Metadata-Version: 2.4
Name: neuroviz-keras
Version: 0.1.0
Summary: Real-time neural network weight visualizer for TensorFlow/Keras training
Author: NeuroViz Contributors
License: MIT
Project-URL: Homepage, https://github.com/neuroviz/neuroviz
Project-URL: Issues, https://github.com/neuroviz/neuroviz/issues
Keywords: tensorflow,keras,visualization,neural-network,weights,deep-learning
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Science/Research
Classifier: Intended Audience :: Developers
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Scientific/Engineering :: Visualization
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.8
Description-Content-Type: text/markdown
Requires-Dist: tensorflow>=2.0
Requires-Dist: flask>=2.0
Requires-Dist: flask-socketio>=5.0
Requires-Dist: numpy>=1.19
Dynamic: requires-python

# 🧠 NeuroViz

**Real-time neural network weight visualizer for TensorFlow/Keras training.**

Stop guessing what's happening inside your model. NeuroViz gives you a live web dashboard that shows the actual weight values of every neuron while training — not just loss and accuracy.

## ✨ Features

- **🔴 Live Grayscale Neuron Map** — Neurons glow from black (near-zero) to white (high magnitude), giving you an instant X-ray of your network
- **🗺️ Weight Heatmaps** — Click any layer to see its full weight matrix as a color-coded heatmap
- **📊 Weight Distributions** — Live histograms showing how weight values are distributed per layer
- **📈 Training Metrics** — Loss and accuracy curves updated in real time
- **⏱️ Epoch Timeline** — Scrub through past epochs to see how weights evolved
- **⚡ Zero Config** — Just add one callback, dashboard opens automatically

## 🚀 Installation

```bash
pip install neuroviz
```

## 📖 Usage

```python
from neuroviz import NeuroViz

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

# Just add the callback — that's it!
model.fit(x_train, y_train,
          epochs=20,
          validation_data=(x_val, y_val),
          callbacks=[NeuroViz()])
```

The dashboard will automatically open at `http://localhost:5050`.

### Options

```python
NeuroViz(
    port=5050,           # Web server port (default: 5050)
    open_browser=True,   # Auto-open browser (default: True)
    update_freq='epoch', # Update frequency: 'epoch' or 'batch' (default: 'epoch')
)
```

## 🖼️ What You'll See

A dark-themed, premium dashboard with:

1. **Network Diagram** — Interactive visualization of your model architecture. Each neuron is colored on a grayscale: black = weight near zero, white = high weight magnitude.
2. **Layer Inspector** — Click on any layer to see a heatmap of all its weights.
3. **Distribution Panel** — Histograms of weight values for each layer.
4. **Metrics Chart** — Live loss and accuracy curves.

## 📋 Requirements

- Python >= 3.8
- TensorFlow >= 2.0

## 📄 License

MIT
