AdaSeq: An All-in-One Library for Developing State-of-the-Art Sequence Understanding Models

AdaSeq (Alibaba Damo Academy Sequence understanding toolkit) is an easy-to-use all-in-one library, built on ModelScope, that allows researchers and developers to train custom models for sequence understanding tasks, including word segmentation, POS tagging, chunking, NER, entity typing, relation extraction, etc.

⚠️Notice: This project is under quick development. This means some interfaces could be changed in the future.

---

Features:

• State-of-the-Art: we provide plenty of cutting-edge models, training methods and useful toolkits for sequence understanding tasks.
• Easy-to-Use: one line of command is all you need to obtain the best model.
• Extensible: easily register new tasks, models, modules, criterions, optimizers, lr_schedulers and training methods.

What's New:

• 2022-11: [EMNLP 2022] Released NPCRF code
• 2022-11: [EMNLP 2022] Released BABERT models

Previous updates

Supported Models

• Transformer-based CRF
• Partial CRF
• Retrieval Augmented NER
• Global-Pointer
• Multi-label Entity Typing
• ...

Quick Start

Requirements and Installation

Python version >= 3.7

git clone https://github.com/modelscope/adaseq.git
pip install -r requirements.txt -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

train a model

python scripts/train.py -c examples/bert_crf/configs/resume.yaml

test a model

python scripts/test.py -c examples/bert_crf/configs/resume.yaml -cp ${checkpoint_path}

Tutorials

• Tutorial 1: Training a Model & Configuration Explanation
• Tutorial 2: Preparing Custom Dataset
• Tutorial 3: Hyperparameter Tuning with Grid Search
• Tutorial 4: Training with Multiple GPUs

License

This project is licensed under the Apache License (Version 2.0).