malayalam U L M Fit Seq2 Seq

Introduction

The malayalam-ULMFit-Seq2Seq model is developed for translating Malayalam text to English. It utilizes the ULMFit architecture and is a work in progress aimed at improving translation accuracy for the Malayalam language.

Architecture

This model is based on the ULMFit (Universal Language Model Fine-tuning) architecture. It is pre-trained on the Malayalam Language Model using fastai, which is known for its efficiency in NLP tasks. The model utilizes SentencePiece for tokenization, with a vocabulary size of 10,000.

Training

The model was trained using the Malayalam Samanantar Dataset, which includes English-Malayalam parallel texts. The dataset is available on Kaggle. However, the model is not yet fine-tuned to achieve state-of-the-art accuracy.

Guide: Running Locally

1. Installation
   Ensure you have Python and pip installed. You can install the necessary package with:

   pip install -Uqq huggingface_hub["fastai"]
   

2. Loading the Model
   Use the following Python code to load and use the model:

   from huggingface_hub import from_pretrained_fastai
   learner = from_pretrained_fastai('hugginglearners/malayalam-ULMFit-Seq2Seq')
   
   original_xtext = 'കേൾക്കുന്ന എല്ലാ കാര്യങ്ങളും എനിക്കു മനസിലായില്ല'
   predicted_text = learner.predict(original_xtext)
   print(f'Predicted text: {predicted_text}')
   

3. Hardware Suggestions
   For efficient processing, using a cloud GPU service such as Google Colab, AWS, or Azure is recommended.

License

The licensing details for this model have not been explicitly stated in the provided information. Users should refer to the source repository or contact the authors for specific licensing terms.