| title: Text Summarization — SmolLM2 LoRA | |
| emoji: 📝 | |
| colorFrom: blue | |
| colorTo: green | |
| sdk: gradio | |
| sdk_version: 5.0.0 | |
| app_file: app.py | |
| pinned: false | |
| # Text Summarization with SmolLM2-1.7B (LoRA Fine-Tuned) | |
| A Gradio app for article summarization using SmolLM2-1.7B fine-tuned with LoRA on the CNN/DailyMail dataset. | |
| ## How it works | |
| 1. Paste an article into the text box (or click a sample article) | |
| 2. Adjust temperature and max tokens if desired | |
| 3. Click **Summarize** to generate a concise summary | |
| ## Model | |
| - **Base**: [HuggingFaceTB/SmolLM2-1.7B](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B) | |
| - **Fine-tuning**: LoRA (r=8, alpha=16) on CNN/DailyMail | |
| - **Hardware**: Kaggle P100 (16GB GPU) | |
| - **Hub**: [arinbalyan/summarization-lora](https://huggingface.co/arinbalyan/summarization-lora) | |
| ## Training | |
| See the [Kaggle notebook](https://www.kaggle.com/code/arinbalyan/fine-tune-smollm2-for-summarization) for the full training pipeline. | |