generated from allenai/allennlp-template-config-files
/
flair_clone.jsonnet
181 lines (178 loc) · 7.09 KB
/
flair_clone.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
local transformer_model_name = std.extVar("EMBEDDING_MODEL_NAME");
local corpus_name = std.extVar("CORPUS");
local embedding_dim = std.parseInt(std.extVar("EMBEDDING_DIMS")); # uniquely determined by transformer_model
local features = {
"features": {
"nuc_children": {"source_key": "nuc_children"},
"sat_children": {"source_key": "sat_children"},
"genre": {"source_key": "genre", "label_namespace": "genre"},
"u1_discontinuous": {"source_key": "u1_discontinuous", "label_namespace": "discontinuous"},
"u2_discontinuous": {"source_key": "u2_discontinuous", "label_namespace": "discontinuous"},
"u1_issent": {"source_key": "u1_issent", "label_namespace": "issent"},
"u2_issent": {"source_key": "u2_issent", "label_namespace": "issent"},
"unit1_case": {"source_key": "unit1_case", "label_namespace": "case"},
"unit2_case": {"source_key": "unit2_case", "label_namespace": "case"},
"u1_depdir": {"source_key": "u1_depdir", "label_namespace": "depdir"},
"u2_depdir": {"source_key": "u2_depdir", "label_namespace": "depdir"},
"u1_func": {"source_key": "u1_func", "label_namespace": "func"},
"u2_func": {"source_key": "u2_func", "label_namespace": "func"},
"length_ratio": {"source_key": "length_ratio"},
"same_speaker": {"source_key": "same_speaker", "label_namespace": "same_speaker"},
"doclen": {"source_key": "doclen"},
//"distance": {"source_key": "distance"},
"distance": {
"source_key": "distance",
"xform_fn": {
"type": "bins",
"bins": [[-1e9, -8], [-8, -2], [-2, 0], [0, 2], [2, 8], [8, 1e9]]
},
"label_namespace": "distance"
},
"u1_position": {
"source_key": "u1_position",
"xform_fn": {
"type": "bins",
"bins": [[0.0, 0.1], [0.1, 0.2], [0.2, 0.3], [0.3, 0.4], [0.4, 0.5], [0.5, 0.6], [0.6, 0.7], [0.7, 0.8], [0.8, 0.9], [0.9, 1.0], [1.0, 1e9]]
//"bins": [[0.0, 0.2], [0.2, 0.4], [0.4, 0.6], [0.6, 0.8], [0.8, 1.0], [1.0, 1e9]]
},
"label_namespace": "u1_position"
},
"u2_position": {
"source_key": "u2_position",
"xform_fn": {
"type": "bins",
"bins": [[0.0, 0.1], [0.1, 0.2], [0.2, 0.3], [0.3, 0.4], [0.4, 0.5], [0.5, 0.6], [0.6, 0.7], [0.7, 0.8], [0.8, 0.9], [0.9, 1.0], [1.0, 1e9]]
//"bins": "bins": [[0.0, 0.2], [0.2, 0.4], [0.4, 0.6], [0.6, 0.8], [0.8, 1.0], [1.0, 1e9]]
},
"label_namespace": "u2_position"
},
"lex_overlap_length": {
"source_key": "lex_overlap_length",
"xform_fn": {
"type": "bins",
"bins": [[0, 2], [2, 7], [7, 1e9]]
},
"label_namespace": "lex_overlap"
}
},
"corpus": corpus_name,
// By default, we will use all features for a corpus, but they can be overridden below.
// The values inside the array need to match a key under the "features" dict above.
