Myanmargpt

🚀 Myanmar - GPT

Myanmar GPT is a model designed to understand the Myanmar (Burmese) language. It aims to make the Myanmar language accessible in the GPT2 model, trained on a private Myanmar - language dataset created by MinSiThu.

🚀 Quick Start

Myanmar GPT is a model trained on a private Myanmar language dataset made by MinSiThu. The project's goal is to enable the Myanmar language in the GPT2 Model. Fine - tuning the MyanmarGPT model simplifies the process of building a custom Myanmar language model compared to using alternative language models.

The reports on training the MyanmarGPT model are visualized at MyanmarGPT Report. Variants of the Burmese Language - Enabled Models can be found at https://github.com/MinSiThu/MyanmarGPT. There is also a 1.42 - billion - parameter MyanmarGPT - Big model with multi - language support, which can be found here.

Currently, Myanmar GPT has four main variant versions:

• MyanmarGPT
• MyanmarGPT - Big
• MyanmarGPT - Chat
• MyanmarGPTX

📦 Installation

!pip install transformers

💻 Usage Examples

Basic Usage

from transformers import pipeline

generator = pipeline(model="jojo-ai-mst/MyanmarGPT")
outputs = generator("အီတလီ",do_sample=False)

print(outputs)
# [{'generated_text': 'အီတလီနိုင်ငံသည် ဥရောပတိုက်၏ တောင်ဘက်တွင် မြေထဲပင်လယ်ထဲသို့ ထိုးထွက်နေသော ကျွန်းဆွယ်ကြီးတစ်ခုဖြစ်၍ ပုံသဏ္ဌာန်အားဖြင့် မြင်းစီးဖိနပ်နှင့် တူလေသည်။ မြောက်ဘက်မှ တောင်ဘက်အငူစွန်းအထိ မိုင်ပေါင်း ၇၅ဝ ခန့် ရှည်လျား၍၊ ပျမ်းမျှမိုင် ၁ဝဝ မှ ၁၂ဝ ခန့်ကျယ်သည်။ အီတလီနိုင်ငံ၏ အကျယ်အဝန်းမှာ ဆာဒင်းနီးယားကျွန်း၊ စစ္စလီကျွန်းနှင့် အနီးပတ်ဝန်းကျင်ရှိ ကျွန်းကလေးများ အပါအဝင် ၁၁၆,၃၅၀ စတုရန်းမိုင်ရှိသည်။ '}]

Advanced Usage

import torch
from transformers import GPT2Tokenizer, GPT2LMHeadModel

model = GPT2LMHeadModel.from_pretrained("jojo-ai-mst/MyanmarGPT")
tokenizer = GPT2Tokenizer.from_pretrained("jojo-ai-mst/MyanmarGPT")

def generate_text(prompt, max_length=300, temperature=0.8, top_k=50):
    input_ids = tokenizer.encode(prompt, return_tensors="pt").cuda() # remove .cude() if only cpu
    output = model.generate(
        input_ids,
        max_length=max_length,
        temperature=temperature,
        top_k=top_k,
        pad_token_id=tokenizer.eos_token_id,
        do_sample=True
    )
    for result in output:
      generated_text = tokenizer.decode(result, skip_special_tokens=True)
      print(generated_text)

generate_text("အီတလီ ")

📚 Documentation

RoadMap for Burmese Language and Artificial Intelligence

The creator started the MyanmarGPT project, which has had a huge impact on Myanmar. Thus, the project continues as a movement called MyanmarGPT Movement. The MyanmarGPT Movement encourages everyone to initiate AI projects in Myanmar.

📄 License

MyanmarGPT is free to use for everyone. However, any project derived/finetuned from MyanmarGPT, using MyanmarGPT internally, or modifying MyanmarGPT related to MyanmarGPT must mention the citation below in the corresponding project's page.

@software{MyanmarGPT,
  author = {{MinSiThu}},
  title = {MyanmarGPT},
  version={1.1-SweptWood},
  url = {https://huggingface.co/jojo-ai-mst/MyanmarGPT},
  urldate = {2023-12-14},
  date = {2023-12-14},
}

For contact, reach the author via https://www.linkedin.com/in/min-si-thu/