#7979th of 90 in Open models
llama3-chinese-chat
Curated Llama 3 Chinese fine-tunes with training data and deployment guides
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 4.1k
- License
- MIT
- Last commit
- Feb 2026
- Language
- Python
4,108 on 25 Apr 20264,131 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
A Chinese-language index of Llama 3 and 3.1 fine-tunes (SFT, DPO, ORPO, MoE, long-context, multimodal), plus the author's own 8B Chinese weights and datasets. It includes deployment notes for vLLM, Ollama, LM Studio, a hand-written API and Streamlit demos, and a Transformers inference script with optional 4-bit loading. Most entries are links to models hosted on ModelScope, Hugging Face and OpenCSG.
Who it is for: Chinese-speaking users picking or running Llama 3 Chinese fine-tunes
Strengths
- Links many community Llama 3 Chinese variants: SFT, DPO, ORPO, MoE, long-context, multimodal
- Publishes training data and DPO settings (beta 0.5, LoRA rank 128) for its own models
- Covers vLLM (OpenAI format), Ollama, LM Studio, Streamlit and plain Transformers
- GGUF build of the 3.1 DPO model runs with one ollama command
Weaknesses
- Mostly a link list; model weights are hosted elsewhere, not in the repo
- README is in Chinese only, with no English documentation
- Base+SFT model trained on common dialogue only and can answer off-target
- Several listed items (70B, Chinese 262k, music, novel writing) are only marked as planned
What it needs
- GPU optional
- Needs transformers 4.40.1, streamlit, peft, bitsandbytes, Ollama, vLLM, LM Studio
- Models: Llama 3 8B, Llama 3.1 8B, GGUF, Hugging Face Transformers, LoRA adapters
Also in Open models
See all 90| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 82 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 80 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 79 out of 100 |