DeepTrain Documentation

Full knowledge and control of the train state.

Installation

pip install deeptrain (without data; see how to run examples ), or clone repository

Examples

MNIST AutoEncoder

Timeseries Classification

Model Health Monitoring

Tracking Weights

Reproducibility

Flexible batch_size

Table of Contents

• Why DeepTrain?
  • What does it do?
  • When is it suitable (and not)?
  • Features
• Basic Example
  • Model maker
  • Model configs
  • DataGenerator (train) configs
  • DataGenerator (validation) configs
  • TrainGenerator configs
  • Create training objects
  • Train
  • Delve deeper
• Advanced Example
  • Configuration
  • Create visualization callback
  • Create training objects
  • Train
  • Phase 2
  • New session w/ changed hyperparams
  • Create new train objects
  • Train
  • Inspect generated logs
• Recommended Usage
• Examples
  • Timeseries Classification
  • Model Health Monitoring
  • Reproducibility
  • Preprocessor & batching logic
  • Flexible batch_size & Faster SSD Loading
  • Model Auto-naming
• Callbacks
  • Basic callbacks
  • MNIST callbacks
• Introspection
  • Inspecting gradients
  • Inspecting internals
• How to ...?
  • Change default configs
  • Run examples
  • Save train state
  • Load train state
  • Use custom train / evaluation function
• How does ... work?
  • TrainGenerator
  • DataGenerator
  • DataLoader
• API Reference
• Framework Comparison
  • DeepTrain vs Pytorch Lightning