fastai v2.7.8

Concepts

• Concepts - Data loading
• Concepts - Model training
• Concepts - Learner lifecycle

Learner

• Learner - Create an instance
• Learner - Init, Attributes
• Learner - Training methods
• Learner - Inference methods

Visualize

• Diagnostics - How to debug
• Show - Inputs, targets, predictions
• Show - Images
• Show - Text, points, boxes, tables
• Plot - Training loop

Evaluate

• Model evaluation - Interpretation
• Metrics - 1/2
• Metrics - 2/2

Training

• Learner - Training loop 1/2
• Learner - Training loop 2/2
• Learner - Customize training loop
• Learner - Callbacks 1/2
• Learner - Callbacks 2/2
• Learner - Context managers

Inference

• Learner - Public methods call tree
• Learner - validate and Recorder
• Learner - get_preds and Loss function
• Learner - predict and DataLoader
• Learner - show_results and DataLoader

DataLoader

• DataLoaders - Create an instance
• DataLoaders - Interface
• DataLoader - Interface
• DataLoader - Init
• DataLoader - iter() and next()
• DataLoader - TfmdDL
• DataLoader - TfmdDL subclasses

Dataset

• Dataset - Interface
• Dataset - TfmdLists
• Dataset - Datasets
• Dataset - Tabular datasets

DataBlock

• DataBlock
• DataBlock - Init data pipeline
• TransformBlocks - Labels
• TransformBlocks - Vision
• TransformBlocks - Text

Download

• Download datasets and models
• Download - Directories config

Load

• Dataset, Model, Learner - Directories
• Get items
• Splitters
• Getters

Transform

• Transforms for labels

Vision

• Type Transforms - Vision
• Item Transforms - Vision
• Data augmentation - Vision 1/4
• Data augmentation - Vision 2/4
• Data augmentation - Vision 3/4
• Data augmentation - Vision 4/4

Text

• Type Transforms - Text
• Item Transforms - Text

Tabular

• Tabular datasets and transforms

Optimizer

• Optimizer
• Optimizers
• Optimizer - Hyperparameters scheduling

Model

• Create model - Vision
• Create Model - Text, GAN

Pytorch

• Modules - Functions, Shapes, Pooling
• Modules - Combine layers, In Out
• Modules - Activations, Norms
• Modules - Convolutions, Attention
• Modules - Text sequences, Dropout
• Modules - Unet, GAN, Tabular

Loss

• Loss function - Interface
• Loss functions - Classification, Regression

Summary

• Summary - DataBlock
• Summary - DataLoaders
• Summary - Learner
• Implementation notes
• Distributed Training