# thu-coai/CDial-GPT

A Large-scale Chinese Short-Text Conversation Dataset and Chinese pre-training dialog models

Repository: https://github.com/thu-coai/CDial-GPT
Canonical: https://ross.abutalabs.com/products/cdial-gpt
Language: Python
License: MIT
License Family: permissive
Topics: dialogue, pytorch, gpt, gpt-2, lccc, text-generation
Last push: 2023-06-12T21:26:44+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2277, "days_push": 1178, "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 1962, forks 258 (observed 2026-08-28T04:05:59.587914+00:00)

## What it is
CDial-GPT provides LCCC, a large-scale cleaned Chinese short-text conversation dataset, along with Chinese GPT-2-based dialogue models pre-trained and fine-tuned on it. The codebase, adapted from TransferTransfo with HuggingFace Transformers, supports pre-training and fine-tuning of Chinese conversational models.

## Use cases
- train a Chinese chatbot
- get a cleaned Chinese conversation dataset
- fine-tune a GPT-2 model on Chinese dialogue
- research multi-turn Chinese conversation generation
- load LCCC dataset via HuggingFace datasets
- pre-train a Chinese dialogue model

## When to choose
- you need a large, quality-filtered Chinese dialogue corpus
- you want a ready-made Chinese conversational GPT baseline
- you are doing NLP research on Chinese short-text conversation

## When to avoid
- you need multilingual or English dialogue data
- you need production-grade chatbot infrastructure rather than models and data
- you need actively maintained code - the project is in maintenance mode

## Facets
- artifact type: dataset
- maturity: maintenance
- function: machine-learning, nlp, llm-training, chatbot, data-science
- domain: large-language-models, chatbots, machine-learning
- platform: python
- tags: chinese, dialogue-dataset, gpt-2, pytorch, text-generation, pretrained-model, lccc, conversation-modeling, natural-language-processing

## Member repositories
- thu-coai/CDial-GPT (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:59.587914+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 2026-08-30T03:05:36.158899+00:00, confidence not recorded.
  - readme: https://github.com/thu-coai/CDial-GPT (fetched 2026-08-28T04:05:59.587914+00:00, sha 38421de587ab)
- Data as of 2026-08-30T08:39:29.467469+00:00.