forked from jungokasai/graph_parser
-
Notifications
You must be signed in to change notification settings - Fork 4
/
graph_parser_main.py
executable file
·170 lines (153 loc) · 11.2 KB
/
graph_parser_main.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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
from graph_parser_model import run_model, run_model_test
import os
from argparse import ArgumentParser
import pickle
import sys
parser = ArgumentParser()
subparsers = parser.add_subparsers(title='different modes', dest = 'mode', description='train or test')
train_parser=subparsers.add_parser('train', help='train parsing')
# train options
train_parser.add_argument("--model", dest="model", help="model", default='Parsing_Model')
## data information
train_parser.add_argument("--base_dir", dest="base_dir", help="base directory for data")
train_parser.add_argument("--text_train", dest="text_train", help="text data for training")
train_parser.add_argument("--jk_train", dest="jk_train", help="jk data for training")
train_parser.add_argument("--jw_train", dest="jw_train", help="jw data for training")
train_parser.add_argument("--tag_train", dest="tag_train", help="tag data for training")
train_parser.add_argument("--arc_train", dest="arc_train", help="arc data for training")
train_parser.add_argument("--rel_train", dest="rel_train", help="rel data for training")
train_parser.add_argument("--text_test", dest="text_test", help="text data for testing")
train_parser.add_argument("--jk_test", dest="jk_test", help="jk data for testing")
train_parser.add_argument("--tag_test", dest="tag_test", help="tag data for testing")
train_parser.add_argument("--arc_test", dest="arc_test", help="arc data for testing")
train_parser.add_argument("--rel_test", dest="rel_test", help="rel data for testing")
train_parser.add_argument("--conllu_test", dest="conllu_test", help="conllu data for testing")
## model configuration
### LSTM config
train_parser.add_argument("--lstm", dest="lstm", help="rnn architecutre", type = int, default = 1)
train_parser.add_argument("--bidirectional", dest="bi", help="bidirectional LSTM", type = int, default = 1)
train_parser.add_argument("--max_epochs", dest="max_epochs", help="max_epochs", type=int, default = 100)
train_parser.add_argument("--num_layers", dest="num_layers", help="number of layers", type=int, default = 2)
train_parser.add_argument("--units", dest="units", help="hidden units size", type=int, default = 64)
train_parser.add_argument("--hidden_p", dest="hidden_p", help="keep fraction of hidden units", type=float, default = 1.0)
train_parser.add_argument("--dropout_p", dest="dropout_p", help="keep fraction", type=float, default = 1.0)
train_parser.add_argument("--word_embeddings_file", dest="word_embeddings_file", help="embeddings file", default = 'glovevector/glove.6B.100d.txt')
train_parser.add_argument("--elmo", help="elmo", type=int, default=1)
train_parser.add_argument("--word_dropout", dest="word_dropout", help="keep fraction", type=float, default = 1.0)
train_parser.add_argument("--word_dropout_alpha", dest="word_dropout_alpha", help="keep fraction", type=float, default = -1.0)
train_parser.add_argument("--word_dropout_jw", dest="word_dropout_jw", help="keep fraction", type=float, default = 1.0)
## minus for not doing word dropout with frequency
### MLP config
train_parser.add_argument("--mlp_num_layers", dest="mlp_num_layers", help="number of MLP layers", type=int, default = 1)
train_parser.add_argument("--arc_mlp_units", dest="arc_mlp_units", help="MLP units size", type=int, default = 64)
train_parser.add_argument("--rel_mlp_units", dest="rel_mlp_units", help="MLP units size", type=int, default = 32)
train_parser.add_argument("--joint_mlp_units", dest="joint_mlp_units", help="MLP units size", type=int, default = 500)
