import torch import torch.nn.functional as F from torch import nn import copy, math from models.position_encoding import build_position_encoding class TransformerEncoder(nn.Module): def __init__(self, enc_layer, num_layers, use_dense_pos=False): super().__init__() self.layers = nn.ModuleList([copy.deepcopy(enc_layer) for i in range(num_layers)]) self.num_layers = num_layers self.use_dense_pos = use_dense_pos def forward(self, src, pos, padding_mask=None): if self.use_dense_pos: ## pos encoding at each MH-Attention block (q,k) output, pos_enc = src, pos for layer in self.layers: output, att_map = layer(output, pos_enc, padding_mask) else: ## pos encoding at input only (q,k,v) output, pos_enc = src + pos, None for layer in self.layers: output, att_map = layer(output, pos_enc, padding_mask) return output, att_map class EncoderLayer(nn.Module): def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1, activation="relu", use_dense_pos=False): super().__init__() self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout) # Implementation of Feedforward model self.linear1 = nn.Linear(d_model, dim_feedforward) self.dropout = nn.Dropout(dropout) self.linear2 = nn.Linear(dim_feedforward, d_model) self.norm1 = nn.LayerNorm(d_model) self.norm2 = nn.LayerNorm(d_model) self.dropout1 = nn.Dropout(dropout) self.dropout2 = nn.Dropout(dropout) self.activation = _get_activation_fn(activation) def with_pos_embed(self, tensor, pos): return tensor if pos is None else tensor + pos def forward(self, src, pos, padding_mask): q = k = self.with_pos_embed(src, pos) src2, attn = self.self_attn(q, k, value=src, key_padding_mask=padding_mask) src = src + self.dropout1(src2) src = self.norm1(src) src2 = self.linear2(self.dropout(self.activation(self.linear1(src)))) src = src + self.dropout2(src2) src = self.norm2(src) return src, attn class TransformerDecoder(nn.Module): def __init__(self, dec_layer, num_layers, use_dense_pos=False, return_intermediate=False): super().__init__() self.layers = nn.ModuleList([copy.deepcopy(dec_layer) for i in range(num_layers)]) self.num_layers = num_layers self.use_dense_pos = use_dense_pos self.return_intermediate = return_intermediate def forward(self, tgt, tgt_pos, memory, memory_pos, tgt_padding_mask, src_padding_mask, tgt_attn_mask=None): intermediate = [] if self.use_dense_pos: ## pos encoding at each MH-Attention block (q,k) output = tgt tgt_pos_enc, memory_pos_enc = tgt_pos, memory_pos for layer in self.layers: output, att_map = layer(output, tgt_pos_enc, memory, memory_pos_enc, tgt_padding_mask, src_padding_mask, tgt_attn_mask) if self.return_intermediate: intermediate.append(output) else: ## pos encoding at input only (q,k,v) output = tgt + tgt_pos tgt_pos_enc, memory_pos_enc = None, None for layer in self.layers: output, att_map = layer(output, tgt_pos_enc, memory, memory_pos_enc, tgt_padding_mask, src_padding_mask, tgt_attn_mask) if self.return_intermediate: intermediate.append(output) if self.return_intermediate: return torch.stack(intermediate) return output, att_map class DecoderLayer(nn.Module): def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1, activation="relu", use_dense_pos=False): super().__init__() self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout) self.corr_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout) # Implementation of Feedforward model self.linear1 = nn.Linear(d_model, dim_feedforward) self.dropout = nn.Dropout(dropout) self.linear2 = nn.Linear(dim_feedforward, d_model) self.norm1 = nn.LayerNorm(d_model) self.norm2 = nn.LayerNorm(d_model) self.norm3 = nn.LayerNorm(d_model) self.dropout1 = nn.Dropout(dropout) self.dropout2 = nn.Dropout(dropout) self.dropout3 = nn.Dropout(dropout) self.activation = _get_activation_fn(activation) def with_pos_embed(self, tensor, pos): return tensor if pos is None else tensor + pos def forward(self, tgt, tgt_pos, memory, memory_pos, tgt_padding_mask, memory_padding_mask, tgt_attn_mask): q = k = self.with_pos_embed(tgt, tgt_pos) tgt2, attn = self.self_attn(q, k, value=tgt, key_padding_mask=tgt_padding_mask, attn_mask=tgt_attn_mask) tgt = tgt + self.dropout1(tgt2) tgt = self.norm1(tgt) tgt2, attn = self.corr_attn(query=self.with_pos_embed(tgt, tgt_pos), key=self.with_pos_embed(memory, memory_pos), value=memory, key_padding_mask=memory_padding_mask) tgt = tgt + self.dropout2(tgt2) tgt = self.norm2(tgt) tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt)))) tgt = tgt + self.dropout3(tgt2) tgt = self.norm3(tgt) return tgt, attn def _get_activation_fn(activation): """Return an activation function given a string""" if activation == "relu": return F.relu if activation == "gelu": return F.gelu if activation == "glu": return F.glu raise RuntimeError(F"activation should be relu/gelu, not {activation}.") #----------------------------------------------------------------------------------- ''' copy from the implementatoin of "attention-is-all-you-need-pytorch-master" by Yu-Hsiang Huang ''' class MultiHeadAttention(nn.Module): ''' Multi-Head Attention module ''' def __init__(self, n_head, d_model, d_k, d_v, dropout=0.1): super().__init__() self.n_head = n_head self.d_k = d_k self.d_v = d_v self.w_qs = nn.Linear(d_model, n_head * d_k, bias=False) self.w_ks = nn.Linear(d_model, n_head * d_k, bias=False) self.w_vs = nn.Linear(d_model, n_head * d_v, bias=False) self.fc = nn.Linear(n_head * d_v, d_model, bias=False) self.attention = ScaledDotProductAttention(temperature=d_k ** 0.5) self.dropout = nn.Dropout(dropout) self.layer_norm = nn.LayerNorm(d_model, eps=1e-6) def forward(self, q, k, v, mask=None): d_k, d_v, n_head = self.d_k, self.d_v, self.n_head sz_b, len_q, len_k, len_v = q.size(0), q.size(1), k.size(1), v.size(1) residual = q # Pass through the pre-attention projection: b x lq x (n*dv) # Separate different heads: b x lq x n x dv q = self.w_qs(q).view(sz_b, len_q, n_head, d_k) k = self.w_ks(k).view(sz_b, len_k, n_head, d_k) v = self.w_vs(v).view(sz_b, len_v, n_head, d_v) # Transpose for attention dot product: b x n x lq x dv q, k, v = q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2) if mask is not None: mask = mask.unsqueeze(1) # For head axis broadcasting. q, attn = self.attention(q, k, v, mask=mask) # Transpose to move the head dimension back: b x lq x n x dv # Combine the last two dimensions to concatenate all the heads together: b x lq x (n*dv) q = q.transpose(1, 2).contiguous().view(sz_b, len_q, -1) q = self.dropout(self.fc(q)) q += residual q = self.layer_norm(q) return q, attn class ScaledDotProductAttention(nn.Module): ''' Scaled Dot-Product Attention ''' def __init__(self, temperature, attn_dropout=0.1): super().__init__() self.temperature = temperature self.dropout = nn.Dropout(attn_dropout) def forward(self, q, k, v, mask=None): attn = torch.matmul(q / self.temperature, k.transpose(2, 3)) if mask is not None: attn = attn.masked_fill(mask == 0, -1e9) attn = self.dropout(F.softmax(attn, dim=-1)) output = torch.matmul(attn, v) return output, attn