M2M100_418M_HAU_EN_REL_FT

Introduction

The M2M100_418M_HAU_EN_REL_FT is a text-to-text generation model that facilitates translations between Hausa and English. It utilizes the m2m_100 architecture and is implemented using the Transformers library in PyTorch.

Architecture

The model is based on the m2m_100 architecture, which is designed for multilingual translation tasks. This architecture supports multiple languages, allowing for direct translation without the need for pivot languages.

Training

The model was fine-tuned on a dataset that includes English and Hausa, leveraging the capabilities of the m2m_100 architecture to enhance language translation tasks. The specific details of the training dataset and parameters are not disclosed in the provided document.

Guide: Running Locally

To run the M2M100_418M_HAU_EN_REL_FT model locally, follow these steps:

1. Install Dependencies: Ensure you have Python and PyTorch installed. Use pip to install the Transformers library.

   pip install transformers
   

2. Download the Model: Access the model from Hugging Face's model hub.

   from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer
   
   model = M2M100ForConditionalGeneration.from_pretrained("masakhane/m2m100_418M_hau_en_rel_ft")
   tokenizer = M2M100Tokenizer.from_pretrained("masakhane/m2m100_418M_hau_en_rel_ft")
   

3. Load and Test: Use the tokenizer and model to perform translations between Hausa and English.

   text = "Translate this text"
   inputs = tokenizer(text, return_tensors="pt")
   outputs = model.generate(**inputs)
   translated_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)
   

4. Run on Cloud GPUs: For faster performance, consider using cloud services such as AWS, GCP, or Azure, which offer GPU instances suitable for machine learning tasks.

License

The documentation does not explicitly mention the licensing terms for the M2M100_418M_HAU_EN_REL_FT model. Users should check the Hugging Face model card for specific licensing information.