# dbiir/UER-py

Open Source Pre-training Model Framework in PyTorch & Pre-trained Model Zoo

Repository: https://github.com/dbiir/UER-py
Canonical: https://ross.abutalabs.com/products/uer-py
Homepage: https://github.com/dbiir/UER-py/wiki
Language: Python
License: Apache-2.0
License Family: permissive
Topics: bert, pre-training, fine-tuning, gpt, chinese, natural-language-processing, pytorch, elmo, classification, ner, t5, unilm, roberta, albert, clue, gpt-2, model-zoo, bart, pegasus, xlm-roberta
Last push: 2024-05-09T11:12:55+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2702, "days_push": 846, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3112, forks 520 (observed 2026-08-28T04:07:43.987051+00:00)

## What it is
UER-py is a PyTorch framework for pre-training transformer language models (BERT, GPT-2, T5, ELMo, etc.) and fine-tuning them on downstream NLP tasks, with a modular encoder/decoder architecture. It also ships a model zoo of pre-trained checkpoints, particularly strong for Chinese NLP.

## Use cases
- pre-train a BERT model from scratch on my own corpus
- fine-tune a pre-trained transformer for text classification
- download Chinese pre-trained language models
- train a GPT-2 model in PyTorch
- run named entity recognition with a fine-tuned BERT
- experiment with modular transformer architectures for NLP research

## When to choose
- you need to pre-train or fine-tune medium-sized (<1B parameter) text models in PyTorch
- you work on Chinese NLP tasks and want a strong model zoo
- you want modular, research-friendly control over transformer components

## When to avoid
- you need multi-modal or very large model training - use its successor TencentPretrain instead
- you just want inference with an off-the-shelf model via Hugging Face Transformers
- you need production serving infrastructure rather than training tooling

## Facets
- artifact type: framework
- maturity: maintenance
- function: llm-training, machine-learning, nlp, deep-learning
- domain: machine-learning, deep-learning, large-language-models
- platform: python, cross-platform
- tags: pytorch, bert, pre-training, fine-tuning, model-zoo, chinese-nlp, gpt-2, t5, transformers, natural-language-processing, gpu

## Member repositories
- dbiir/UER-py (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:43.987051+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 2026-08-30T07:26:42.065336+00:00, confidence not recorded.
  - readme: https://github.com/dbiir/UER-py (fetched 2026-08-28T04:07:43.987051+00:00, sha edd54b530a91)
  - homepage: https://github.com/dbiir/UER-py/wiki (fetched 2026-08-29T09:41:51.122010+00:00, sha 19d3ba8540ca)
- Data as of 2026-08-30T08:39:29.467469+00:00.