# 🚀 Running PALADIM Demos from Hugging Face

You can run PALADIM demos **without cloning this repository**! Everything loads directly from Hugging Face Hub.

## 🎯 Quick Options

### Option 1: Python Script (Fastest)

```powershell
# Install packages
pip install transformers peft torch

# Run demo
python demo_from_huggingface.py
```

This loads the model from https://huggingface.co/nickagge/paladim-sentiment and runs inference.

---

### Option 2: Google Colab (No Installation)

1. **Open this notebook**: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/nickagge/paladim/blob/main/PALADIM_Demo.ipynb)

2. **Enable GPU** (optional but faster):
   - Click: `Runtime` → `Change runtime type` → `T4 GPU`

3. **Run all cells** (`Ctrl+F9` or `Runtime` → `Run all`)

That's it! The model loads from Hugging Face automatically.

---

### Option 3: Jupyter Notebook (Local)

```powershell
# Install Jupyter
pip install jupyter transformers peft torch

# Start Jupyter
jupyter notebook

# Open: PALADIM_Demo.ipynb
```

Then run all cells in the notebook.

---

### Option 4: Copy-Paste Code (Any Python Environment)

```python
# Install: pip install transformers peft torch

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from peft import PeftModel

# Load model from Hugging Face
model = AutoModelForSequenceClassification.from_pretrained(
    "distilbert-base-uncased", num_labels=2
)
tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased")
model = PeftModel.from_pretrained(model, "nickagge/paladim-sentiment")

# Test it
text = "This is amazing!"
inputs = tokenizer(text, return_tensors="pt", truncation=True)
with torch.no_grad():
    outputs = model(**inputs)
    prediction = torch.argmax(outputs.logits, dim=-1).item()

print("Sentiment:", "Positive" if prediction == 1 else "Negative")
```

---

## 📦 What Gets Downloaded?

When you run any demo, these files automatically download from Hugging Face:

- **Base model**: `distilbert-base-uncased` (~250 MB)
- **PALADIM adapters**: `nickagge/paladim-sentiment` (~3.5 MB)
- **Tokenizer**: Vocabulary and config files (~1 MB)

**Total**: ~255 MB (cached after first run)

---

## 🌐 Where Does It Run?

| Environment | Setup | GPU | Cost |
|------------|-------|-----|------|
| **Google Colab** | Click notebook link | ✅ Free T4 | $0 |
| **Local PC** | `pip install ...` | Optional | $0 |
| **Jupyter** | `jupyter notebook` | Optional | $0 |
| **Kaggle** | Upload notebook | ✅ Free GPU | $0 |
| **Any Python** | Copy-paste code | Optional | $0 |

---

## 💻 System Requirements

### Minimum (CPU)
- Python 3.8+
- 4GB RAM
- 500 MB disk space
- Works on any laptop

### Recommended (GPU)
- NVIDIA GPU with CUDA
- 8GB VRAM
- Faster inference (~10x)

---

## 🎯 Available Demos

| Demo | Description | File | How to Run |
|------|-------------|------|------------|
| **From HF** | Load model from cloud | `demo_from_huggingface.py` | `python demo_from_huggingface.py` |
| **Basic Usage** | Simple inference | `demo_usage.py` | `python demo_usage.py` |
| **Continual Learning** | Full PALADIM pipeline | `demo_continual_learning.py` | `python demo_continual_learning.py` |
| **Training** | Train from scratch | `train.py` | `python train.py` |
| **Benchmark** | Test performance | `train.py` | `python train.py --run_benchmark` |
| **Jupyter Notebook** | Interactive demo | `PALADIM_Demo.ipynb` | Open in Colab/Jupyter |

---

## 🔧 Troubleshooting

### "No module named 'transformers'"
```powershell
pip install transformers peft torch
```

### "Connection error" or "Model not found"
- Check internet connection
- Verify model exists: https://huggingface.co/nickagge/paladim-sentiment

### "CUDA out of memory"
```python
# Use CPU instead
device = "cpu"
model.to(device)
```

### Model predicts all one class
- This is a demo model trained on limited data
- For production, train on more diverse data
- See `train.py` for training examples

---

## 📚 Next Steps

1. ✅ **Run a demo** - Try `demo_from_huggingface.py` first
2. 📖 **Read the docs** - See `GETTING_STARTED.md` for full guide
3. 🧪 **Train your own** - Use `train.py` with your data
4. 🚀 **Deploy** - Use the model in your applications
5. 📤 **Share** - Upload your trained models to Hugging Face

---

## 🔗 Resources

- **Model on Hugging Face**: https://huggingface.co/nickagge/paladim-sentiment
- **GitHub Repository**: https://github.com/nickagge/paladim
- **Issues/Support**: https://github.com/nickagge/paladim/issues

---

## 💡 Example Use Cases

### 1. Sentiment Analysis API
```python
from flask import Flask, request
# ... load PALADIM model ...

@app.route('/analyze', methods=['POST'])
def analyze():
    text = request.json['text']
    sentiment = predict_sentiment(text)
    return {'sentiment': sentiment}
```

### 2. Batch Processing
```python
texts = ["text1", "text2", "text3", ...]
results = [analyze_sentiment(t) for t in texts]
```

### 3. Real-time Analysis
```python
import streamlit as st
text = st.text_area("Enter text:")
if text:
    sentiment = analyze_sentiment(text)
    st.write(f"Sentiment: {sentiment}")
```

---

**🎉 Happy experimenting with PALADIM!**