Tutorials

• Get Started
  • Installation
    • Build LiBai from Source
  • Quick Run
    • Train Bert-large Model Parallelly
    • Train VisionTransformer on ImageNet Dataset
  • LiBai Model Zoo
    • Parallelism Mode in LiBai
    • Supported Models in LiBai
    • Baselines
  • Benchmarks
    • Settings
    • Main Results
• Basics
  • Config System
    • Configs in LiBai
    • Get the Default Config
    • LazyConfig Best Practices
  • Features in LiBai
    • Automatic Mixed Precision Training
    • Gradient Clipping
    • Gradient Accumulation
    • Activation Checkpointing
    • ZeRO
  • Training
    • Trainer Abstraction
    • Customize a DefaultTrainer
    • Logging of Metrics
  • Evaluation
    • Customize Evaluator using DatasetEvaluator
    • Run Evaluator Manually
  • Training & Evaluation in Command Line
    • Training
    • Evaluation
    • Quickly check in the respective loop
  • Load and Save a Checkpoint in LiBai
  • Write Models
    • Construct Models in LiBai
    • Import the model in config
  • Write Dataloaders
    • Build Common Dataloaders
  • Build New Project on LiBai
    • Introduction
    • Main Function Entry
    • Build Config
    • Start Training
  • Distributed Configuration
    • Distributed Setting Example
    • Update Distributed Config with Command Line
  • Auto Parallel Training
    • Installation
    • Train/Evaluate model in auto-parallel mode
• Advanced Tutorials
  • How to Customize Dataloader
    • How the Existing Dataloader Works
    • Use Custom Dataloader
  • How to Customize Parallelism
    • Define your own Parallel Model with LiBai.layers
    • Write your own Pipeline Parallel Model