-
Notifications
You must be signed in to change notification settings - Fork 2
/
config_v2.yaml
103 lines (80 loc) · 2.22 KB
/
config_v2.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
# 实验名称
experiment: "can" # "ori_CAN_with_counting_predicted_class_info_in_full_crohme_with_ori_train_with_gpus1_bs8_lr1" # "ori_CAN_with_counting_predicted_class_info_in_full_crohme_with_gpus1_bs8_lr1" # "ori_CAN_with_counting_predicted_class_info_in_full_crohme_with_gpus1_bs8_lr1_ddp" # "ori_CAN_in_full_crohme"
sim_loss:
type: l2
use_flag: False
context_loss: False
word_state_loss: False
counting_decoder:
use_flag: True
in_channel: 684
out_channel: 111
# 随机种子
seed: 20211024
# 训练参数
epochs: 50
batch_size: 8 # 8
workers: 5 # 0
train_parts: 1
valid_parts: 1
valid_start: 0 # 1000000000
save_start: 0 # 220
optimizer: Adadelta
lr: 0.01 # 1
lr_decay: cosine
step_ratio: 10
step_decay: 5
eps: 1e-6
weight_decay: 1e-4
beta: 0.9
# output_counting_feature: False
# output_channel_attn_feature: False
# counting_loss_ratio: 1
dropout: True
dropout_ratio: 0.5
relu: True
gradient: 100
gradient_clip: True
use_label_mask: False
use_loc_mask: False
# 训练数据
train_image_path: '/liuzhuang7/CROHME/train_images.pkl'
train_label_path: '/liuzhuang7/CROHME/train_labels.txt'
eval_image_path: '/liuzhuang7/CROHME/14_test_images.pkl'
eval_label_path: '/liuzhuang7/CROHME/14_test_labels.txt'
16_eval_image_path: '/liuzhuang7/CROHME/16_test_images.pkl'
16_eval_label_path: '/liuzhuang7/CROHME/16_test_labels.txt'
19_eval_image_path: '/liuzhuang7/CROHME/19_test_images.pkl'
19_eval_label_path: '/liuzhuang7/CROHME/19_test_labels.txt'
matrix_path: '/liuzhuang7/CROHME/symbol_statistic_v1.pkl'
word_path: '/liuzhuang7/CROHME/words_dict.txt'
data_process:
reverse_color: False
equal_range: False
# collate_fn
collate_fn: collate_fn
densenet:
ratio: 16
growthRate: 24
reduction: 0.5
bottleneck: True
use_dropout: True
encoder:
input_channel: 1
out_channel: 684
decoder:
net: Decoder_v3
cell: 'GRU'
input_size: 256
hidden_size: 256
out_size: 256
attention:
attention_dim: 512
word_conv_kernel: 1
whiten_type: None
max_step: 256
optimizer_save: False
finetune: True
checkpoint_dir: 'checkpoints'
checkpoint: "checkpoints/can-l2-context-word_2022-12-08-10-19_decoder-Decoder_v3/2014_can-l2-context-word_2022-12-08-10-19_decoder-Decoder_v3_WordRate-0.9037_ExpRate-0.5720_179.pth"
log_dir: 'logs_v2'