A flexible pytorch template for Natural Language Processing based on Bert.

Now, it just support the NLC (Natural Language Classification), NLI (Natural Language Inference) and other simple classification mission. It will support the NER and Machine Comprehension in the future

• pytorch-bert-template
  • Requirements
  • How to use
  • Features
    • Folder Structure
  • Usage
    • Config file format
    • Using config files
    • Resuming from checkpoints
    • Using Multiple GPU
  • Customization
    • Custom CLI options
    • Data Loader
    • Processor
    • Agent
    • Model
    • Testing
    • Eval
    • Predict
    • Additional logging
    • Validation data
    • Checkpoints
  • TensorboardX Visualization
  • TODOs
  • Acknowledgments

• Rename trainer to agent
• Finish the document
• Finish the ParameterSearch Function.
• Monitor and control the agent by Wechat
• Realize the NER
• Realize the Machine Comprehension

This project is inspired by the project pytorch-template by victoresque