File size: 7,491 Bytes
158b61b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
import copy
import unittest

import torch

import onmt
import onmt.inputters
import onmt.opts
from onmt.model_builder import build_embeddings, \
    build_encoder, build_decoder
from onmt.utils.parse import ArgumentParser

parser = ArgumentParser(description='train.py')
onmt.opts.model_opts(parser)
onmt.opts._add_train_general_opts(parser)

# -data option is required, but not used in this test, so dummy.
opt = parser.parse_known_args(['-data', 'dummy'])[0]


class TestModel(unittest.TestCase):

    def __init__(self, *args, **kwargs):
        super(TestModel, self).__init__(*args, **kwargs)
        self.opt = opt

    def get_field(self):
        src = onmt.inputters.get_fields("text", 0, 0)["src"]
        src.base_field.build_vocab([])
        return src

    def get_batch(self, source_l=3, bsize=1):
        # len x batch x nfeat
        test_src = torch.ones(source_l, bsize, 1).long()
        test_tgt = torch.ones(source_l, bsize, 1).long()
        test_length = torch.ones(bsize).fill_(source_l).long()
        return test_src, test_tgt, test_length

    def embeddings_forward(self, opt, source_l=3, bsize=1):
        '''
        Tests if the embeddings works as expected

        args:
            opt: set of options
            source_l: Length of generated input sentence
            bsize: Batchsize of generated input
        '''
        word_field = self.get_field()
        emb = build_embeddings(opt, word_field)
        test_src, _, __ = self.get_batch(source_l=source_l, bsize=bsize)
        if opt.decoder_type == 'transformer':
            input = torch.cat([test_src, test_src], 0)
            res = emb(input)
            compare_to = torch.zeros(source_l * 2, bsize,
                                     opt.src_word_vec_size)
        else:
            res = emb(test_src)
            compare_to = torch.zeros(source_l, bsize, opt.src_word_vec_size)

        self.assertEqual(res.size(), compare_to.size())

    def encoder_forward(self, opt, source_l=3, bsize=1):
        '''
        Tests if the encoder works as expected

        args:
            opt: set of options
            source_l: Length of generated input sentence
            bsize: Batchsize of generated input
        '''
        if opt.rnn_size > 0:
            opt.enc_rnn_size = opt.rnn_size
        word_field = self.get_field()
        embeddings = build_embeddings(opt, word_field)
        enc = build_encoder(opt, embeddings)

        test_src, test_tgt, test_length = self.get_batch(source_l=source_l,
                                                         bsize=bsize)

        hidden_t, outputs, test_length = enc(test_src, test_length)

        # Initialize vectors to compare size with
        test_hid = torch.zeros(self.opt.enc_layers, bsize, opt.enc_rnn_size)
        test_out = torch.zeros(source_l, bsize, opt.dec_rnn_size)

        # Ensure correct sizes and types
        self.assertEqual(test_hid.size(),
                         hidden_t[0].size(),
                         hidden_t[1].size())
        self.assertEqual(test_out.size(), outputs.size())
        self.assertEqual(type(outputs), torch.Tensor)

    def nmtmodel_forward(self, opt, source_l=3, bsize=1):
        """
        Creates a nmtmodel with a custom opt function.
        Forwards a testbatch and checks output size.

        Args:
            opt: Namespace with options
            source_l: length of input sequence
            bsize: batchsize
        """
        if opt.rnn_size > 0:
            opt.enc_rnn_size = opt.rnn_size
            opt.dec_rnn_size = opt.rnn_size
        word_field = self.get_field()

        embeddings = build_embeddings(opt, word_field)
        enc = build_encoder(opt, embeddings)

        embeddings = build_embeddings(opt, word_field, for_encoder=False)
        dec = build_decoder(opt, embeddings)

        model = onmt.models.model.NMTModel(enc, dec)

        test_src, test_tgt, test_length = self.get_batch(source_l=source_l,
                                                         bsize=bsize)
        outputs, attn = model(test_src, test_tgt, test_length)
        outputsize = torch.zeros(source_l - 1, bsize, opt.dec_rnn_size)
        # Make sure that output has the correct size and type
        self.assertEqual(outputs.size(), outputsize.size())
        self.assertEqual(type(outputs), torch.Tensor)


def _add_test(param_setting, methodname):
    """
    Adds a Test to TestModel according to settings

    Args:
        param_setting: list of tuples of (param, setting)
        methodname: name of the method that gets called
    """

    def test_method(self):
        opt = copy.deepcopy(self.opt)
        if param_setting:
            for param, setting in param_setting:
                setattr(opt, param, setting)
        ArgumentParser.update_model_opts(opt)
        getattr(self, methodname)(opt)
    if param_setting:
        name = 'test_' + methodname + "_" + "_".join(
            str(param_setting).split())
    else:
        name = 'test_' + methodname + '_standard'
    setattr(TestModel, name, test_method)
    test_method.__name__ = name


'''
TEST PARAMETERS
'''
opt.brnn = False

test_embeddings = [[],
                   [('decoder_type', 'transformer')]
                   ]

for p in test_embeddings:
    _add_test(p, 'embeddings_forward')

tests_encoder = [[],
                 [('encoder_type', 'mean')],
                 # [('encoder_type', 'transformer'),
                 # ('word_vec_size', 16), ('rnn_size', 16)],
                 []
                 ]

for p in tests_encoder:
    _add_test(p, 'encoder_forward')

tests_nmtmodel = [[('rnn_type', 'GRU')],
                  [('layers', 10)],
                  [('input_feed', 0)],
                  [('decoder_type', 'transformer'),
                   ('encoder_type', 'transformer'),
                   ('src_word_vec_size', 16),
                   ('tgt_word_vec_size', 16),
                   ('rnn_size', 16)],
                  [('decoder_type', 'transformer'),
                   ('encoder_type', 'transformer'),
                   ('src_word_vec_size', 16),
                   ('tgt_word_vec_size', 16),
                   ('rnn_size', 16),
                   ('position_encoding', True)],
                  [('coverage_attn', True)],
                  [('copy_attn', True)],
                  [('global_attention', 'mlp')],
                  [('context_gate', 'both')],
                  [('context_gate', 'target')],
                  [('context_gate', 'source')],
                  [('encoder_type', "brnn"),
                   ('brnn_merge', 'sum')],
                  [('encoder_type', "brnn")],
                  [('decoder_type', 'cnn'),
                   ('encoder_type', 'cnn')],
                  [('encoder_type', 'rnn'),
                   ('global_attention', None)],
                  [('encoder_type', 'rnn'),
                   ('global_attention', None),
                   ('copy_attn', True),
                   ('copy_attn_type', 'general')],
                  [('encoder_type', 'rnn'),
                   ('global_attention', 'mlp'),
                   ('copy_attn', True),
                   ('copy_attn_type', 'general')],
                  [],
                  ]

if onmt.models.sru.check_sru_requirement():
    #   """ Only do SRU test if requirment is safisfied. """
    # SRU doesn't support input_feed.
    tests_nmtmodel.append([('rnn_type', 'SRU'), ('input_feed', 0)])

for p in tests_nmtmodel:
    _add_test(p, 'nmtmodel_forward')