Welcome to MMEngine’s documentation!

You can switch between Chinese and English documents in the lower-left corner of the layout.

Get Started

• Introduction
• Installation
• 15 minutes to get started with MMEngine

Examples

• Train a GAN

Common Usage

• Resume Training
• Distributed Training
• Speed up Training
• Save Memory on GPU
• Better performance optimizers
• Visualize Training Logs
• Set Random Seed
• Debug Tricks
• Calculate the FLOPs and Parameters of Model
• Setting the Frequency of Logging, Checkpoint Saving, and Validation
• EpochBasedTraining to IterBasedTraining

Tutorials

• Runner
  • Example codes of the runner
    • A beginer-friendly example
    • Explanations on example codes
  • Best practice of the Runner - config files
  • Basic dataflow
  • Why use the runner? (Optional reading)
  • Suggestions on next steps
• Dataset and DataLoader
  • Details on dataloader
    • sampler and shuffle
    • DefaultSampler
    • The obscure collate_fn
  • Details on dataset
    • Use torchvision datasets
    • Customize your dataset
    • Use MMEngine BaseDataset
• Model
  • Runner and model
  • Interface introduction
  • DataPreprocessor
• Evaluation
  • Evaluation in model training or testing
    • Using a single evaluation metric
    • Using multiple evaluation metrics
    • Customizing evaluation metrics
  • Using offline results for evaluation
• OptimWrapper
  • OptimWrapper vs Optimizer
    • Model training
    • Get learning rate/momentum
    • Export/load state dicts
    • Use multiple optimizers
    • Configure the OptimWapper in Runner
  • Advanced usages
    • Set different hyperparamters for different types of parameters
    • Set different hyperparamters for different model modules
    • Customize optimizer construction policies
    • Adjust hyperparameters during training
• Parameter Scheduler
  • Usage
    • Use a single LRScheduler
    • Combine multiple LRSchedulers (e.g. learning rate warm-up)
  • How to adjust other hyperparameters
    • Momentum
    • Generic parameter scheduler
• Hook
  • Built-in Hooks
    • LoggerHook
    • CheckpointHook
    • ParamSchedulerHook
    • IterTimerHook
    • DistSamplerSeedHook
    • RuntimeInfoHook
    • EMAHook
    • EmptyCacheHook
    • SyncBuffersHook
  • Customize Your Hooks

Advanced tutorials

• Registry
• Config
• BaseDataset
• Data transform
• Weight initialization
• Visualization
• Abstract Data Element
• Distribution Communication
• Logging
• File IO
• Global manager (ManagerMixin)
• Use modules from other libraries
• Test time augmentation
• Model Complexity Analysis

Design

• Hook
• Runner
• Evaluation
• Visualization
• Logging
• Inference

Migration guide

• Migrate Runner from MMCV to MMEngine
• Migrate Hook from MMCV to MMEngine
• Migrate Model from MMCV to MMEngine
• Migrate parameter scheduler from MMCV to MMEngine
• Migrate Data Transform to OpenMMLab 2.0

API Reference

• mmengine.analysis
  • ActivationAnalyzer
  • FlopAnalyzer
  • mmengine.analysis.activation_count
  • mmengine.analysis.flop_count
  • mmengine.analysis.parameter_count
  • mmengine.analysis.parameter_count_table
  • mmengine.analysis.get_model_complexity_info
• mmengine.registry
  • Registry
  • DefaultScope
  • mmengine.registry.build_from_cfg
  • mmengine.registry.build_model_from_cfg
  • mmengine.registry.build_runner_from_cfg
  • mmengine.registry.build_scheduler_from_cfg
  • mmengine.registry.count_registered_modules
  • mmengine.registry.traverse_registry_tree
  • mmengine.registry.init_default_scope
• mmengine.config
  • Config
  • ConfigDict
  • DictAction
• mmengine.runner
  • Runner
  • Loop
  • Checkpoints
  • AMP
  • Miscellaneous
• mmengine.hooks
  • Hook
  • CheckpointHook
  • EMAHook
  • LoggerHook
  • NaiveVisualizationHook
  • ParamSchedulerHook
  • RuntimeInfoHook
  • DistSamplerSeedHook
  • IterTimerHook
  • SyncBuffersHook
  • EmptyCacheHook
  • ProfilerHook
  • NPUProfilerHook
  • PrepareTTAHook
  • EarlyStoppingHook
• mmengine.model
  • Module
  • Model
  • EMA
  • Model Wrapper
  • Weight Initialization
  • Utils
• mmengine.optim
  • Optimizer
  • Scheduler
• mmengine.evaluator
  • Evaluator
  • Metric
  • Utils
• mmengine.structures
  • BaseDataElement
  • InstanceData
  • LabelData
  • PixelData
• mmengine.dataset
  • Dataset
  • Dataset Wrapper
  • Sampler
  • Utils
• mmengine.infer
  • BaseInferencer
• mmengine.device
  • mmengine.device.get_device
  • mmengine.device.get_max_cuda_memory
  • mmengine.device.is_cuda_available
  • mmengine.device.is_npu_available
  • mmengine.device.is_mlu_available
  • mmengine.device.is_mps_available
• mmengine.hub
  • mmengine.hub.get_config
  • mmengine.hub.get_model
• mmengine.logging
  • MMLogger
  • MessageHub
  • HistoryBuffer
  • mmengine.logging.print_log
• mmengine.visualization
  • Visualizer
  • visualization Backend
• mmengine.fileio
  • File Backend
  • File Handler
  • File IO
  • Parse File
• mmengine.dist
  • dist
  • utils
• mmengine.utils
  • Manager
  • Path
  • Package
  • Version
  • Progress Bar
  • Miscellaneous
• mmengine.utils.dl_utils
  • TimeCounter
  • mmengine.utils.dl_utils.collect_env
  • mmengine.utils.dl_utils.load_url
  • mmengine.utils.dl_utils.has_batch_norm
  • mmengine.utils.dl_utils.is_norm
  • mmengine.utils.dl_utils.mmcv_full_available
  • mmengine.utils.dl_utils.tensor2imgs
  • mmengine.utils.dl_utils.TORCH_VERSION
  • mmengine.utils.dl_utils.set_multi_processing
  • mmengine.utils.dl_utils.torch_meshgrid
  • mmengine.utils.dl_utils.is_jit_tracing

Notes

• Changelog of v0.x
  • v0.7.4 (06/03/2023)
  • v0.7.3 (04/28/2023)
  • v0.7.2 (04/06/2023)
  • v0.7.1 (04/03/2023)
  • v0.7.0 (03/16/2023)
  • v0.6.0 (02/24/2023)
  • v0.5.0 (01/20/2023)
  • v0.4.0 (12/28/2022)
  • v0.3.2 (11/24/2022)
  • v0.3.1 (11/09/2022)
  • v0.3.0 (11/02/2022)
  • v0.2.0 (10/11/2022)
• Contributing to OpenMMLab
  • Pull Request Workflow
  • Guidance
  • Python Code style
  • PR Specs

Switch Language

• English
• 简体中文

Indices and tables

• Index

• Module Index

• Search Page