Welcome to ABLATOR Documentation!
ABLATOR is a machine learning library that helps you test and compare several model variants to find the one that performs the best; an ablation experiment. Running several experiments in parallel in cumbersome. ABLATOR will take care of the hard work.
Quick Overview
Here is a quick overview of ABLATOR documentations' contents. Usages of ABLATOR are arranged as following sections. Please refer to each section for detailed instructions.
ABLATOR Tutorials
This section will introduce the comprehensive usages of ABLATOR, including the basic usages and advanced usages. Please refer to this section for detailed instructions.
ABLATOR Modules
ABLATOR is composed of several core modules. This is the section introducing how ABLATOR works with these modules.
API Reference
This section is the API reference of ABLATOR. Please refer to this section for detailed usages of ABLATOR modules and functions.
More Examples
ABLATOR is capable of handling various types of deep learning experiments. Please visit this section for more examples of ABLATOR use cases.
Getting Started
Installations
Other installation options are also available.
Quick Start
To get started with ABLATOR quickly, try it out in the demo codes below, where a simple CNN will be trained and evaluated with ABLATOR.
Learn Basics
For more basic usages of ABLATOR, please refer to the Basic Tutorials section below.
How ABLATOR Works
ABLATOR is composed of several core modules. Please refer to this section for detailed usages of each module of ABLATOR and learn how ABLATOR works.
Configuration module
In ABLATOR, the configuration system is used as a framework or structure for defining experiments. With this system, ABLATOR creates and sets up experiments, incorporating the appropriate configurations.
Training module
Other building blocks of ABLATOR are the training module, which launch the experiment that has been configured with the configuration module.
Analysis module
The analysis module has tools that allow you to observe the correlation between the studied hyperparameters and the model's performance.
API Reference
For more detailed information about ABLATOR modules and APIs, please refer to the API Reference.
ABLATOR Community
Visit ABLATOR on Github
ABLATOR is an open-source project. Visit ABLATOR on Github to learn more and feel free to make your contributions.
Meet the developers
ABLATOR is developed and maintained by Deep USC Research Group from University of Southern California.
Follow ABLATOR on social media
Slack Workspace
Discord Community
ABLATOR Official Twitter