minsithu/myanmargpt

The very first and largest Burmese GPT in Myanmar

MyanmarGPT - Myanmar Generative Pretrained Transformer

The very first and largest usable Burmese Language GPT in Myanmar

• Free to use and open-source
• Lightweight and Accurate
• Burmese + International Languages (Total 61 Languages)
• And, It is Awesome!

MyanmarGPT is the very first and largest usable Burmese language GPT in Myanmar with strong community contributions. It was created by me, Min Si Thu.

These two models are trained by using private property datasets, manually cleaned by Min Si Thu.

There are two versions of MyanmarGPT at the moment, 2023 December.

• MyanmarGPT - 128 M parameters
• MyanmarGPT-Big - 1.42 B parameters

Extended, released in 2024, January 28.

• MyanmarGPT-Chat - 128 M parameters

Released in 2024, February 23.

• MyanmarGPTX - Faster, Lightweight and Multiplatform

MyanmarGPT

MyanmarGPT is 128 million parameters Burmese Language Model. It is very lightweight and easy to use on all devices.

MyanmarGPT-Big

MyanmarGPT-Big is a 1.42 billion parameters Multi-Language Model. It is an enterprise-level LLM for Burmese Language mainly and other languages. Currently supports 61 Languages.

MyanmarGPT-Chat

Fine-tuned on MyanmarGPT, question answering model for the Burmese language. With the knowledge of "A Brief History of the World"

How to use - Tutorial on Building MyanmarGPT-Chat on local machine

MyanmarGPTX

Fine-tuned on MyanmarGPT-Chat, question answering model for the Burmese language. Faster, lightweight and multiplatform available model.

How to use MyanmarGPT and MyanmarGPT-Big

Install hugging face transformer

pip install transformers

MyanmarGPT

# Using Pipeline

from transformers import pipeline

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

# Using AutoTokenizer and CausalLM

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))

MyanmarGPT-Big

# Using Pipeline

from transformers import pipeline

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

# Using AutoTokenizer and CausalLM

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jojo-ai-mst/MyanmarGPT-Big")
model = AutoModelForCausalLM.from_pretrained("jojo-ai-mst/MyanmarGPT-Big")


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))

About MyanmarGPT interview with DVB

[<iframe width="898" height="505" src="https://www.youtube.com/embed/RujWqJwmrLM" title="Chat GPT (AI) ကို မြန်မာလို သုံးစွဲနိုင်တော့မလား - DVB Youth Voice" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen></iframe>](https://youtu.be/RujWqJwmrLM)

Contributors

• Min Si Thu