-
Notifications
You must be signed in to change notification settings - Fork 2
/
meta_evaluator.yaml
141 lines (131 loc) · 3.47 KB
/
meta_evaluator.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
system:
# Path where all intermediate and final results will be written:
root_folder: e:/GPM_Sim/meta_evaluation
# Path where data was stored using the CdmProcessor class:
sequence_data_folder: e:/GPM_Sim/pretraining/person_sequence
# Path where train data was stored using the GeneralPretrainedModelTools package:
train_data_folder: e:/GPM_Sim/train
# Names of the sub folders holding the labels of the train data:
train_label_sub_folders:
- label_c1000000
- label_c1000011
- label_c1000022
- label_c1000033
- label_c1000044
- label_c1000055
- label_c1000066
- label_c1000077
- label_c1000088
- label_c1000099
# Path where train data was stored using the GeneralPretrainedModelTools package:
test_data_folder: e:/GPM_Sim/test
# Names of the sub folders holding the labels of the train data:
test_label_sub_folders:
- label_c1000000
- label_c1000011
- label_c1000022
- label_c1000033
- label_c1000044
- label_c1000055
- label_c1000066
- label_c1000077
- label_c1000088
- label_c1000099
# Maximum number of CPU cores to use:
max_cores: 5
# Batch size for training and testing: (default: 32)
batch_size: 32
checkpoint_every: 50
pretrained models:
- name: pretrained_model_1
learning objectives:
masked_concept_learning: yes
mask_one_concept_per_visit: yes
masked_visit_concept_learning: yes
truncate_type: random
label_prediction: no
training:
train_fraction: 0.8
num_epochs: 100
num_freeze_epochs: 1
learning_rate: 0.001
weight_decay: 0.01
max_batches:
model:
max_sequence_length: 512
concept_embedding: yes
segment_embedding: yes
age_embedding: yes
date_embedding: yes
visit_order_embedding: yes
visit_concept_embedding: yes
hidden_size: 768
num_hidden_layers: 12
num_attention_heads: 12
intermediate_size: 3072
hidden_act: gelu
embedding_combination_method: sum
hidden_dropout_prob: 0.1
attention_probs_dropout_prob: 0.1
- name: pretrained_model_2
learning objectives:
masked_concept_learning: yes
mask_one_concept_per_visit: yes
masked_visit_concept_learning: yes
truncate_type: random
label_prediction: no
training:
train_fraction: 0.8
num_epochs: 100
num_freeze_epochs: 1
learning_rate: 0.001
weight_decay: 0.01
max_batches:
model:
max_sequence_length: 512
concept_embedding: yes
segment_embedding: yes
age_embedding: yes
date_embedding: yes
visit_order_embedding: yes
visit_concept_embedding: yes
hidden_size: 768
num_hidden_layers: 12
num_attention_heads: 12
intermediate_size: 3072
hidden_act: gelu
embedding_combination_method: concat
hidden_dropout_prob: 0.1
attention_probs_dropout_prob: 0.1
fine-tuned models:
- name: fine_tuned_model_1
pretrained models:
- name: pretrained_model_1
pretrained_epoch: 100
- name: pretrained_model_1
pretrained_epoch: 50
- name: pretrained_model_2
learning objectives:
label_prediction: yes
truncate_type: tail
training:
train_fraction: 1
num_epochs: 100
num_freeze_epochs: 1
learning_rate: 0.001
weight_decay: 0.01
max_batches:
- name: fine_tuned_model_2
pretrained models:
- name: pretrained_model_1
- name: pretrained_model_2
learning objectives:
label_prediction: yes
truncate_type: tail
training:
train_fraction: 1
num_epochs: 100
num_freeze_epochs: 0
learning_rate: 0.001
weight_decay: 0.01
max_batches: