👑 Easy-to-use and powerful NLP and LLM library with 🤗 Awesome model zoo, supporting wide-range of NLP tasks from research to industrial applications, including 🗂Text Classification, 🔍 Neural Search, ❓ Question Answering, ℹ️ Information Extraction, 📄 Document Intelligence, 💌 Sentiment Analysis etc.
New Features
Pipelines
CodeGen
Text2Image
TextClassification
Prompt Learning
新增Prompt Learning训练框架,支持PET、P-Tuning、RGL等经典模型的快速实现。#2894
Neural Search
UIE
新增多语言模型UIE-M-Base和UIE-M-Large,支持中英文混合抽取及Taskflow一键调用。#3192 新增基于封闭域模型GlobalPointer的UIE数据蒸馏方案,支持Taskflow一键部署。#3136
Compression API
TextGeneration
logits_processor
https://github.com/PaddlePaddle/PaddleNLP/pull/3018get_decoder_start_token_id()
方法获取模型decoder_start_token_id
https://github.com/PaddlePaddle/PaddleNLP/pull/3018get_input_embeddings()
和set_input_embeddings()
方法获取 embeddings https://github.com/PaddlePaddle/PaddleNLP/pull/3018get_vocab()
方法获取词表 https://github.com/PaddlePaddle/PaddleNLP/pull/3018TextSummarization
Bug Fix
model_max_input_size
配置字段 https://github.com/PaddlePaddle/PaddleNLP/pull/3127bos_token_id
出错问题 https://github.com/PaddlePaddle/PaddleNLP/pull/3058model_max_length
问题 https://github.com/PaddlePaddle/PaddleNLP/pull/3018Others
New Contributors