-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathexperiments.conf
119 lines (94 loc) · 2.67 KB
/
experiments.conf
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
best {
data_dir = ./data_dir/
# Computation limits.
max_top_antecedents = 50
max_training_sentences = 5
top_span_ratio = 0.4
max_num_extracted_spans = 3900
max_num_speakers = 20
max_segment_len = 256
dataset = "ontonotes"
mention_sigmoid = false
# Learning
bert_learning_rate = 1e-5
task_learning_rate = 2e-4
adam_eps = 1e-8
adam_weight_decay = 1e-2
warmup_ratio = 0.1
max_grad_norm = 1 # Set 0 to disable clipping
gradient_accumulation_steps = 1
# Model hyperparameters.
coref_depth = 1 # when 1: no higher order (except for cluster_merging)
coarse_to_fine = true
fine_grained = true
dropout_rate = 0.3
ffnn_size = 1000
ffnn_depth = 1
num_epochs = 24
feature_emb_size = 20
max_span_width = 30
use_metadata = true
use_features = true
use_segment_distance = true
model_heads = true
use_width_prior = true # For mention score
use_distance_prior = true # For mention-ranking score
# Other.
conll_eval_path = ${best.data_dir}/dev.english.v4_gold_conll # gold_conll file for dev
conll_test_path = ${best.data_dir}/test.english.v4_gold_conll # gold_conll file for test
genres = ["bc", "bn", "mz", "nw", "pt", "tc", "wb"]
eval_frequency = 1000
report_frequency = 100
log_root = ${best.data_dir}
mention_proposer = outside
}
spanbert_base = ${best}{
num_docs = 2802
bert_learning_rate = 2e-05
task_learning_rate = 0.0001
coref_depth = 1
max_segment_len = 384
ffnn_size = 3000
cluster_ffnn_size = 1000
max_training_sentences = 3
neg_sample_rate=0.2
bert_tokenizer_name = bert-base-cased
bert_pretrained_name_or_path = SpanBERT/spanbert-base-cased
}
spanbert_base_greedy = ${best}{
num_docs = 2802
bert_learning_rate = 2e-05
task_learning_rate = 0.0001
coref_depth = 1
max_segment_len = 384
ffnn_size = 3000
cluster_ffnn_size = 1000
max_training_sentences = 3
bert_tokenizer_name = bert-base-cased
bert_pretrained_name_or_path = SpanBERT/spanbert-base-cased
mention_proposer = greedy
}
spanbert_large = ${best}{
num_docs = 2802
bert_learning_rate = 1e-05
task_learning_rate = 0.0003
max_segment_len = 512
ffnn_size = 3000
cluster_ffnn_size = 3000
max_training_sentences = 3
neg_sample_rate=0.2
bert_tokenizer_name = bert-base-cased
bert_pretrained_name_or_path = SpanBERT/spanbert-large-cased
}
spanbert_large_greedy = ${best}{
num_docs = 2802
bert_learning_rate = 1e-05
task_learning_rate = 0.0003
max_segment_len = 512
ffnn_size = 3000
cluster_ffnn_size = 3000
max_training_sentences = 3
bert_tokenizer_name = bert-base-cased
bert_pretrained_name_or_path = SpanBERT/spanbert-large-cased
mention_proposer = greedy
}