# Facico/Chinese-Vicuna

Chinese-Vicuna: A Chinese Instruction-following LLaMA-based Model —— 一个中文低资源的llama+lora方案，结构参考alpaca

Repository: https://github.com/Facico/Chinese-Vicuna
Canonical: https://ross.abutalabs.com/products/chinese-vicuna
Homepage: https://github.com/Facico/Chinese-Vicuna
Language: C
License: Apache-2.0
License Family: permissive
Topics: llama, alpaca, chinese, vicuna
Last push: 2025-04-18T02:41:35+00:00

## Health v2 (maintenance only)
Score: 38/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 17, release rhythm 35, longevity 90
- inputs: {"age_days": 1260, "days_push": 502, "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 4113, forks 405 (observed 2026-08-28T04:08:35.771216+00:00)

## What it is
Chinese-Vicuna is a low-resource LLaMA+LoRA solution for building Chinese instruction-following language models, structured after Alpaca. It provides fine-tuning, generation, and CPU inference code that can train LLaMA-7B on a single RTX 2080Ti and multi-round chatbots on a single RTX 3090.

## Use cases
- fine-tune llama on chinese instruction data with a single consumer gpu
- train a chinese multi-round chatbot on a 3090
- run a llama-based chinese model on cpu with int4 quantization
- adapt llama to a chinese domain like medical or legal text
- convert and quantize original facebook llama checkpoints
- train llama 13b with qlora 4-bit on an 11gb gpu

## When to choose
- you need a chinese instruction-tuned llama on limited gpu memory
- you want parameter-efficient lora/qlora fine-tuning of llama
- you need cpu-only inference of a llama model in c++

## When to avoid
- you need a state-of-the-art multilingual model rather than a llama-era fine-tune
- you want a managed api instead of self-hosted training and inference
- you need languages other than chinese

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-training, llm-inference, chatbot, machine-learning
- domain: large-language-models, machine-learning, artificial-intelligence
- platform: python, cpp
- tags: llama, lora, qlora, chinese-nlp, instruction-tuning, fine-tuning, low-resource, alpaca, natural-language-processing, gpu, linux

## Member repositories
- Facico/Chinese-Vicuna (main) score 38

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:35.771216+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 2026-08-29T18:23:10.807654+00:00, confidence not recorded.
  - readme: https://github.com/Facico/Chinese-Vicuna (fetched 2026-08-28T04:08:35.771216+00:00, sha 618a25546932)
  - homepage: https://github.com/Facico/Chinese-Vicuna (fetched 2026-08-29T09:14:35.608796+00:00, sha 13751a39b8a1)
- Data as of 2026-08-30T08:39:29.467469+00:00.