File size: 3,988 Bytes
4275cae
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
import torch
import torch.nn as nn
from torch.nn.utils.rnn import pack_padded_sequence

def init_weight(m):
    if isinstance(m, nn.Conv1d) or isinstance(m, nn.Linear) or isinstance(m, nn.ConvTranspose1d):
        nn.init.xavier_normal_(m.weight)
        # m.bias.data.fill_(0.01)
        if m.bias is not None:
            nn.init.constant_(m.bias, 0)

            
class MovementConvEncoder(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(MovementConvEncoder, self).__init__()
        self.main = nn.Sequential(
            nn.Conv1d(input_size, hidden_size, 4, 2, 1),
            nn.Dropout(0.2, inplace=True),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Conv1d(hidden_size, output_size, 4, 2, 1),
            nn.Dropout(0.2, inplace=True),
            nn.LeakyReLU(0.2, inplace=True),
        )
        self.out_net = nn.Linear(output_size, output_size)
        self.main.apply(init_weight)
        self.out_net.apply(init_weight)

    def forward(self, inputs):
        inputs = inputs.permute(0, 2, 1)
        outputs = self.main(inputs).permute(0, 2, 1)
        # print(outputs.shape)
        return self.out_net(outputs)



class TextEncoderBiGRUCo(nn.Module):
    def __init__(self, word_size, pos_size, hidden_size, output_size, device):
        super(TextEncoderBiGRUCo, self).__init__()
        self.device = device

        self.pos_emb = nn.Linear(pos_size, word_size)
        self.input_emb = nn.Linear(word_size, hidden_size)
        self.gru = nn.GRU(hidden_size, hidden_size, batch_first=True, bidirectional=True)
        self.output_net = nn.Sequential(
            nn.Linear(hidden_size * 2, hidden_size),
            nn.LayerNorm(hidden_size),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(hidden_size, output_size)
        )

        self.input_emb.apply(init_weight)
        self.pos_emb.apply(init_weight)
        self.output_net.apply(init_weight)
        self.hidden_size = hidden_size
        self.hidden = nn.Parameter(torch.randn((2, 1, self.hidden_size), requires_grad=True))

    # input(batch_size, seq_len, dim)
    def forward(self, word_embs, pos_onehot, cap_lens):
        num_samples = word_embs.shape[0]

        pos_embs = self.pos_emb(pos_onehot)
        inputs = word_embs + pos_embs
        input_embs = self.input_emb(inputs)
        hidden = self.hidden.repeat(1, num_samples, 1)

        cap_lens = cap_lens.data.tolist()
        emb = pack_padded_sequence(input_embs, cap_lens, batch_first=True)

        gru_seq, gru_last = self.gru(emb, hidden)

        gru_last = torch.cat([gru_last[0], gru_last[1]], dim=-1)

        return self.output_net(gru_last)


class MotionEncoderBiGRUCo(nn.Module):
    def __init__(self, input_size, hidden_size, output_size, device):
        super(MotionEncoderBiGRUCo, self).__init__()
        self.device = device

        self.input_emb = nn.Linear(input_size, hidden_size)
        self.gru = nn.GRU(hidden_size, hidden_size, batch_first=True, bidirectional=True)
        self.output_net = nn.Sequential(
            nn.Linear(hidden_size*2, hidden_size),
            nn.LayerNorm(hidden_size),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(hidden_size, output_size)
        )

        self.input_emb.apply(init_weight)
        self.output_net.apply(init_weight)
        self.hidden_size = hidden_size
        self.hidden = nn.Parameter(torch.randn((2, 1, self.hidden_size), requires_grad=True))

    # input(batch_size, seq_len, dim)
    def forward(self, inputs, m_lens):
        num_samples = inputs.shape[0]

        input_embs = self.input_emb(inputs)
        hidden = self.hidden.repeat(1, num_samples, 1)

        cap_lens = m_lens.data.tolist()
        emb = pack_padded_sequence(input_embs, cap_lens, batch_first=True, enforce_sorted=False)

        gru_seq, gru_last = self.gru(emb, hidden)

        gru_last = torch.cat([gru_last[0], gru_last[1]], dim=-1)

        return self.output_net(gru_last)