How-to guides

Choose a path by the outcome you need. Running the train → predict → evaluate workflows is covered by the configuration reference — each of its pages starts with the commands that run its workflow.

Start here

Read Adopt KonfAI from PyTorch, MONAI, or nnU-Net if you already use PyTorch, MONAI, nnU-Net, or Lightning. Read Processing large images when regional I/O and bounded host memory matter. Using custom models, transforms, augmentations, and losses covers extension contracts and Docker covers a containerized runtime. Apps and agent automation are first-class sections in the main navigation: Using KonfAI Apps and Agent workflows (MCP server).

• Adopt KonfAI from PyTorch, MONAI, or nnU-Net
• Processing large images
• Using custom models, transforms, augmentations, and losses
• Docker