cotk documentation

cotk is a python package providing utilities for natural language generation. It contains benchmark data loader, word vector loader, pretrained baseline models and other useful utilities for evaluating your models fairly with baselines.

Tutorials

• Installation
  • Requirements
  • Install from pip
  • Install from source code
• Quick Start
  • Dataloader
  • Metrics
  • Predefined Models
• Practice: Implement a GRU Language Model
  • Preparing the data
  • Training models
  • Evaluations
  • Additional: Word Vector
• CLI Usage
  • Import
  • Config
• Extending Cotk: More Data, More Metrics!
  • Add A New Dataset
  • Add A New Task
  • Add A New Metric
• Frequently Asked Questions
  • On windows, why my codes seem running from beginning after calling metric.close ?

Package Reference

• Data Loader
  • Overview
  • Dataloader
  • LanguageProcessing
  • LanguageGeneration
  • SingleTurnDialog
  • MultiTurnDialog
  • SentenceClassification
  • Field
  • Tokenizer
  • Vocab
  • Context
• Word Vector
  • WordVector
  • GeneralWordVector
• Metric
  • Hash Value
  • Basic Classes
  • Metric class
  • Metric-like class
• Resources
  • File ID
  • Word Vector
  • Datasets
• file_utils
• resources_processor

Model Zoo

• LanguageGeneration
  • Language Model (TensorFlow)
  • VAE (TensorFlow)
  • SeqGAN (TensorFlow)
• SingleTurnDialog
  • Seq2Seq (PyTorch)
  • Seq2Seq (TensorFlow)
• MultiTurnDialog
  • HRED (TensorFlow)
  • CVAE (TensorFlow)

Indices and tables

• Index

• Module Index