Model configuration using Cerebras Model Zoo

Ways to port your model

Cerebras Model Zoo provides tools to facilitate porting your model to Cerebras ecosystem, from using one of Cerebras Model Zoo models out of the box, to writing your own implementation using Cerebras APIs. This section provides documentation for each of these cases:

Stage	Data Processing and Dataloaders	Define model architecture
Beginner: Getting started with Cerebras Ecosystem	Use data preprocessing from Cerebras Model Zoo	Use model implementation in Cerebras Model Zoo and customize hyperparameters in the params yaml file
Intermediate: Use your own data and hyperparameters	Create your own data preprocessing
Advanced: Define your own model using Cerebras Model Zoo tools		Port your PyTorch or code using run function in Cerebras Model Zoo and Cerebras Model Zoo Supported Operations API
Early access: Define your model using Cerebras PyTorch API		Have more flexibility porting your code with Cerebras PyTorch API

• Cerebras Model Zoo YAML parameters
  • Model parameters
  • Data loader parameters
  • Optimizer parameters
  • Runconfig parameters
• Convert checkpoints and configurations
  • Overview
  • Location
  • Definition
  • Usage
  • Models supported
  • Examples
  • Upgrading Checkpoints and Configs to the current release
  • Frequently Asked Questions

previous

Creating Custom Dataloaders for CS-2

next

Cerebras Model Zoo YAML parameters