Python Script Tuning

Define your own objective function using the Python API for flexible tuning. Choose from four tuner types provided by ZenithTune and write a Python script to run optimization.

Quick Start

• Basic Usage

Tuning Tips

ZenithTune has other useful features.

• Switch to Maximize
• Exclude Trial from Tuning
• Resume and Analyze from Existing Tuning Results
• Set Initial Values for Tuning Parameters
• Receive Metadata from Objective Function
• Tune Environment Variables
• Tune Scripts and Configuration Files

Advanced Tuning

You can expect further performance improvements by using more advanced tuning features.

• Parallel Execution of Tuning Scripts
• CPU Affinity Tuning
• Limiting Training Data Sample Count
• Auto Pruning

Tuning for Distributed Learning Frameworks

With ZenithTune, you can easily tune applications using widely used distributed learning frameworks.

• DeepSpeed
• Megatron-LM
• MMEngine

Kubernetes Job Tuning

ZenithTune provides tuning capabilities for PyTorchJob running on Kubernetes clusters.

• Kubernetes PyTorchJob Tuning
• Job Tuning Example with PyTorchJobTuner