FrameTrain
Local AI model training & fine-tuning — a cross-platform desktop app.
FrameTrain is a desktop application for training and fine-tuning AI models locally on your own hardware — no cloud required. It brings LoRA/QLoRA fine-tuning of Hugging Face models, PyTorch-based training with GPU acceleration (NVIDIA CUDA and Apple Silicon / Metal MPS), a visual neural-network builder, dataset management, live training monitoring and automatic model versioning together in one GUI.
Because training runs entirely on your machine, your datasets and models never leave your device — GDPR-compliant by design.
This is the official source repository for the FrameTrain desktop app.
Ready-to-install builds are on the website and under Releases.
Features
Fine-tuning & training
- LoRA and QLoRA fine-tuning of large language models
- PyTorch-based training pipeline
- Import Hugging Face models or load local ones
- Export trained models, including GGUF for local inference
- Ollama integration for running results locally
Synapse Builder — visual neural networks
- Build custom architectures on a drag-and-drop canvas (powered by React Flow)
- Transformer, CNN and LSTM building blocks
- Train directly from the canvas
Data & workflow
- Dataset import, Parquet support, train/validation/test splitting
- Live training monitoring with loss curves
- AI training analysis & coach for diagnosing runs
- Automatic model versioning and comparison of training runs
Runs everywhere, stays private
- Windows, macOS and Linux
- NVIDIA CUDA and Apple Silicon (M1/M2/M3/M4 via Metal MPS) — no CUDA required on Mac
- Fully local and offline
See the app in action on the screenshots tour.
Download
Get the latest build for your platform:
Requirements: a 64-bit OS; for LLM fine-tuning an NVIDIA GPU or an Apple Silicon Mac is recommended (8 GB unified memory minimum, 16 GB+ for 7B models). See the Apple Silicon guide for what your Mac can handle.
Build from source
FrameTrain is a Tauri 2 app with a React + TypeScript frontend.
Prerequisites
Setup
# install frontend dependencies
npm install
# run the desktop app in development (hot reload)
npm run tauri:dev
# build a production desktop binary for your platform
npm run tauri:build
Frontend-only commands (without the desktop shell) are also available:
npm run dev # Vite dev server
npm run build # type-check + build the web frontend
npm run preview # preview the built frontend
Tech stack
- Shell: Tauri 2 (Rust) — small, secure, native binaries
- Frontend: React 18, TypeScript, Vite, Tailwind CSS
- Visual builder: React Flow (
@xyflow/react)
- Training core: PyTorch, Hugging Face, PEFT/LoRA, QLoRA, BitsAndBytes
- Testing: Vitest, Testing Library
Project structure
src/ React + TypeScript frontend (UI, AI logic, datasets, Synapse Builder)
src-tauri/ Rust backend (native shell, AI proxy, file system, updater)
scripts/ Build and maintenance scripts
Learn more
License
FrameTrain is proprietary software. See LICENSE for the terms that apply to this repository.
Made by FrameSphere.
