Model Card for myemoji-gemma-3-270m-it-onnx

This model is a fine-tuned version of google/gemma-3-270m-it, exported to ONNX (Open Neural Network Exchange) format for optimized on-device performance. It is designed to act as an "emoji translator," converting input text into relevant emojis.

Model Details

• Model Name: myemoji-gemma-3-270m-it-onnx

• Base Model: google/gemma-3-270m-it

• Architecture: Gemma 3 (270M parameters)

• Format: ONNX (optimized for web/mobile inference)

• Fine-tuning Task: Text-to-Emoji translation

• Language: English (Input), Emojis (Output)

Intended Use

This model is lightweight (270M parameters) and formatted in ONNX, making it ideal for running directly in the browser (using Transformers.js) or on mobile devices without requiring heavy server-side compute.

Primary Use Case:

• Translating sentences into emojis (e.g., "I am happy" -> "😊").

• Fun/creative text augmentation in web applications.

How to Use (Transformers.js)

You can run this model in a web browser using Transformers.js.

import { pipeline } from '@xenova/transformers';

// Allocate the pipeline
const generator = await pipeline('text-generation', 'manuelaschrittwieser/myemoji-gemma-3-270m-it-onnx');

// Generate emojis
const output = await generator('I am so excited for the party!', {
  max_new_tokens: 20,
});

console.log(output);
// Expected output: 😊🎉🎈

Training Data

This model was fine-tuned using the Gemma 3 270M Instruction Tuned model as a base. It was likely trained on a dataset of text-emoji pairs to learn the mapping between natural language sentiments/objects and their corresponding emoji representations.

Hardware & Optimization

• Parameter Count: 270M (Ultra-compact)

• Export Format: ONNX

• Hardware Compatibility: CPU, WebGPU, NPU (via ONNX Runtime)

License

This model is built upon Google's Gemma model. Users must comply with the Gemma Terms of Use.

• License: Gemma Terms of Use / Apache 2.0 (Check original base model for specifics)