fine-tuned-model-indonesian

fine-tuned-model-indonesian is an Indonesian Language Model based on Llama-3.1-8B fine-tuned using Unsloth and Hugging Face's TRL library. This model is optimized for general Indonesian conversational tasks, document analysis, and Legal AI Retrieval-Augmented Generation (RAG) pipelines.

---

📌 Model Summary

---

💡 Key Features

• High Fluency in Bahasa Indonesia: Fine-tuned on Indonesian domain-specific corpora to maintain grammatical consistency and natural phrasing.
• RAG & Legal Document Integration: Excellent for contextual question answering, document extraction, and structured information synthesis.
• Resource Efficient: Built upon 4-bit NF4 quantization, enabling fast inference and low VRAM footprint on consumer GPUs.

---

⚙️ Recommended Inference Parameters

To maintain optimal response quality and prevent repetitive loops, use the following generation configuration:

---

🚀 How to Use

1. Using Hugging Face transformers

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "hendrimardani/fine-tuned-model-indonesian"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype=torch.bfloat16,
    load_in_4bit=True
)

messages = [
    {
        "role": "system", 
        "content": "Anda adalah asisten AI berpengalaman yang menjawab pertanyaan dalam Bahasa Indonesia secara ringkas, jelas, dan faktual."
    },
    {
        "role": "user", 
        "content": "Jelaskan tentang aturan waktu kerja dan lembur berdasarkan ketentuan hukum yang berlaku di Indonesia."
    }
]

prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=512,
    do_sample=True,
    temperature=0.3,
    repetition_penalty=1.15,
    eos_token_id=tokenizer.eos_token_id,
    pad_token_id=tokenizer.pad_token_id if tokenizer.pad_token_id is not None else tokenizer.eos_token_id
)

response = tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True)
print(response)