-
Notifications
You must be signed in to change notification settings - Fork 10
/
config_sp.jsonnet
227 lines (226 loc) · 7.31 KB
/
config_sp.jsonnet
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
218
219
220
221
222
223
224
225
226
local train_data_path = "./data/train.json";
local validation_data_path = "./data/dev.json";
local elmo_embedding_path = "./data/elmo_embeddings/all/";
local domain_slot_list_path = "./ontology/domain_slot_list_sp.txt";
local base_dim = 612; # 512 + 100
local symbol_dim = 128;
local num_use_val = 25; # number of values that uses value list
local dropout = 0.5;
local bi_dropout = 0.5;
local save_mem = false;
local vocab_min_count = 0;
local loss_scale_by_num_values = false;
local is_embedding_trainable = false;
local use_pre_calc_elmo_embeddings = true;
local use_graph = true;
local word_embeddings = "elmo"; # glove or elmo
local phrase_layer_type = "gru"; # gru or stacked_self_attention
{
"dataset_reader": {
"type": "dstqa",
"lazy": false,
"tokenizer": {
"type": "word",
"word_splitter": {
"type": "spacy",
"pos_tags": false,
"ner": false,
},
},
"token_indexers": if use_pre_calc_elmo_embeddings == false then {
[if word_embeddings == "glove" then "tokens" else "elmo"]:
if word_embeddings == "glove" then {"type": "single_id"}
else {"type": "elmo_characters"},
"token_characters": {
"type": "characters",
"character_tokenizer": {
"byte_encoding": "utf-8",
"end_tokens": [
260
],
"start_tokens": [
259
]
},
"min_padding_length": 5
}
} else {
"token_characters": {
"type": "characters",
"character_tokenizer": {
"byte_encoding": "utf-8",
"end_tokens": [
260
],
"start_tokens": [
259
]
},
"min_padding_length": 5
}
},
"domain_slot_list_path": domain_slot_list_path,
},
"iterator": {
"track_epoch": true,
"type": "bucket",
"batch_size": 36,
"max_instances_in_memory": 1000,
"sorting_keys": [
[
"dialogs",
"num_tokens"
]
]
},
"vocabulary": {
"min_count": {
"tokens": vocab_min_count
},
},
"model": {
"use_graph": use_graph,
"loss_scale_by_num_values": loss_scale_by_num_values,
"use_pre_calc_elmo_embeddings": use_pre_calc_elmo_embeddings,
"elmo_embedding_path": elmo_embedding_path,
"base_dim": base_dim,
"domain_slot_list_path": domain_slot_list_path,
"type": "dstqa",
"dropout": dropout,
"bi_dropout": bi_dropout,
"word_embeddings": word_embeddings,
"initializer": [],
"token_indexers": {
[if word_embeddings == "glove" then "tokens" else "elmo"]:
if word_embeddings == "glove" then {"type": "single_id"}
else {"type": "elmo_characters"},
"token_characters": {
"type": "characters",
"character_tokenizer": {
"byte_encoding": "utf-8",
"end_tokens": [
260
],
"start_tokens": [
259
]
},
"min_padding_length": 5
}
},
"text_field_char_embedder": {
"token_characters": {
"type": "character_encoding",
"dropout": 0.0,
"embedding": {
"embedding_dim": 20,
"num_embeddings": 262
},
"encoder": {
"type": "cnn",
"embedding_dim": 20,
"ngram_filter_sizes": [
5
],
"num_filters": 100
}
}
},
"text_field_embedder": {
"allow_unmatched_keys": true,
[if word_embeddings == "glove" then "tokens" else "elmo"]:
if word_embeddings == "glove" then {
"type": "embedding",
"embedding_dim": 300,
"pretrained_file": "http://nlp.stanford.edu/data/glove.840B.300d.zip",
"trainable": is_embedding_trainable
}
else {
"type": "elmo_token_embedder",
"do_layer_norm": false,
"requires_grad": is_embedding_trainable,
"dropout": 0.0,
"options_file": "https://s3-us-west-2.amazonaws.com/allennlp/models/elmo/2x2048_256_2048cnn_1xhighway/elmo_2x2048_256_2048cnn_1xhighway_options.json",
"weight_file": "https://s3-us-west-2.amazonaws.com/allennlp/models/elmo/2x2048_256_2048cnn_1xhighway/elmo_2x2048_256_2048cnn_1xhighway_weights.hdf5",
},
},
"symbol_embedder": {
"symbols": {
"type": "embedding",
"num_embeddings": 40,
"embedding_dim": symbol_dim
}
},
"phrase_layer":
if phrase_layer_type == "gru" then {
"type": "gru",
"bidirectional": true,
"hidden_size": base_dim / 2,
"input_size": base_dim + symbol_dim + 2 * num_use_val,
"num_layers": 1
} else {
"type": "stacked_self_attention",
"hidden_dim": base_dim / 2,
"projection_dim": base_dim / 2,
"feedforward_hidden_dim": base_dim / 2,
"num_attention_heads": 4,
"input_dim": base_dim + symbol_dim + 2 * num_use_val,
"num_layers": 1,
"use_positional_encoding": true
},
"class_prediction_layer": {
"input_dim": base_dim,
"num_layers": 1,
"hidden_dims": [base_dim],
"activations": ["linear"],
},
"span_prediction_layer": {
"input_dim": base_dim,
"num_layers": 1,
"hidden_dims": [base_dim],
"activations": ["linear"],
},
"span_label_predictor": {
"input_dim": base_dim,
"num_layers": 2,
"hidden_dims": [base_dim, 3],
"activations": ["relu", "linear"],
},
"span_end_encoder": {
"input_dim": base_dim,
"num_layers": 1,
"hidden_dims": base_dim,
"activations": "relu",
},
"span_start_encoder": {
"input_dim": base_dim,
"num_layers": 1,
"hidden_dims": base_dim,
"activations": "relu",
},
},
"train_data_path": train_data_path,
"validation_data_path": validation_data_path,
"trainer": {
"num_serialized_models_to_keep": 1,
"keep_serialized_model_every_num_seconds": 60 * 30,
"cuda_device": 0,
"num_epochs": 1000,
"optimizer": {
"type": "adam",
"lr": 0.001,
}
},
"validation_iterator": {
"track_epoch": true,
"type": "bucket",
"batch_size": 36,
"max_instances_in_memory": 1000,
"sorting_keys": [
[
"dialogs",
"num_tokens"
]
]
}
}