Welcome to PaddleNLP

PaddleNLP is the natural language processing development library of PaddlePaddle, featuring user-friendly text domain APIs, multi-scenario application examples, and high-performance distributed training. It aims to enhance the modeling efficiency in the text domain for PaddlePaddle developers and provides a wealth of NLP application examples.

• User-friendly Text Domain APIs

  • Offers a rich set of industry-grade preset task capabilities Taskflow and comprehensive text domain APIs: supports a wide range of Chinese dataset loading with Dataset API, enables flexible and efficient data preprocessing with Data API, provides over 60+ pretrained word vectors with Embedding API, and offers over 100+ pretrained models with Transformer API, significantly boosting the efficiency of NLP task modeling.

• Multi-scenario Application Examples

  • Covers NLP application examples from academic to industry levels, including fundamental NLP technologies, NLP system applications, and related extended applications. Developed entirely based on the new API system of the PaddlePaddle core framework 2.0, it provides developers with best practices in the text domain of PaddlePaddle.

• High-performance Distributed Training

  • Based on the leading automatic mixed precision optimization strategy of the PaddlePaddle core framework, combined with the distributed Fleet API, it supports 4D hybrid parallel strategies, efficiently completing large-scale pretrained model training.

• Project GitHub: PaddlePaddle/PaddleNLP

• Project Gitee: https://gitee.com/paddlepaddle/PaddleNLP

• GitHub Issue Feedback: PaddlePaddle/PaddleNLP#issues

• WeChat Group: Scan the QR code on WeChat and fill out the questionnaire to join the group for in-depth communication with numerous community developers and the official team.

Model Library List

• Introduction to Popular Models
• Model Library List

Quick Start

• Installation
• Text Generation
• Quick Training
• Quick Inference

Paddle LLM Training

• Paddle LLM Main Documentation
• LLM - Pre-training Documentation
• LLM - Fine-tuning Documentation
• LLM - DPO Documentation
• LLM - RLHF Documentation
• Model Merging Documentation

Paddle LLM Inference

• Docker Deployment - Quick Start Guide
• LLM Inference Tutorial
• Practical Optimization
• Static Graph Model List
• Inference Quantization Tutorial for Various Models
• Heterogeneous Device Inference
• LLM - Quantization Tutorial

PaddleNLP Toolkit

• PaddleNLP One-Stop Prediction: Taskflow API
• Single input
• Single input
• Multiple inputs
• Multiple inputs
• Pre-trained Word Embeddings

PaddleNLP Tutorials

• Transformer Pre-trained Model
• Trainer API Training Tutorial
• Dialogue Template Tutorial
• Multi-Turn Dialogue Fine-Tuning Tutorial
• How to Use chat_template for Training
• How to Customize System Prompt
• Chinese Sentiment Analysis Tutorial
• Model Compression Tutorial
• Data Distillation Tutorial
• Torch2Paddle Weight Conversion Tutorial

Evaluation Metrics

• Evaluation Metrics

Practical Tutorials

• AI Studio Notebook

Advanced Guide

• Model Compression
• Large-Scale Distributed Training

Community Collaboration

• How to Contribute Models
• How to Contribute Datasets
• How to Contribute Documentation Examples
• How to Join Interest Groups

FAQ

• PaddleNLP Frequently Asked Questions (Continuously Updated)
• ⭐️【Selection】Top 5 NLP Questions
• ⭐️【Theory Section】General NLP Issues
• ⭐️【Practical Section】PaddleNLP Practical Issues

API Reference

• paddlenlp.data
• paddlenlp.datasets
• paddlenlp.layers
• paddlenlp.losses
• paddlenlp.metrics
• paddlenlp.ops
• paddlenlp.seq2vec
• paddlenlp.taskflow
• paddlenlp.trainer
• paddlenlp.transformers
• paddlenlp.utils

Indices and tables

• Index

• Module Index

• Search Page