Unsloth
开源微调工具，主打在消费级显卡上训练与运行模型：官方称相比标准实现训练速度快约 2 倍、显存占用省约 70%，支持 Llama、Qwen、Gemma、Mistral 等主流家族，可直接导出 GGUF 与 4bit 权重，附 Colab 与本地脚本。适合把开源模型调成垂直任务专用版。 ⭐ GitHub Stars: 78k+ 🔗 文档: https://unsloth.ai/docs
An open-source fine-tuning toolkit aimed at consumer GPUs: the project reports roughly 2x faster training with about 70 percent less VRAM than standard implementations, supports the main Llama, Qwen, Gemma and Mistral families, and exports GGUF plus 4-bit weights directly. Colab notebooks and local scripts are provided, which makes task-specific model tuning practical on one card. ⭐ GitHub Stars: 78k+ 🔗 Docs: https://unsloth.ai/docs
📋 Product Info
🎯 Why This Tool
Unsloth belongs to the Open Source LLMs category on hedirbase. Tagged with: fine-tuning,llm,lora,quantization,open-source.
An open-source fine-tuning toolkit aimed at consumer GPUs: the project reports roughly 2x faster training with about 70 percent less VRAM than standard implementations, supports the main Llama, Qwen,