LoRAbit

LLM Trainer based on MLX

Only for Mac

Free

Mac

MLX-powered LoRA training, kept local and organized. Prepare datasets, check readiness, train adapters, and export results in one workspace. LoRAbit is an on-device LoRA training workspace built for users who want a clear, local workflow for preparing datasets, checking model readiness, running preflight analysis, training adapters with MLX, and exporting reusable results. LoRAbit helps you keep the full workflow organized in one place: • Create and manage local training sessions • Choose local MLX-ready model folders or safetensors model folders • Prepare datasets from common local text, document, and structured data formats • Expand selected folders into supported dataset files • Review selected files before preprocessing • Run a conservative preflight check before training • Train LoRA adapters locally with MLX • Export MLX adapter bundles and inference-ready MLX model folders • Review logs, reports, and generated artifacts inside the active workspace LoRAbit is designed around local ownership. Your selected models, datasets, logs, training runs, and exported artifacts stay inside the workspace you choose. The app does not require an account and does not depend on a backend service for its core workflow. This version also adds broader localization and appearance controls, so the app can follow your system language and theme preferences more naturally. --- If you experience any inconvenience or have suggestions for improvement, please let me know. I will review them carefully and reflect them as quickly as possible.

  • This app hasn’t received enough ratings or reviews to display an overview.

This update improves the workspace and LoRA training workflow. • A clearer path from model selection and dataset preparation to preflight, training, and export. The configuration window now consistently follows the existing light or dark appearance. • A new User Guide button opens a LoRA introduction, walkthrough, glossary, and troubleshooting guide. A sample dataset and practice configuration are included. • The session library is now separate from the current workspace, with search, renaming, and multi-selection deletion. • Improved training progress display and log detail, with copying and deletion by selection or filter. • Refined MLX dataset processing and memory management, with improved reliability when stopping training, continuing from an adapter, or exporting the result. • More reliable configuration saving and restoration, with a retry action after a failed save. • The developer’s app catalog is available from the sidebar, configuration window, and Help menu. This version requires an Apple Silicon Mac running macOS 27 or later.

The developer, JEAHOON BANG, indicated that the app’s privacy practices may include handling of data as described below. For more information, see the developer’s privacy policy .

  • Data Not Collected

    The developer does not collect any data from this app.

    Privacy practices may vary, for example, based on the features you use or your age. Learn More

    The developer has not yet indicated which accessibility features this app supports. Learn More

    Seller
    JEAHOON BANG
    Size
    10.4 MB
    Category
    Utilities
    Compatibility
    Requires macOS 27.0 or later and a Mac with Apple M1 chip or later.
    • Mac
      Requires macOS 27.0 or later and a Mac with Apple M1 chip or later.
    Languages
    English and 5 more
    • English, French, German, Japanese, Korean, Spanish
    Age Rating
    4+
    Copyright
    © JeaHoon, Bang