---
language:
- my
- en
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
tags:
- myanmar
- burmese
- llm
- chat
- instruction-following
- conversational
- autoregressive
base_model: MiniMaxAI/MiniMax-M2.7
datasets:
- amkyawdev/myanmar-v3-clean
- amkyawdev/burme-coder-max
- amkyawdev/mm-llm-coder-agent-dataset
- saillab/alpaca-myanmar_burmese-cleaned
---

# 🐉 Myanmar Ghost

**Advanced Myanmar Language Model (LLM)**

Fine-tuned on MiniMax-M2.7 with QLoRA for Myanmar language understanding.

## 💬 Features

- 🗣️ **Myanmar Chat** - Natural conversation in Burmese
- 📝 **Instruction Following** - Follow complex Myanmar instructions  
- 💻 **Code Generation** - Write Myanmar code and documentation
- 🌐 **Translation** - Myanmar ↔ English
- 📖 **Summarization** - Summarize Myanmar text
- ❓ **QA** - Answer questions in Myanmar

## 📊 Training Data

| Dataset | Samples |
|---------|---------|
| myanmar-v3-clean | 877,706 |
| burme-coder-max | 1,000,000 |
| mm-llm-coder-agent | 4,000,020 |
| alpaca-myanmar | 41,601 |

**Total: ~6M instruction samples**

## 🚀 Quick Start

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load model
model_name = "amkyawdev/myanmar-ghost"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    load_in_4bit=True,
    device_map="auto"
)

# Generate
prompt = """### Instruction:
မြန်မာစာမေးပွဲအကြောင်း ရှင်းပါ

### Response:
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## 📋 Requirements

```
torch>=2.0.0
transformers>=4.40.0
bitsandbytes>=0.40.0
peft>=0.4.0
accelerate>=0.20.0
```

## 📜 License

Apache 2.0

## 👤 Author

**Aung Myo Kyaw (amkyawdev)**