t5 efficient mini grammar correction

Introduction

The T5-Efficient-MINI model is designed for grammar correction tasks in English. It was trained on a subset of the C4_200M dataset with additional random typos introduced to enhance error correction capabilities. This model does not require prefixes for its specific task.

Architecture

The model is based on the T5-Efficient-MINI architecture, leveraging the transformer framework for text-to-text generation. It supports deployment in various formats, including PyTorch and ONNX, making it flexible for different inference environments.

Training

The model was trained as part of the Full Stack Deep Learning course. It utilized a subset of the C4_200M dataset, with the introduction of typos using the nlpaug library to simulate more realistic grammar correction scenarios.

Guide: Running Locally

1. Setup Environment: Ensure you have Python installed along with libraries like transformers, torch, and onnxruntime if using ONNX.
2. Clone Repository: Pull the model's code and weights from its Hugging Face repository.
3. Load Model: Use Hugging Face's transformers library to load the model and tokenizer.
4. Inference: Run inference on sample sentences to test grammar correction.
5. Hardware Suggestion: For optimal performance, especially during training or large-scale inference, consider using cloud GPUs such as those provided by AWS, Google Cloud, or Azure.

License

The T5-Efficient-MINI model is released under the MIT License, allowing for wide usage and modification in both personal and commercial projects.