shrdlu9's picture
Upload 15 files
f57e181
***** Test results *****
Thu Sep 22 06:41:21 2022
Task: ner
Model path: bert-base-uncased
Data path: ./data/ud/
Tokenizer: bert-base-uncased
Batch size: 32
Epoch: 6
Learning rate: 2e-05
LR Decay End Factor: 0.3LR Decay End Epoch: 5Sequence length: 96
Training: True
Num Threads: 24
Num Sentences: 0
Max Grad Norm: 0.0
Use GNN: False
Syntax graph style: dep
Use label weights: False
Clip value: 50
precision recall f1-score support
CARDINAL 0.7133 0.6503 0.6803 612
DATE 0.6922 0.7254 0.7084 1045
EVENT 0.4429 0.3875 0.4133 80
FAC 0.3390 0.3974 0.3659 151
GPE 0.8456 0.8714 0.8583 1936
LANGUAGE 0.5135 0.2468 0.3333 77
LAW 0.4130 0.3333 0.3689 57
LOC 0.5934 0.4977 0.5414 217
MONEY 0.5370 0.4754 0.5043 61
NORP 0.6211 0.7536 0.6809 422
ORDINAL 0.8208 0.8304 0.8256 171
ORG 0.5289 0.5869 0.5564 857
PERCENT 0.3333 0.4722 0.3908 36
PERSON 0.7192 0.7885 0.7523 1371
PRODUCT 0.2705 0.3367 0.3000 98
QUANTITY 0.3485 0.4340 0.3866 53
SEP] 0.0000 0.0000 0.0000 0
TIME 0.6071 0.6355 0.6210 214
WORK_OF_ART 0.3000 0.2538 0.2750 130
micro avg 0.6487 0.7110 0.6784 7588
macro avg 0.5073 0.5093 0.5033 7588
weighted avg 0.6821 0.7110 0.6946 7588
Special token predictions: 0