train_parser.add_argument("--mlp_prob", dest="mlp_prob", help="MLP units size", type=float, default = 1.0)
### Input Config
train_parser.add_argument("--stag_dim", dest="stag_dim", help="supertag dimension", type=int, default = 5)
train_parser.add_argument("--jk_dim", dest="jk_dim", help="jakcknife dimension", type=int, default = 5)
train_parser.add_argument("--embedding_dim", dest="embedding_dim", help="embedding dim", type=int, default = 100)
train_parser.add_argument("--early_stopping", dest="early_stopping", help="early stopping", type=int, default = 5)
train_parser.add_argument("--input_p", dest="input_dp", help="keep fraction for input", type=float, default = 1.0)
## Char Encoding
## Default same as Ma and Hovy 2016
train_parser.add_argument("--chars_dim", dest="chars_dim", help="character embedding dim", type=int, default = 30)
train_parser.add_argument("--chars_window_size", dest="chars_window_size", help="character embedding dim", type=int, default = 3)
train_parser.add_argument("--nb_filters", dest="nb_filters", help="nb_filters", type=int, default = 30)
### Train Config
train_parser.add_argument("--lrate", dest="lrate", help="lrate", type=float, default = 0.001)
train_parser.add_argument("--seed", dest="seed", help="set seed", type= int, default = 0)
### Scores
#train_parser.add_argument("--punc_test", dest="punc_test", help="punctuation data for testing")
train_parser.add_argument("--content_test", dest="content_test", help="content data for testing")
train_parser.add_argument("--metrics", nargs='+', dest="metrics", help="content data for testing")
## Output Options
train_parser.add_argument("--predicted_arcs_file", dest="predicted_arcs_file", help="filename for predicted arcs")
train_parser.add_argument("--predicted_rels_file", dest="predicted_rels_file", help="filename for predicted rels")
train_parser.add_argument("--predicted_conllu_file", dest="predicted_conllu_file", help="filename for predicted conllu")
#train_parser.add_argument("--predicted_arcs_file_greedy", dest="predicted_arcs_file_greedy", help="filename for predicted arcs")
#train_parser.add_argument("--predicted_rels_file_greedy", dest="predicted_rels_file_greedy", help="filename for predicted rels")
#train_parser.add_argument("--predicted_stags_file", dest="predicted_stags_file", help="filename for predicted rels") ## for joint
#train_parser.add_argument("--predicted_pos_file", dest="predicted_pos_file", help="filename for predicted rels") ## for joint
## test options
test_parser=subparsers.add_parser('test', help='test parser')
### data information
test_parser.add_argument("--base_dir", dest="base_dir", help="base directory for data")
test_parser.add_argument("--text_test", dest="text_test", help="text data for testing")
test_parser.add_argument("--jk_test", dest="jk_test", help="jk data for testing")
test_parser.add_argument("--tag_test", dest="tag_test", help="tag data for testing")
test_parser.add_argument("--arc_test", dest="arc_test", help="tag data for testing")
test_parser.add_argument("--rel_test", dest="rel_test", help="tag data for testing")
test_parser.add_argument("--conllu_test", dest="conllu_test", help="conllu data for testing")
### Model Information
test_parser.add_argument("--model", dest="modelname", help="model name")
### Output Options
test_parser.add_argument("--get_accuracy", help="compute tag accuracy", action="store_true", default=False)
test_parser.add_argument("--save_tags", dest="save_tags", help="save 1-best tags")
test_parser.add_argument("--predicted_arcs_file", dest="predicted_arcs_file", help="filename for predicted arcs")
test_parser.add_argument("--predicted_rels_file", dest="predicted_rels_file", help="filename for predicted rels")
test_parser.add_argument("--predicted_conllu_file", dest="predicted_conllu_file", help="filename for predicted conllu")
