Welcome to SkiNet documentation!¶ Getting started Software development process Development environment Docker Lightning Studio Running tests Linting and type checking Kaggle notebooks Azure Azure storage setup Identification of datasets Access data in Azure Blob Storage Running experiments on Azure Configuration guide YAML blocks → models GENERAL_CONFIG Run and experiment naming Cross-cutting rules Minimal working config Data Data ISIC 2017 dataset Metadata lazy loading Datasets and dataloaders Quickstart Dataset Class Dataset Splits Modifications to the default Dataset class Augmentation of data Note about modification to PyTorch’s default dataloader Motivation Dataloaders and subprocesses How SkiNet prevents new worker spawning: RepeatDataLoader Usage with ISIC 2017 Custom collate function Model and training Network Design Architecture — U-Net 2D (custom, from scratch) Architecture diagram Block design Candidate blocks (considered, not selected) Model selection Training setup Augmentation pipeline (training only) Inference pipeline Training Set up training Lightning model Model architecture Best threshold selection Callbacks Ways to start training inside a configured environment (Docker container) Launching training from Lightning Studio Training monitoring Reproducibility Exporting to ONNX Quick start CLI reference What the script does ONNX graph iOS / mobile preprocessing constants Auto-discovery rules (--run mode) Optional dependencies Checkpoint key remapping Performance and debugging Debugging Debug configuration NaN / loss divergence GPU memory exhaustion (OOM) DataLoader bottleneck diagnosis Checkpoint / resume errors Plotting Logging Plot images and masks side-by-side Plot images overlaid with masks Plot augmented data Contributing Documentation Building documentation locally (for debugging or preview) Deploy on Github Pages Experiments Experiments API Reference API Reference Experiment Config Lightning Model Model Configs Training Configs Training Utilities Transform Configs Data Configs Datasets API Reference Logging and Callbacks Experiment Keys Optuna Utilities