Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/kyzhouhzau/Bert-BiLSTM-CRF
This model base on bert-as-service. Model structure : bert-embedding bilstm crf.
https://github.com/kyzhouhzau/Bert-BiLSTM-CRF
Last synced: about 1 month ago
JSON representation
This model base on bert-as-service. Model structure : bert-embedding bilstm crf.
- Host: GitHub
- URL: https://github.com/kyzhouhzau/Bert-BiLSTM-CRF
- Owner: kyzhouhzau
- Created: 2018-12-18T12:37:13.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2018-12-18T13:55:54.000Z (almost 6 years ago)
- Last Synced: 2024-08-02T08:10:01.028Z (4 months ago)
- Language: Python
- Size: 975 KB
- Stars: 38
- Watchers: 4
- Forks: 12
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- awesome-bert - kyzhouhzau/Bert-BiLSTM-CRF - as-service. Model structure : bert-embedding bilstm crf. , (BERT NER task:)
README
# bert_Bilstm_Crf
+ bert-as-service:https://github.com/hanxiao/bert-as-service
+ BERT:https://github.com/google-research/bert
+ BiLSTM-CRF:https://github.com/kyzhouhzau/
This model based on google's Bert and hanxiao's bert-as-service this just a try to use bert in a word-embedding mode.
This model still has some problems need to solve.
such as:
1. the bert-service can catch the speed of model training.
2. should i use trainable=True for each embeddings in each sentence
3. The result on dev data was not very perfect!
4. only trained 9177 batch. if more maybe better!
5. others
Run:
First:start bert service
```
bert-serving-start -max_seq_len 250 -pooling_strategy NONE -pooling_layer -4 -3 -2 -1 -model_dir /home/kyzhou/bert/uncased_L-12_H-768_A-12 -num_wor 4
```
Second:
```
python train.py
```
Third:
```
python eval.py
```
Result:
loss:
on dev set:
