forked from OpenNMT/OpenNMT-py
-
Notifications
You must be signed in to change notification settings - Fork 0
/
java_med2m_features.yaml
155 lines (144 loc) · 6.24 KB
/
java_med2m_features.yaml
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
data:
corpus_1:
path_src: ../synpos/data/java-med2m-processed/train_src.txt
path_tgt: ../synpos/data/java-med2m/java-med2m.TargetType.seq.train.target.txt
src_feats:
feat_0: ../synpos/data/java-med2m-processed/train_feat0.txt
feat_1: ../synpos/data/java-med2m-processed/train_feat1.txt
feat_2: ../synpos/data/java-med2m-processed/train_feat2.txt
feat_3: ../synpos/data/java-med2m-processed/train_feat3.txt
feat_4: ../synpos/data/java-med2m-processed/train_feat4.txt
feat_5: ../synpos/data/java-med2m-processed/train_feat5.txt
feat_6: ../synpos/data/java-med2m-processed/train_feat6.txt
feat_7: ../synpos/data/java-med2m-processed/train_feat7.txt
feat_8: ../synpos/data/java-med2m-processed/train_feat8.txt
feat_9: ../synpos/data/java-med2m-processed/train_feat9.txt
feat_10: ../synpos/data/java-med2m-processed/train_feat10.txt
feat_11: ../synpos/data/java-med2m-processed/train_feat11.txt
feat_12: ../synpos/data/java-med2m-processed/train_feat12.txt
feat_13: ../synpos/data/java-med2m-processed/train_feat13.txt
feat_14: ../synpos/data/java-med2m-processed/train_feat14.txt
feat_15: ../synpos/data/java-med2m-processed/train_feat15.txt
feat_16: ../synpos/data/java-med2m-processed/train_feat16.txt
feat_17: ../synpos/data/java-med2m-processed/train_feat17.txt
feat_18: ../synpos/data/java-med2m-processed/train_feat18.txt
feat_19: ../synpos/data/java-med2m-processed/train_feat19.txt
feat_20: ../synpos/data/java-med2m-processed/train_feat20.txt
feat_21: ../synpos/data/java-med2m-processed/train_feat21.txt
feat_22: ../synpos/data/java-med2m-processed/train_feat22.txt
feat_23: ../synpos/data/java-med2m-processed/train_feat23.txt
transforms: [filterfeats, onmt_tokenize, inferfeats] #, filtertoolong]
weight: 1
valid:
path_src: ../synpos/data/java-med-processed/val_src.txt
path_tgt: ../synpos/data/java-med/java-med.TargetType.seq.val.target.txt
src_feats:
feat_0: ../synpos/data/java-med-processed/val_feat0.txt
feat_1: ../synpos/data/java-med-processed/val_feat1.txt
feat_2: ../synpos/data/java-med-processed/val_feat2.txt
feat_3: ../synpos/data/java-med-processed/val_feat3.txt
feat_4: ../synpos/data/java-med-processed/val_feat4.txt
feat_5: ../synpos/data/java-med-processed/val_feat5.txt
feat_6: ../synpos/data/java-med-processed/val_feat6.txt
feat_7: ../synpos/data/java-med-processed/val_feat7.txt
feat_8: ../synpos/data/java-med-processed/val_feat8.txt
feat_9: ../synpos/data/java-med-processed/val_feat9.txt
feat_10: ../synpos/data/java-med-processed/val_feat10.txt
feat_11: ../synpos/data/java-med-processed/val_feat11.txt
feat_12: ../synpos/data/java-med-processed/val_feat12.txt
feat_13: ../synpos/data/java-med-processed/val_feat13.txt
feat_14: ../synpos/data/java-med-processed/val_feat14.txt
feat_15: ../synpos/data/java-med-processed/val_feat15.txt
feat_16: ../synpos/data/java-med-processed/val_feat16.txt
feat_17: ../synpos/data/java-med-processed/val_feat17.txt
feat_18: ../synpos/data/java-med-processed/val_feat18.txt
feat_19: ../synpos/data/java-med-processed/val_feat19.txt
feat_20: ../synpos/data/java-med-processed/val_feat20.txt
feat_21: ../synpos/data/java-med-processed/val_feat21.txt
feat_22: ../synpos/data/java-med-processed/val_feat22.txt
feat_23: ../synpos/data/java-med-processed/val_feat23.txt
transforms: [filterfeats, onmt_tokenize, inferfeats]
src_onmttok_kwargs: "{'mode': 'space'}"
tgt_onmttok_kwargs: "{'mode': 'space'}"
# # Vocab opts
src_vocab: java_med/data_shared/data.vocab.src
tgt_vocab: java_med/data_shared/data.vocab.src
src_feats_vocab:
feat_0: java_med/data/vocab/data.vocab.feat_0
feat_1: java_med/data/vocab/data.vocab.feat_1
feat_2: java_med/data/vocab/data.vocab.feat_2
feat_3: java_med/data/vocab/data.vocab.feat_3
feat_4: java_med/data/vocab/data.vocab.feat_4
feat_5: java_med/data/vocab/data.vocab.feat_5
feat_6: java_med/data/vocab/data.vocab.feat_6
feat_7: java_med/data/vocab/data.vocab.feat_7
feat_8: java_med/data/vocab/data.vocab.feat_8
feat_9: java_med/data/vocab/data.vocab.feat_9
feat_10: java_med/data/vocab/data.vocab.feat_10
feat_11: java_med/data/vocab/data.vocab.feat_11
feat_12: java_med/data/vocab/data.vocab.feat_12
feat_13: java_med/data/vocab/data.vocab.feat_13
feat_14: java_med/data/vocab/data.vocab.feat_14
feat_15: java_med/data/vocab/data.vocab.feat_15
feat_16: java_med/data/vocab/data.vocab.feat_16
feat_17: java_med/data/vocab/data.vocab.feat_17
feat_18: java_med/data/vocab/data.vocab.feat_18
feat_19: java_med/data/vocab/data.vocab.feat_19
feat_20: java_med/data/vocab/data.vocab.feat_20
feat_21: java_med/data/vocab/data.vocab.feat_21
feat_22: java_med/data/vocab/data.vocab.feat_22
feat_23: java_med/data/vocab/data.vocab.feat_23
feat_merge: 'sharemlp'
feat_vec_size: 32
#src_vocab_size: 2000
src_vocab_size: 32000
share_vocab: True
share_embeddings: True
copy_attn: True
save_data: java_med_data
overwrite: False
save_checkpoint_steps: 1000
keep_checkpoint: 30
seed: 3435
train_steps: 600000
valid_steps: 1000
warmup_steps: 8000
report_every: 100
# Uri:
#early_stopping: 10
#early_stopping_criteria: accuracy
valid_batch_size: 8192
decoder_type: transformer
encoder_type: transformer
word_vec_size: 512
rnn_size: 512
#layers: 6
enc_layers: 6
dec_layers: 6
transformer_ff: 2048
heads: 8
accum_count: [4]
accum_steps: [0]
optim: adam
adam_beta1: 0.9
adam_beta2: 0.998
decay_method: noam
learning_rate: 2.0
max_grad_norm: 0.0
queue_size: 500
bucket_size: 32768
batch_size: 4096
batch_type: tokens
normalization: tokens
dropout: 0.1
attention_dropout: [0.1]
label_smoothing: 0.1
max_generator_batches: 2
param_init: 0.0
param_init_glorot: 'true'
#position_encoding: 'true'
max_relative_positions: 200
world_size: 2
gpu_ranks:
- 0
- 1