# brightmart/roberta_zh

RoBERTa中文预训练模型: RoBERTa for Chinese

Repository: https://github.com/brightmart/roberta_zh
Canonical: https://ross.abutalabs.com/products/roberta_zh
Language: Python
License Family: other
Topics: roberta, chinese, bert, pre-trained-language-models, pre-trained, gpt2
Last push: 2024-07-22T15:02:06+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": 2557, "days_push": 772, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2791, forks 408 (observed 2026-08-28T04:07:22.094960+00:00)

## What it is
A repository of pre-trained RoBERTa models for Chinese, implemented in TensorFlow with PyTorch versions also provided, loadable directly as BERT models. It includes multiple model sizes trained on ~30GB of Chinese text covering news, community Q&A, and encyclopedic data.

## Use cases
- fine-tune a Chinese language model for text classification
- download a pretrained RoBERTa checkpoint for Chinese NLP tasks
- improve Chinese question answering with a stronger BERT-like encoder
- compare Chinese pretrained models like BERT, XLNet, and RoBERTa
- use a Chinese transformer model in TensorFlow or PyTorch

## When to choose
- you need a strong pretrained Chinese encoder compatible with BERT loading code
- you work in TensorFlow or PyTorch and want ready-made Chinese model weights
- you want a smaller 6-layer or 12-layer model for faster experimentation

## When to avoid
- you need a maintained library with active development and a license
- you need multilingual or English-only models
- you want the latest architectures like LLMs or instruction-tuned models

## Facets
- artifact type: dataset
- maturity: maintenance
- function: machine-learning, nlp, deep-learning
- domain: machine-learning, large-language-models
- platform: python
- tags: roberta, chinese, pretrained-models, bert, tensorflow, pytorch, model-weights, natural-language-processing, gpu

## Member repositories
- brightmart/roberta_zh (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:22.094960+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 2026-08-30T08:15:15.883358+00:00, confidence not recorded.
  - readme: https://github.com/brightmart/roberta_zh (fetched 2026-08-28T04:07:22.094960+00:00, sha dd3033c78e96)
- Data as of 2026-08-30T08:39:29.467469+00:00.