hier_lstm/seq_pred.py at master · lightingghost/hier_lstm · GitHub

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
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
import mxnet as mx
import os
import sys
import numpy as np
from normal_lstm import HyperPara, lstm_model, get_input_shapes
from copy import copy
import json


file = open('data/idx2word', 'r')
idx2word = json.load(file)
file.close()


class Model:
    def __init__(self, sym_gen, init_paras, pretrained, idx2word, ctx=mx.gpu()):
        self.idx2word = idx2word
        self.state_dict = copy(init_paras)
        self.sym_gen = sym_gen
        self.pretrained = pretrained


    def predict(self, data, label, data_len=300, label_len=30):
        self.state_dict['data'] = (1, data_len)
        self.state_dict['label'] = (1, 30)

        self.model_exec = self.sym_gen(data_len).simple_bind(ctx=mx.cpu(), **self.state_dict)

        for key in self.model_exec.arg_dict.keys():
            if key in self.pretrained.arg_params:
                self.pretrained.arg_params[key].copyto(self.model_exec.arg_dict[key])
        for name, shape in self.state_dict.items():
            mx.nd.zeros(shape).copyto(self.model_exec.arg_dict[name])

        mx.nd.array(data).copyto(self.model_exec.arg_dict['data'])
        mx.nd.array(label).copyto(self.model_exec.arg_dict['label'])
        self.model_exec.forward()

        prob = self.model_exec.outputs[0].asnumpy()
        idxs = np.argmax(prob, axis=1)

        doc_vec = np.hstack([self.model_exec.outputs[i+1].asnumpy() for i in range(3)])

        pred = [self.idx2word[str(i)] for i in idxs if str(i) in self.idx2word]

        return idxs, doc_vec

def translate(seq, dict=idx2word):
    result = []
    for idx in seq:
        if idx == 1:
            continue
        elif idx == 2:
            result.append('<unk>')
        elif idx == 0:
            break
        else:
            result.append(dict[str(idx)])
    return ' '.join(result)

def predict(epoch):
    #model para
    _dict_len       = 55496
    _test           = True
    _num_lstm_layer = 3
    _input_size     = _dict_len + 3
    _num_hidden     = 512
    _num_embed      = 300
    _num_label      = _dict_len + 3
    _dropout        = 0.5
    #opt para
    _learning_rate  = 0.001
    #training para
    _devs           = [mx.cpu()]
    _batch_size     = 1
    _num_epoch      = 4


    #data
    data_name = 'data'
    label_name = 'label'
    if _test:
        data_path = os.path.join('data', 'ndata1000.npy')
        label_path = os.path.join('data', 'label1000.npy')
    else:
        name = 'val'
        data_path = os.path.join('data', 'normal_lstm', name + '_data.npy')
        label_path = os.path.join('data', 'normal_lstm', name + '_label.npy')
    data = np.load(data_path)
    label = np.load(label_path)
    _nsamples = label.shape[0]


    # #model

    def sym_gen(seq_len):
        enc_para = HyperPara(num_lstm_layer = _num_lstm_layer,
                            seq_len        = seq_len,
                            input_size     = _input_size,
                            num_hidden     = _num_hidden,
                            num_embed      = _num_embed,
                            num_label      = None,
                            dropout        = _dropout)
        dec_para = HyperPara(num_lstm_layer = _num_lstm_layer,
                            seq_len        = 30,
                            input_size     = None,
                            num_hidden     = _num_hidden,
                            num_embed      = None,
                            num_label      = _num_label,
                            dropout        = _dropout)

        data_name = 'data'
        label_name = 'label'
        sym = lstm_model(data_name, label_name, enc_para, dec_para)
        return sym

    enc_para = HyperPara(num_lstm_layer = _num_lstm_layer,
                         seq_len        = 300,
                         input_size     = _input_size,
                         num_hidden     = _num_hidden,
                         num_embed      = _num_embed,
                         num_label      = None,
                         dropout        = _dropout)
    dec_para = HyperPara(num_lstm_layer = _num_lstm_layer,
                         seq_len        = 30,
                         input_size     = None,
                         num_hidden     = _num_hidden,
                         num_embed      = None,
                         num_label      = _num_label,
                         dropout        = _dropout)

    init_dict = get_input_shapes(enc_para, dec_para, _batch_size)

    checkpoint_path = os.path.join('checkpoint0', 'auto_sum')
    pretrained_model = mx.model.FeedForward.load(checkpoint_path, epoch)

    file = open('data/idx2word', 'r')
    idx2word = json.load(file)
    file.close()

    pre_model = Model(sym_gen, init_dict, pretrained_model, idx2word)

    print('Previous model load complete.')

    data_lens = np.argwhere(data == -1)[:, 1]
    # for i in range(_nsamples):
    #     data[i, data_lens[i]] = 0
    doc_vecs = []

    idxs = np.random.permutation(_nsamples)

    for i in range(_nsamples):
        if data_lens[i] <= 3:
            continue
        t_data = data[i, :data_lens[i]].reshape((1, data_lens[i]))
        t_label = label[i, :].reshape((1, 30))
        print(i)
        pred, doc_vec = pre_model.predict(t_data, t_label, data_lens[i])
        print(translate(t_data[0]))
        print('----')
        print(translate(pred))
        print('----')
        print(translate(t_label[0]))
        doc_vecs.append(doc_vec)

    return np.vstack(doc_vecs)

if __name__ == '__main__':
    doc_vecs = predict(30)
    np.save('doc_vecs_ils.npy', doc_vecs)