User Guide# Quick Start# Quick Start Step 1: Use Trainer Step 2: Get Log Step 3: Display of progress Step 4: Save Model Trainer# Trainer and Evaluator Concepts Trainer at a glance Snapshot Handler Runtime Logic for Custom Training and Evaluation Default Logic (ppe.handler.Logic) CodeBlock Logic (ppe.handler.Logic) Runtimes for Custom Devices Use of ppe.to to transfer modules and batches to custom devices Runtime Registry CodeBlocks for Abstracting Logic Steps Executing CodeBlocks Extensions# Extensions Manager Extensions How to use Creating Extensions Reporting Distributed Snapshot Utilities# Lazy Modules Config Basic Substitution pytorch_pfn_extras.onnx Installation API CUDA (CuPy Interoperability)