#test_parser.add_argument("--predicted_arcs_file_greedy", dest="predicted_arcs_file_greedy", help="filename for predicted arcs")
#test_parser.add_argument("--predicted_rels_file_greedy", dest="predicted_rels_file_greedy", help="filename for predicted rels")
test_parser.add_argument("--predicted_stags_file", dest="predicted_stags_file", help="filename for predicted rels") ## for joint
test_parser.add_argument("--predicted_pos_file", dest="predicted_pos_file", help="filename for predicted rels") ## for joint
test_parser.add_argument("--save_probs", dest="save_probs", help="save probabilities", action="store_true", default=False)
test_parser.add_argument("--get_weight", dest="get_weight", help="save stag weight", action="store_true", default=False)
### Scores
#test_parser.add_argument("--punc_test", dest="punc_test", help="punctuation data for testing")
test_parser.add_argument("--content_test", dest="content_test", help="content data for testing")
test_parser.add_argument("--metrics", nargs='+', dest="metrics", help="content data for testing")
## Jabberwocky
test_parser.add_argument("--top_k", dest="top_k", help="Keep top 300", action="store_true", default=False)
test_parser.add_argument("--k", dest="k", help="Keep k", type=int, default=0)
opts = parser.parse_args()
if opts.mode == "train":
print(opts.metrics)
# opts.base_dir = 'sample_data'
# opts.text_train = 'sample_data/sents/train.txt'
# opts.tag_train = 'sample_data/predicted_stag/train.txt'
# opts.jk_train = 'sample_data/predicted_pos/train.txt'
# opts.arc_train = 'sample_data/arcs/train.txt'
# opts.rel_train = 'sample_data/rels/train.txt'
# opts.text_test = 'sample_data/sents/dev.txt'
# opts.tag_test = 'sample_data/predicted_stag/dev.txt'
# opts.jk_test = 'sample_data/predicted_pos/dev.txt'
# opts.arc_test = 'sample_data/arcs/dev.txt'
# opts.rel_test = 'sample_data/rels/dev.txt'
params = ['bi', 'num_layers', 'units', 'hidden_p', 'dropout_p', 'mlp_num_layers', 'arc_mlp_units', 'rel_mlp_units', 'stag_dim', 'jk_dim', 'embedding_dim', 'input_dp', 'chars_dim', 'nb_filters', 'chars_window_size', 'lrate', 'seed', 'elmo']
model_dir = '{}/'.format(opts.model) + '-'.join(map(lambda x: str(getattr(opts, x)), params))
opts.model_dir = os.path.join(opts.base_dir, model_dir)
if opts.word_dropout < 1.0:
opts.model_dir += '-wd{}'.format(opts.word_dropout)
if opts.word_dropout_alpha > 0.0:
opts.model_dir += '-wa{}'.format(opts.word_dropout_alpha)
if opts.word_dropout_jw < 1.0:
opts.model_dir += '-wj{}'.format(opts.word_dropout_jw)
#if 'glovevector' == opts.word_embeddings_file[:11]:
# opts.model_dir += '-glove'
print('Model Dirctory: {}'.format(opts.model_dir))
if not os.path.isdir(opts.model_dir):
os.makedirs(opts.model_dir)
with open(os.path.join(opts.model_dir, 'options.pkl'), 'wb') as fhand:
pickle.dump(opts, fhand)
opts.predicted_arcs_file = os.path.join(opts.model_dir, opts.predicted_arcs_file)
if not os.path.isdir(os.path.dirname(opts.predicted_arcs_file)):
os.makedirs(os.path.dirname(opts.predicted_arcs_file))
opts.predicted_rels_file = os.path.join(opts.model_dir, opts.predicted_rels_file)
if not os.path.isdir(os.path.dirname(opts.predicted_rels_file)):
os.makedirs(os.path.dirname(opts.predicted_rels_file))
opts.predicted_conllu_file = os.path.join(opts.model_dir, opts.predicted_conllu_file)
if not os.path.isdir(os.path.dirname(opts.predicted_conllu_file)):
os.makedirs(os.path.dirname(opts.predicted_conllu_file))
run_model(opts)
if opts.mode == "test":
with open(os.path.join(os.path.dirname(opts.modelname), 'options.pkl'), 'rb') as foptions:
options=pickle.load(foptions)
## additional features not present in old models
if not hasattr(options, 'word_dropout'):
setattr(options, 'word_dropout', 1.0)
if not hasattr(options, 'word_dropout_alpha'):
setattr(options, 'word_dropout_alpha', 0.0)
if not hasattr(options, 'word_dropout_jw'):
setattr(options, 'word_dropout_jw', 1.0)
run_model_test(options, opts)