Custom Model

This guide will walk you through the process of implementing a custom model with Krypton ML. We'll create a simple sentiment analysis model as an example.

Step 1: Create the Custom Model

First, let's create a simple custom model. Create a new file called custom_sentiment_model.py in folder app:

from krypton_ml.core.models import KryptonCustomModel
from typing import Dict, Any

class SentimentModel(KryptonCustomModel):
    def __init__(self):
        # Initialize your model here
        # For this example, we're using a very simple rule-based approach
        self.positive_words = set(['good', 'great', 'excellent', 'amazing', 'wonderful'])
        self.negative_words = set(['bad', 'terrible', 'awful', 'horrible', 'poor'])

    def predict(self, input: Dict[str, Any]) -> Dict[str, Any]:
        text = input.get('text', '').lower()
        words = text.split()
        
        positive_count = sum(1 for word in words if word in self.positive_words)
        negative_count = sum(1 for word in words if word in self.negative_words)
        
        if positive_count > negative_count:
            sentiment = 'positive'
        elif negative_count > positive_count:
            sentiment = 'negative'
        else:
            sentiment = 'neutral'
        
        return {"sentiment": sentiment, "confidence": abs(positive_count - negative_count) / len(words)}

Step 2: Configure Krypton ML

Create a configuration file for Krypton ML. Name it krypton_config.yaml:

krypton:
  server:
    host: "0.0.0.0"
    port: 8000
  models:
    - name: custom-sentiment-model
      type: custom
      module_path: ./app
      callable: custom_sentiment_model.SentimentModel
      endpoint: sentiment-analysis
      description: "A simple custom sentiment analysis model"
      tags:
        - custom
        - sentiment-analysis

Make sure to replace ./app with the actual folder path to your custom_sentiment_model.py file.

Step 3: Run Krypton ML Server

Now, start the Krypton ML server with your configuration:

krypton krypton_config.yaml

Step 4: Test the Model

You can now test your custom sentiment analysis model using a simple curl command or any API client:

curl -X POST http://localhost:8000/sentiment-analysis \
     -H "Content-Type: application/json" \
     -d '{"text": "This product is amazing and wonderful!"}'

This should return a JSON response with the sentiment analysis result.

Conclusion

You've successfully implemented and deployed a custom model using Krypton ML. This example demonstrates how to create a custom model that adheres to the KryptonCustomModel interface and how to integrate it into your Krypton ML workflow. You can extend this example by implementing more sophisticated custom models as needed.