-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.log
53 lines (53 loc) · 1.69 KB
/
train.log
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
--------------------------------------
some config:
data_dir = ./resource/data
output_dir = ./output
cache_dir = ./cache
model_name = R-BERT
mode = 0
seed = 1234
cuda = 0
plm_root_dir = ./resource
plm_name = bert-base-uncased
epoch = 5
max_len = 128
batch_size = 16
lr = 2e-05
other_lr = 2e-05
weight_decay = 0.0
adam_epsilon = 1e-08
gradient_accumulation_steps = 1
warmup_proportion = 0.1
max_grad_norm = 1.0
dropout = 0.1
device = cuda:0
model_dir = ./output/R-BERT
plm_dir = ./resource/bert-base-uncased
--------------------------------------
start to load data ...
get the result of tokenization from ./resource/data/train.json
get the result of tokenization from ./resource/data/test.json
finish loading train!
finish loading dev!
finish loading test!
finish!
--------------------------------------
traning model parameters (except PLM layers):
cls_mlp.weight : torch.Size([768, 768])
cls_mlp.bias : torch.Size([768])
entity_mlp.weight : torch.Size([768, 768])
entity_mlp.bias : torch.Size([768])
dense.weight : torch.Size([19, 2304])
dense.bias : torch.Size([19])
--------------------------------------
start to train the model ...
[001] train_loss: 1.672 | dev_loss: 0.605 | micro f1 on dev: 0.8458 >>> save models!
[002] train_loss: 0.509 | dev_loss: 0.510 | micro f1 on dev: 0.8737 >>> save models!
[003] train_loss: 0.260 | dev_loss: 0.568 | micro f1 on dev: 0.8747 >>> save models!
[004] train_loss: 0.135 | dev_loss: 0.592 | micro f1 on dev: 0.8868 >>> save models!
[005] train_loss: 0.071 | dev_loss: 0.624 | micro f1 on dev: 0.8906 >>> save models!
--------------------------------------
start load model ...
--------------------------------------
start test ...
test_loss: 0.624 | micro f1 on test: 0.8906