README.md

---
license: apache-2.0
base_model: scb10x/typhoon2.5-qwen3-4b
tags:
- facebook-scraper
- lora
- typhoon
- fine-tuned
- th
---

# 🤖 Typhoon 2.5 4B - Facebook Scraper LoRA Adapter

This model is fine-tuned using QLoRA technique on data crawled from Facebook posts.

## 👥 Developer & Maintainer Information
* **Fine-Tuner**: **Pansakorn (林金龍) Phothidaen (pphothidaen)** ([LinkedIn](https://www.linkedin.com/in/pansakorn/))
* **Objective**: Academic research, Thai news/post summarization, and data classification.
* **Compliance**: Use of this model must strictly comply with local laws and regulations, privacy guidelines, and international standards.

## 📊 Last Updated Info
* **Timestamp**: `2026-07-17 03:53:31 UTC`
* **Base Model**: [scb10x/typhoon2.5-qwen3-4b](https://huggingface.co/scb10x/typhoon2.5-qwen3-4b)
* **License**: **Apache 2.0 License** (In accordance with the base model license and OpenTyphoon Terms and Conditions)
* **Fine-Tuning Technique**: QLoRA (LoRA Adapter)

## ⚙️ Usage
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

base_model_id = "scb10x/typhoon2.5-qwen3-4b"
lora_model_id = "pphothidaen/typhoon-2-4b-fb-scraper-lora"

model = AutoModelForCausalLM.from_pretrained(base_model_id, device_map="auto")
model = PeftModel.from_pretrained(model, lora_model_id)
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
```