---
base_model: meta-llama/Meta-Llama-3-8B-Instruct
library_name: peft
datasets:
- lordjia/Cantonese_English_Translation
---

# MISHANM/Cantonese_eng_text_generation_Llama3_8B_instruction

This model has undergone meticulous fine-tuning for Cantonese language compatibility. It is equipped to handle question-answering and translation tasks between English and Cantonese. Leveraging sophisticated natural language processing methodologies, it delivers precise and context-sensitive responses, ensuring a comprehensive grasp of Cantonese nuances. Consequently, its outputs are dependable and pertinent across various scenarios.

## Model Details
1. Language: Cantonese
2. Tasks: Question Answering(Cantonese to Cantonese) , Translation (Tibetan to Cantonese)
3. Base Model: meta-llama/Meta-Llama-3-8B-Instruct



# Training Details

The model is trained on approx 109,942 instruction samples.
1. GPUs: 4*AMD Radeon™ PRO V620 
2. Training Time: 61:07:36
  
   


 ## Inference with HuggingFace
 ```python3
 
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load the fine-tuned model and tokenizer
model_path = "MISHANM/Cantonese_eng_text_generation_Llama3_8B_instruction"

model = AutoModelForCausalLM.from_pretrained(model_path,device_map="auto")

tokenizer = AutoTokenizer.from_pretrained(model_path)

# Function to generate text
def generate_text(prompt, max_length=500, temperature=0.9):
    # Format the prompt according to the chat template
    messages = [
        {
            "role": "system",
            "content": "You are a Cantonese language expert and linguist, with same knowledge give response in Cantonese language.",
        },
        {"role": "user", "content": prompt}
    ]

    # Apply the chat template
    formatted_prompt = f"<|system|>{messages[0]['content']}<|user|>{messages[1]['content']}<|assistant|>"

    # Tokenize and generate output
    inputs = tokenizer(formatted_prompt, return_tensors="pt")
    output = model.generate(  
        **inputs, max_new_tokens=max_length, temperature=temperature, do_sample=True
    )
    return tokenizer.decode(output[0], skip_special_tokens=True)

# Example usage
prompt = """佢日日搭的士出入，好似幾百萬未開頭噉。"""
translated_text = generate_text(prompt)
print(translated_text)



```

## Citation Information
```
@misc{MISHANM/Cantonese_eng_text_generation_Llama3_8B_instruction,
  author = {Mishan Maurya},
  title = {Introducing Fine Tuned LLM for Cantonese Language},
  year = {2025},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  
}
```


- PEFT 0.12.0