MyanmarGPT: Pioneering the Burmese Language in the World of Transformers

In the ever-evolving landscape of natural language processing, MyanmarGPT emerges as a groundbreaking development, marking the advent of the very first and largest usable Burmese language Generative Pretrained Transformer (GPT) in Myanmar. Created by the visionary Min Si Thu, MyanmarGPT stands as a testament to the power of community contributions and the democratization of language models.

Key Features:

1. Free and Open-Source: MyanmarGPT is not just a technological achievement; it's a commitment to accessibility. As the very first usable Burmese language GPT, it is freely available to the public and operates on an open-source basis. This ethos of openness fosters collaboration and innovation within the community.

2. Lightweight and Accurate: Offering a lightweight design without compromising accuracy, MyanmarGPT is tailored for easy integration across various devices. Its efficient architecture ensures seamless usage, making it a versatile tool for a diverse range of applications.

3. Burmese + International Languages: MyanmarGPT transcends linguistic boundaries by supporting a total of 61 languages, making it a truly inclusive language model. While its primary focus is on the Burmese language, its multilingual capabilities reflect a global perspective, catering to diverse linguistic needs.

4. Two Distinct Versions: MyanmarGPT is presented in two versions as of December 2023:

   • MyanmarGPT (128 M parameters): A lightweight Burmese Language Model suitable for various applications.

   • MyanmarGPT-Big (1.42 B parameters): An enterprise-level Multi-Language Model primarily designed for Burmese and extended to support 61 languages.

There are two versions of MyanmarGPT at the moment, 2023 December.
- [MyanmarGPT](https://huggingface.co/jojo-ai-mst/MyanmarGPT) - 128 M parameters
- [MyanmarGPT-Big](https://huggingface.co/jojo-ai-mst/MyanmarGPT-Big) - 1.42 B parameters

How to Use MyanmarGPT and MyanmarGPT-Big:

To harness the power of MyanmarGPT, users can follow these simple steps:

1. Install Hugging Face Transformer:

    pip install transformers
   

2. MyanmarGPT:

    from transformers import pipeline
   
    pipe = pipeline("text-generation", model="jojo-ai-mst/MyanmarGPT")
    outputs = pipe("အီတလီ", do_sample=False)
    print(outputs)
   

3. MyanmarGPT-Big:

    from transformers import pipeline
   
    pipe = pipeline("text-generation", model="jojo-ai-mst/MyanmarGPT-Big")
    outputs = pipe("အီတလီ", do_sample=False)
    print(outputs)
   

4. Advanced Usage Example:

    import torch
    from transformers import AutoTokenizer, AutoModelForCausalLM
   
    tokenizer = AutoTokenizer.from_pretrained("jojo-ai-mst/MyanmarGPT")
    model = AutoModelForCausalLM.from_pretrained("jojo-ai-mst/MyanmarGPT")
   
    input_ids = tokenizer.encode("ချစ်သား", return_tensors='pt')
    output = model.generate(input_ids, max_length=50)
    print(tokenizer.decode(output[0], skip_special_tokens=True))
   

Contributors: Behind the success of MyanmarGPT is a community-driven effort, with Min Si Thu at the helm. This project's success highlights the collaborative nature of technological advancements, where contributions from diverse individuals collectively propel innovation.

Conclusion: MyanmarGPT not only signifies a technological milestone but also symbolizes the empowerment of linguistic diversity in the realm of transformers. As MyanmarGPT continues to evolve, it stands as an inspiration for the global community, showcasing the potential of open-source, multilingual, and community-driven language models. MyanmarGPT is not just a tool; it's a testament to the possibilities that emerge when language technology is made accessible to all.