Introduction

Langtrain provides a complete suite of tools for modern AI development, allowing you to train custom models and orchestrate complex agents without compromising data privacy. Our ecosystem consists of four core pillars:1. Langtrain Studio: A desktop app for managing datasets and fine-tuning runs.2. Langtune: An open-source Python library for fine-tuning text models on your own GPU.3. Langvision: An open-source Python library for fine-tuning vision-language models.4. Langtrain SDK: Python and TypeScript clients for the Langtrain API.

Studio is a native desktop app for your Langtrain account, available for Windows and Linux.

• –Datasets: Upload and organise training files.
• –Fine-tuning runs: Start a run and follow its metrics and logs as it trains.
• –Secure sign-in: Authenticator-app sign-in, with credentials kept in the system keychain.

Langtune is an open-source (MIT) Python library for parameter-efficient fine-tuning of text models such as Llama, Mistral, Qwen and Phi.

• –Methods: LoRA, QLoRA, DoRA, IA³ and prefix tuning, plus DPO, ORPO and KTO for preference training.
• –Runs locally: Training happens on your GPU; it only goes to the cloud when a Langtrain API key is set.
• –Install: pip install "langtrain-ai[train]" langtune.

Langvision is an open-source (MIT) Python library for fine-tuning vision-language models on your own images and text.

• –Models: LLaVA, Qwen-VL, InternVL, CogVLM and Phi-3 Vision.
• –Data: One JSON line per image, with the conversation about it.
• –Export: Save the trained model as ONNX or safetensors.
• –Install: pip install langvision, or pip install langtrain-ai[vision].

Open text models from Hugging Face that Transformers and PEFT support, such as Llama, Mistral, Qwen, Gemma and Phi, and the vision-language models listed above for Langvision. See Models for choosing one.

Ready to get started?

• –Quick start - Train your first model
• –Langtune - Fine-tune on your own GPU
• –Langvision - Fine-tune a vision-language model
• –Langtrain Studio - Download the desktop app