-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_inner_att_model.py
executable file
·155 lines (111 loc) · 6.22 KB
/
run_inner_att_model.py
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
#!/usr/bin/python
# -*- coding: utf-8 -*-
import argparse
from repeval.utils.embeddings import (Glove840B300d, SennaEmbeddings)
from repeval.corpus.snli_corpus import (MultiNLICorpus,
SNLICorpus,
FullNLICorpus)
Corpora = {"MultiNLICorpus": MultiNLICorpus,
"SNLICorpus": SNLICorpus,
"FullNLICorpus": FullNLICorpus}
EmbeddingsList = {"Glove840B300d": Glove840B300d,
"SennaEmbeddings": SennaEmbeddings}
parser = argparse.ArgumentParser(description='PyTorch MultiNLI Classifier')
parser.add_argument('--corpus', type=str, default="MultiNLICorpus",
choices=Corpora,
help='Name of the corpus to use. '
'Choices: ' + " ".join(Corpora.keys()))
parser.add_argument('--embeddings', type=str, default="Glove840B300d",
choices=EmbeddingsList,
help='Name of the embeddings to use. '
'Choices: ' + " ".join(EmbeddingsList.keys()))
parser.add_argument('--context_window_size', type=int, default=1,
help='Size of the context window size')
parser.add_argument('--runs_dir', type=str,
default='./runs',
help='location of the runs directories')
parser.add_argument('--load', type=str, default=None,
help='Model to load either for further training or eval')
parser.add_argument('--write_output', action='store_true',
help='Whether to write evaluation output to a file')
parser.add_argument('--dont_update_db', action='store_false', dest='update_db',
help='Deactivate database writing')
parser.add_argument('--log_interval', type=int, default=200, metavar='N',
help='report interval')
parser.add_argument('--train_embeddings', action='store_true',
help='Enable embedding training. Default: False')
parser.add_argument('--use_char_embeddings', action='store_true',
help='Enable character level embeddigs. Default: False')
parser.add_argument('--dont_train_char_embeddings', action='store_false',
dest='train_char_embeddings',
help='Enable character embedding training. Default: False')
parser.add_argument('--char_merging_method',
choices=['lstm', 'mean', 'sum'],
help='Way in which char embeddings are encoded to produce '
'word reprentations',
default='mean')
parser.add_argument('--char_embedding_dim', type=int, default=20,
help='dimension of the character embedding. Default 20')
parser.add_argument('--pos_tags', action='store_true',
help='Enable POS tag embeddings. Default: False')
parser.add_argument('--pos_embedding_dim', type=int, default=50,
help='dimension of the pos tags embeddings. Default 50')
parser.add_argument('--invert_premises', action='store_true',
help='Invert premises during training')
parser.add_argument('--double_hypotheses', action='store_true',
help='Double hypotheses during training')
parser.add_argument('--differentiate_inputs', action='store_true',
help='Delete common words between premise and hypotheses')
parser.add_argument('--dropout', type=float, default=0.25,
help='dropout applied to layers (0 = no dropout). '
'Default: 0.25')
parser.add_argument('--clip', type=float, default=5,
help='gradient clipping (default=5)')
# Learning Rate
parser.add_argument('--lr', type=float, default=0.001,
help='initial learning rate Default 0.001')
parser.add_argument('--lr_decay', type=float, default=1,
help='learning rate decay (Default: 1, ie. no decay)')
parser.add_argument('--start_decay_at_epoch', type=float, default=1,
help='Epoch at which to begin decaying the learning rate. '
'Default: 1')
parser.add_argument('--update_learning_rate', action='store_true',
help='Do not decay learning rate (may be desirable for '
'some optimzers (e.g. Adam)')
parser.add_argument('--param_init', type=float, default=None,
help='Uniform distribution interval limit for '
'initializing model parameters. '
'[-param_init, param_init]. If None, torch'
'iitialization is used. Default: None')
parser.add_argument('--optim', default='adam',
choices=['sgd', 'adagrad', 'adadelta', 'adam', 'rmsprop'],
help='Optimization method. Default: adam')
parser.add_argument('--lstm_hidden', type=int, default=300,
help='Hidden dimension of the lstm. Default: 300')
parser.add_argument('--lstm_layers', type=int, default=1,
help='Number of layers for the lstm. Default: 1')
parser.add_argument('--epochs', type=int, default=10,
help='Upper epoch limit. Default: 10')
parser.add_argument('--early_stopping_epochs', type=int, default=None,
help="If validation accuracy hasn't improved in "
"early_stopping_epochs, training will stop. "
"Default: None (early stopping will not be applied)")
parser.add_argument('--batch_size', type=int, default=128, metavar='N',
help='batch size. Default: 128')
parser.add_argument('--seed', type=int, default=1111,
help='random seed')
parser.add_argument('--no_cuda', action='store_false', dest='cuda',
help="Don't use CUDA")
parser.add_argument('--mode', type=str, default='train',
choices=('train', 'test'),
help='Whether to train or eval')
def run_main():
import torch.nn as nn
from repeval.models.inner_att import InnerAtt
from repeval.routines import main
args = parser.parse_args()
Model = InnerAtt
loss_function = nn.CrossEntropyLoss()
main(args, Model, Corpora, EmbeddingsList, loss_function)
if __name__ == "__main__":
run_main()