You can deploy this model and start earning money today!

TeXlyre/
grammar-t5-small-onnx

Name: grammar-t5-small-onnx
Availability: InStock

onnx

mit

Views

No views yet

Grammar Synthesis

https://huggingface.co/huytd189/grammar-t5-small converted to ONNX weights for compatibility with Transformers.js.

This repository contains the ONNX weights of the grammar-t5-small model trained on the Grammarly Coedit dataset.

Usage (Transformers Python)

Install PyTorch followed by the requirements:

pip install -r requirements.txt

Exporting ONNX

To modify the original model and re-export it using Optimum, the grammar-t5-small model must be cloned as a submodule.

NOTE: Make sure to install Git LFS before initializing submodules:

bash

git lfs install
# git submodule add https://huggingface.co/huytd189/grammar-t5-small model
git submodule update --init --recursive

Next, run the export script to export the ONNX model:

python export_onnx.py

Inference

To run the inference on the original model:

python infer.py

Or to infer using the ONNX model:

python infer_onnx.py

Usage (Transformers.js)

InstallTransformers.js from NPM:

npm i @huggingface/transformers

Next, run the following example:grammar synthesis:

javascript

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

const generator = await pipeline('text2text-generation', 'TeXlyre/grammar-t5-small-onnx');

const inputTexts = [
    "Rewrite and fix: 'I will drank two bottle'",
    "Rewrite and fix: 'She go to the store'",
    "Rewrite and fix: 'He is more taller than me'",
    "Paraphrase to make it coherent: I've expierienced wet and reiny days. But I survive back then",
    "Correct the grammar: 'She don't like appless.'"
];

for (const inputText of inputTexts) {
    const output = await generator(inputText, {
        max_new_tokens: 100,
        num_beams: 4,
        early_stopping: true
    });
    
    console.log(`Input: ${inputText}`);
    console.log(`Output: ${output[0].generated_text}\n`);
}