"corpus_configs": {
"deu.rst.pcc": ["distance", "u1_depdir", "u2_depdir", "u2_func", "u1_position", "u2_position",
"sat_children", "nuc_children"],
"eng.pdtb.pdtb": ['u2_depdir', 'u2_func', 'u2_issent', 'u2_position', 'length_ratio'],
"eng.rst.gum": ["distance", "same_speaker", "u2_func", "u2_depdir", "unit1_case", "unit2_case", "nuc_children",
"sat_children", "genre", "lex_overlap_length", "u2_discontinuous", "u1_discontinuous",
"u1_position", "u2_position"],
"eng.rst.rstdt": ['u2_discontinuous', 'u1_position', 'u2_position', 'u2_func', 'u2_issent'],
"eng.sdrt.stac": ["same_speaker"],
"eus.rst.ert": ["u2_position"],
"fra.sdrt.annodis": ["u1_depdir", "u1_position"],
"nld.rst.nldt": ['distance', 'u1_depdir', 'sat_children', 'genre', 'u1_position'],
"por.rst.cstn": ['u2_discontinuous', 'u1_position', 'u2_position'],
"rus.rst.rrt": ['distance', 'nuc_children', 'sat_children', 'u1_position', 'u2_position', 'u1_depdir',
'u2_depdir', 'u2_func', 'u1_issent', 'u2_issent'],
"spa.rst.rststb": ["distance", "u1_depdir", "u1_discontinuous", "u2_depdir", "sat_children", "u2_func", "genre"],
"spa.rst.sctb": ['distance', 'u1_position', 'sat_children'],
"tur.pdtb.tdb": ["distance", "u1_depdir", "u2_depdir", "u2_func", "u1_issent", "u2_issent", "length_ratio",
"u1_position", "u2_position"],
"zho.pdtb.cdtb": ["distance", "u1_depdir", "u2_depdir", "u2_func", "u1_issent", "u2_issent", "length_ratio"],
"zho.rst.sctb": ['sat_children', 'nuc_children', 'genre', 'u2_discontinuous', 'u1_discontinuous', 'u1_depdir', 'u1_func'],
"fas.rst.prstc": ['distance', 'nuc_children', 'sat_children', 'u2_discontinuous', 'genre'],
}
};
// For small corpora, make this number reflect the size of train
// For larger corpora, use a smaller number, aiming for 1/3 of total size
local batches_per_epoch = {
"deu.rst.pcc": 541,
"eng.pdtb.pdtb": 3000, // real: 10980
"eng.rst.gum": 1700, // real: 3475
"eng.rst.rstdt": 2000, // real: 4001
"eng.sdrt.stac": 1200, // real: 2395
"eus.rst.ert": 634,
"fas.rst.prstc": 1025,
"fra.sdrt.annodis": 547,
"nld.rst.nldt": 402,
"por.rst.cstn": 1037,
"rus.rst.rrt": 2500, // real: 7217
"spa.rst.rststb": 560,
"spa.rst.sctb": 110,
"tur.pdtb.tdb": 613,
"zho.pdtb.cdtb": 915,
"zho.rst.sctb": 110
};
{
"dataset_reader" : {
"type": "disrpt_2021_rel_flair_clone",
"token_indexers": {
"tokens": {
"type": "pretrained_transformer",
"model_name": transformer_model_name,
"max_length": 511
}
},
"tokenizer": {
"type": "pretrained_transformer",
"model_name": transformer_model_name
},
"features": features
},
"train_data_path": std.extVar("TRAIN_DATA_PATH"),
"validation_data_path": std.extVar("VALIDATION_DATA_PATH"),
"model": {
"type": "disrpt_2021_flair_clone",
"embedder": {
"type": "featureful_bert",
"model_name": transformer_model_name,
"max_length": 511,
"train_parameters": true,
"last_layer_only": true
},
"seq2vec_encoder": {
"type": "bert_pooler",
"pretrained_model": transformer_model_name
},
"feature_dropout": 0.0,
"features": features,
},
"data_loader": {
"batches_per_epoch": batches_per_epoch[corpus_name],
// NOTE: if you need to change batch size, scale batches_per_epoch, which assumes
// a batch size of 4, by an appropriate amount. E.g., if you need to make batch
// size 2, then use `batches_per_epoch[corpus_name] * 2`
"batch_size": 4,
"shuffle": true
},
"trainer": {
"num_epochs": 100,
"patience": 12,
"optimizer": {
"type": "huggingface_adamw",
"lr": 2e-5,
#"weight_decay": 0.05,
#"betas": [0.9, 0.99],
#"parameter_groups": [
# [[".*embedder.*transformer.*"], {"lr": 2e-5}]
#],
},
#"learning_rate_scheduler": {
# "type": "slanted_triangular",
# "num_epochs": 50,
# "cut_frac": 0.1,
#},
"learning_rate_scheduler": {
"type": "reduce_on_plateau",
"factor": 0.6,
"mode": "max",
"patience": 2,
"verbose": true,
"min_lr": 5e-7
},
//"learning_rate_scheduler": {
// "type": "cosine",
// "t_initial": 5,
//},
"validation_metric": "+relation_accuracy"
}
}