/
experiments.conf
97 lines (88 loc) · 2.08 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
# Word embeddings.
glove_300d {
path = data/glove.840B.300d.txt
size = 300
}
glove_300d_filtered {
path = data/glove.840B.300d.txt.filtered
size = 300
}
glove_300d_2w {
path = data/glove_50_300_2.txt
size = 300
}
glove_300d_2w_filtered {
path = data/glove_50_300_2.txt.filtered
size = 300
}
# Main configuration.
best {
# Computation limits.
max_top_antecedents = 50
max_training_sentences = 50
top_span_ratio = 0.4
# Model hyperparameters.
filter_widths = [3, 4, 5]
filter_size = 50
char_embedding_size = 8
char_vocab_path = "data/char_vocab.txt"
context_embeddings = ${glove_300d_filtered}
head_embeddings = ${glove_300d_2w_filtered}
contextualization_size = 200
contextualization_layers = 3
ffnn_size = 150
ffnn_depth = 2
feature_size = 20
max_span_width = 20
use_metadata = true
use_features = true
model_heads = true
coref_depth = 2
lm_layers = 3
lm_size = 1024
num_cdd_pool = 30
use_im = true
im_emb_size = 512
vis_weight = 0.4
ffnn_size_im = 100
ffnn_depth_im = 1
# End-to-End + Visual baseline
use_im_fc = false
im_fc_feat_path = data/resnet152_feat.hdf5
im_fc_feat_size = 2048
im_layer = 0
im_fc_emb_size = 512
im_dropout_rate = 0
# Learning hyperparameters.
max_gradient_norm = 5.0
lstm_dropout_rate = 0.4
lexical_dropout_rate = 0.5
dropout_rate = 0.2
optimizer = adam
learning_rate = 0.001
decay_rate = 0.999
decay_frequency = 100
random_seed = 2019
max_step = 50000
# Other.
train_path = data/train.vispro.1.1.jsonlines
eval_path = data/val.vispro.1.1.jsonlines
lm_path = data/elmo_cache.vispro.hdf5
cdd_path = data/cdd_np.vispro.1.1.jsonlines
lm_cdd_path = data/elmo_cache.vispro_cdd.hdf5
im_obj_label_path = data/mscoco_label.jsonlines
lm_obj_path = data/elmo_cache.vispro_mscoco.hdf5
eval_frequency = 5000
report_frequency = 100
log_root = /home/yuxintong/pr4vd/Visual_PCR/logs
}
best_predict = ${best} {
context_embeddings = ${glove_300d}
head_embeddings = ${glove_300d_2w}
}
e2e_baseline = ${best} {
use_im = false
}
e2e_visual_baseline = ${best} {
use_im_fc = true
}