A toolkit for Relation Extraction and more...
Project description
REx
A toolkit for Relation Extraction and more...
This project has not finished yet, so be careful when using it.
Dependencies
- Python>=3.6
- torch>=1.2.0 : project is developed with torch==1.5.1, should be compatable with >=1.2.0 versions
- numpy>=1.19.0
Installation
pip install -e .
QuickStart
mkdir project_dir && cd project_dir
python -m rex startproject project_name # planned so, but has not finished
Structures
- io: data loading, transformation operations and data building
- utils: data load/dump
- instance: common wrapper to make dict structure data able to access their attributes by
obj.attribute
- vocab: vocabulary building
- transform: data cleaning/transformation, word cutting, and some utils for constructing data
- core: core tools for data preprocess
- dataset: abstract dataset wrapper
- loader: wrapper for data loaders
- trainer: main runner for model training process
- metrics: evaluation measures, including many kinds of metrics
- accuracy
- (all|filtered)_(f1|precision|recall)
- micro
- macro
- average
- (max|all)_(all|filtered)_(prc|auc)
- p@k
- modules: basic modules
- embeddings
- word2vec
- glove
- fasttext
- tf-idf
- elmo
- bert
- encoders
- sentence encoders
- bag encoders
- selectors
- MaxBag
- OneBag
- AttBag
- decoders
- crf
- SimpleCRF
- ConstraintCRF
- span
- SingleSpan
- DualSpan
- crf
- embeddings
- models: entire models
- utils: others for building projects
- config
- logger
Dataset Compatibility Adaptation
- IPRE preprocess
- NYT10
- NYT-H
- SemEval2010-Task8
- TACRED
- ACE05
Support
- Chinese sentence level relation extraction
- Chinese bag level RE
- English sentence level RE
- English bag level RE
Test
python -m unittest -v
Docs
cd docs
sphinx-apidoc -o . ..
make html
LICENCE
MIT
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
pytorch-rex-0.0.0.tar.gz
(2.8 kB
view hashes)
Built Distribution
Close
Hashes for pytorch_rex-0.0.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 6e59bb79506ab4a956cb133b5acc1f06c490d4fc62e21dd264a1ed1be67eb4f6 |
|
MD5 | b30da4037f813e9280380632164aba6a |
|
BLAKE2b-256 | cf1b24acace09538d920b5b2a94d2ccddcd50f0d86c9ecaf528278b67fa37fc9 |