Clarifai Local Runner Setup

This repository demonstrates how to create and set up Clarifai local runners for custom model development. It contains a sample text-to-text model that performs string manipulation operations.

Prerequisites

Before setting up this repository, ensure you have:

• Python 3.11+ installed
• pip package manager
• Git
• A Clarifai account

Using Clarifai CLI for Local Testing

The Clarifai CLI provides several commands for local model development:

1. Install the Clarifai Python SDK

Install the latest version of the Clarifai Python SDK:

pip install -U clarifai

2. Login to Clarifai CLI

clarifai login

Fill the details requested during login

3. Setup the Model Codebase

To setup a new codebase:

clarifai model init

Else if you have a codebase on git already:

clarifai model init <codebase-path> --github-repo <user>/<repo> --github-pat <github-token>

This sets up your codebase to develop.

4. Start your Local Runner

Now, you can start your local runner:

clarifai model local-runner

This connects your local model to the Clarifai platform for testing while keeping the model running locally.

Test your Local Runner

Set Clarifai PAT

Refer this guide on how to obtain one from the platform.

Inference using SDK

Once your runner is started, you can test it with a simple prediction:

import os
from clarifai.client import Model
from clarifai.runners.utils import data_types

os.environ["CLARIFAI_PAT"] = "YOUR_PAT_HERE"

model = Model("https://clarifai.com/<user-id>/local-dev-runner-app/models/local-dev-model",
    deployment_id = 'local-dev-deployment', # Only needed for dedicated deployed models
    base_url='https://api.clarifai.com',
 )

    
# Example model prediction from different model methods: 

response = model.predict(prompt="What is the future of AI?", number_of_letters=3)
print(response)

Advanced Usage

Custom Model Development

To create your own model:

1. Modify the MyModel class in 1/model.py
2. Implement your custom logic in the model methods
3. Update the config.yaml with your model details
4. Test locally using the provided commands
5. Upload to Clarifai platform

Adding Dependencies

To add additional Python packages:

1. Add them to requirements.txt
2. Run pip install -r requirements.txt
3. Import and use them in your model code

Troubleshooting

Common Issues

1. Import Errors: Ensure all dependencies are installed with pip install -r requirements.txt
2. Model Not Loading: Check that your model class inherits from ModelClass and implements load_model()
3. Configuration Errors: Verify your config.yaml has valid user_id, app_id, and model_id values
4. "inference_compute_info not found" Error: Ensure your config.yaml includes the complete inference_compute_info section as shown in the configuration example

Getting Help

• Check the Clarifai Documentation
• Use clarifai --help for CLI command help
• Use clarifai model --help for model-specific commands