Overview

An easy-to-use & supercharged open-source experiment tracker

Aim logs your training runs, enables a beautiful UI to compare them and an API to query them programmatically.

_{Integrates seamlessly with your favorite tools}

Jump to:

• Quick start

• Aim UI

• Using Aim

• Understanding Aim

• Examples

• API reference

If you have questions:

• Open a feature request or report a bug

• Join our slack

• Overview

Quick Start

• Get started with Aim
• Integrate Aim into an existing project
• Migrate from other tools
• Track media and objects
• Next steps

Aim UI

• Overview
• Home Page
• Runs Management
• Explorers
• Table
• Bookmarks
• Tags page
• Reports

Using Aim

• Manage runs
  • Create runs
  • Continue runs
  • Delete runs
  • Upload runs
• Configure runs
  • Reusing Run in your Repo
  • Defining custom location for your Repo
  • Organizing Runs in Experiments
  • Adding tags and params to Run
  • Training Run Reproducibility
  • Capturing Terminal logs
• Query runs and objects
• Query language basics
  • Introduction
  • Searching
  • Searching metrics and images
  • Security restrictions
• Track experiments with aim remote server
  • Overview
  • Prerequisites
  • Server-side setup
  • Client-side setup
  • SSL support
  • Conclusion
• Notify on failed/stuck runs
  • Run progress reporting SDK
• Log messages during training process
  • The logging API
  • Notifying on logged messages
• Logging artifacts with Aim
  • Artifacts logging SDK
  • Storage backends
• Set up the notification service
  • Quick start
  • Configuring notifiers
• Define custom callbacks
  • Callbacks
  • Example
• Host Aim on Kubernetes (K8S)
  • Dockerfile
  • Volume
  • Deployment
  • Service
  • Conclusion
• Run Aim UI on Jupyter Notebook
• Run Aim UI on SageMaker Notebook instance
  • Using Terminal
  • Using Notebook Extension
• Integration guides
  • Pytorch Ignite
  • Pytorch Lightning
  • Hugging Face
  • TF/keras
  • XGBoost
  • Catboost
  • LightGBM

Understanding Aim

• Overview
  • Aim Components
• Data storage - where Aim data is collected
  • Storage structure
  • What is the index container?
  • How data written to/read from the storage?
• Storage indexing - how Aim data is indexed
  • Background
  • How things worked before?
  • Automatic indexing
  • Conclusion
• Concepts
  • Aim Run
  • Aim Repo
  • Run Params
  • Run Sequence
  • Sequence Context
• Running Aim with profiling
  • Why would you need to enable the profiling.
• Storing Metric sequence data for fast access
  • Background
  • What has changed?
  • What to do if the Metric has been logged already?

Examples

• Track and compare GANs with Aim

References

• Aim CLI
  • init
  • version
  • up
  • server
  • runs
  • convert
  • storage
• Aim status watcher CLI
  • start
  • notifiers
• Aim SDK
  • aim.sdk.repo module
  • aim.sdk.run module
  • aim.sdk.objects.image
  • aim.sdk.objects.distribution
  • aim.sdk.objects.audio
  • aim.sdk.objects.text
  • aim.sdk.objects.figure
  • aim.sdk.sequence module
  • aim.sdk.sequences.metric module
  • aim.sdk.sequences.image_sequence module
  • aim.sdk.sequences.distribution_sequence module
  • aim.sdk.sequences.audio_sequence module
  • aim.sdk.sequences.text_sequence module
  • aim.sdk.sequences.figure_sequence module
  • aim.sdk.sequence_collection module
• Aim Storage
  • aim.storage.arrayview module

Community

• Changelog

Indices and tables

• Index

• Module Index