_aligns.append(attn_align)
output = self.layer_norm(output) # (B, L, D)
attns = {"std": attn}
if self._copy:
attns["copy"] = attn
if with_align:
attns["align"] = attn_aligns[self.alignment_layer] # `(B, Q, K)`
# attns["align"] = torch.stack(attn_aligns, 0).mean(0) # All avg
# TODO change the way attns is returned dict => list or tuple (onnx)
return output, attns, hiddens
def _init_cache(self, memory_bank):
self.state["cache"] = {}
for i, layer in enumerate(self.transformer_layers):
layer_cache = {"memory_keys": None, "memory_values": None, "self_keys": None, "self_values": None}
self.state["cache"]["layer_{}".format(i)] = layer_cache