hummingbird
/
fastNLP
Not watched
2
0
Fork 0
Code
Releases 13 Wiki Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
1208 Commits
3 Branches
25 MB
0 Download
 
 
 
 
 
Tree: 0032f7788a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '0032f7788a'
${ noResults }
fastNLP/reproduction/joint_cws_parse
yh 7a21c2a587 1. 增强BertEmbedding使其可以自动判断token_type_ids; 2.增加CrossEntropyLoss中对label dimension的报错处理 5 years ago
..
data 将DataInfo修改为DataBundle 5 years ago
models 1. 增强BertEmbedding使其可以自动判断token_type_ids; 2.增加CrossEntropyLoss中对label dimension的报错处理 5 years ago
README.md 增加fastNLP.embeddings模块并修改对应的现有代码以适配fastNLP.embeddings 5 years ago
__init__.py 增加joint_cws_parse的代码 5 years ago
train.py 1. 增强BertEmbedding使其可以自动判断token_type_ids; 2.增加CrossEntropyLoss中对label dimension的报错处理 5 years ago

README.md

Code for paper A Unified Model for Chinese Word Segmentation and Dependency Parsing

准备数据

  1. 数据应该为conll格式，1, 3, 6, 7列应该对应为'words', 'pos_tags', 'heads', 'labels'.
  2. 将train, dev, test放在同一个folder下，并将该folder路径填入train.py中的data_folder变量里。
  3. 从百度云(提取:ua53)下载预训练vector，放到同一个folder下，并将train.py中vector_folder变量正确设置。

运行代码

python train.py

其它

ctb5上跑出论文中报道的结果使用以上的默认参数应该就可以了(应该会更高一些); ctb7上使用默认参数会低0.1%左右，需要调节
learning rate scheduler.

Delete task

Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.

cancel
confirm

一款轻量级的自然语言处理（NLP）工具包，目标是减少用户项目中的工程型代码，例如数据处理循环、训练循环、多卡运行等

Python Jupyter Notebook Text CSV Markdown