# PaddlePaddle/PaddleNLP

Easy-to-use and powerful LLM and SLM library with awesome model zoo.

Repository: https://github.com/PaddlePaddle/PaddleNLP
Canonical: https://ross.abutalabs.com/products/paddlenlp
Homepage: https://paddlenlp.readthedocs.io
Language: Python
License: Apache-2.0
License Family: permissive
Topics: nlp, embedding, bert, ernie, paddlenlp, pretrained-models, transformers, information-extraction, question-answering, search-engine, semantic-analysis, sentiment-analysis, neural-search, uie, document-intelligence, compression, llm, distributed-training, llama
Last push: 2026-05-23T02:54:28+00:00

## Health v2 (maintenance only)
Score: 63/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 83, release rhythm 16, longevity 100
- inputs: {"age_days": 2035, "days_push": 102, "days_rel": 469, "gap_med": 155, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 12967, forks 3031 (observed 2026-08-28T04:11:00.274163+00:00)

## What it is
PaddleNLP is an easy-to-use NLP and large language model development kit built on the PaddlePaddle deep learning framework, with a large pretrained model zoo. It supports efficient LLM/SLM training, lossless compression, and high-performance inference across multiple hardware platforms including NVIDIA GPUs, Kunlun XPUs, Ascend NPUs, and other accelerators.

## Use cases
- train or fine-tune large language models like Llama, Qwen, DeepSeek, and ERNIE
- run high-performance quantized LLM inference on GPUs and domestic accelerators
- extract structured information from documents with UIE
- build question answering and semantic search systems
- perform sentiment analysis and text classification in Chinese and English
- compress large models with quantization and distillation
- run distributed pretraining with tensor and pipeline parallelism

## When to choose
- you are in the PaddlePaddle ecosystem or need Chinese-focused NLP models like ERNIE and UIE
- you need multi-hardware training and inference including Ascend NPU or Kunlun XPU
- you want an all-in-one kit covering pretraining, fine-tuning, alignment, compression, and deployment
- you need very high-throughput quantized inference for models like DeepSeek V3/R1

## When to avoid
- your stack is PyTorch-based and you prefer Hugging Face Transformers compatibility
- you only need lightweight inference without the PaddlePaddle framework dependency
- your team has no experience with PaddlePaddle and you want the largest community ecosystem

## Facets
- artifact type: library
- maturity: active
- function: llm-training, llm-inference, nlp, machine-learning, deep-learning, rag, transformers
- domain: large-language-models, machine-learning, deep-learning
- platform: python, windows
- tags: paddlepaddle, model-zoo, information-extraction, uie, distributed-training, model-compression, quantization, chinese-nlp, ernie, llama, qwen, deepseek, fine-tuning, rlhf, natural-language-processing, linux, macos, gpu, docker

## Member repositories
- PaddlePaddle/PaddleNLP (main) score 63

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:00.274163+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:13:38.765705+00:00, confidence not recorded.
  - readme: https://github.com/PaddlePaddle/PaddleNLP (fetched 2026-08-28T04:11:00.274163+00:00, sha bbc1306f28dd)
  - homepage: https://paddlenlp.readthedocs.io (fetched 2026-08-29T08:09:31.758147+00:00, sha 718c1c748704)
  - registry_pypi: https://pypi.org/pypi/paddlenlp/json (fetched 2026-08-29T08:09:31.768477+00:00, sha 8a497feb6613)
- Data as of 2026-08-30T08:39:29.467469+00:00.