neurenix init

Initialize a new Neurenix project with folder structure and configuration

The init command creates a new Neurenix project with a standard folder structure, configuration files, and an optional dataset.

Usage

neurenix init [options]

Options

Option	Type	Default	Description
--name	string	neurenix-project	Name of the project directory to create
--template	string	basic	Project template (basic or advanced)
--dataset	string	None	Dataset to download (URL or registered name)
--force	flag	false	Force overwrite if directory exists

Templates

Basic Template

The basic template creates a minimal project structure suitable for getting started quickly:

project-name/
├── config.json
├── train.py
├── data/
├── models/
├── configs/
└── logs/

Configuration (config.json):

• MLP model with layers [128, 64, 10]
• ReLU activation
• Adam optimizer with learning rate 0.001
• Batch size 32, 10 epochs
• Auto device selection

Advanced Template

The advanced template creates a comprehensive project structure for production-ready workflows:

project-name/
├── config.json
├── train.py
├── data/
│   ├── train/
│   ├── val/
│   └── test/
├── models/
├── configs/
├── logs/
├── scripts/
├── notebooks/
└── tests/

Configuration (config.json):

• ResNet-18 model with pretrained weights
• Adam optimizer with learning rate 0.0001
• Cosine learning rate scheduler with warmup
• Batch size 64, 20 epochs
• Mixed precision training
• Data augmentation enabled

Examples

Create a basic project

neurenix init --name my-first-project

Neurenix project 'my-first-project' initialized successfully with basic template.

Create an advanced project

neurenix init --name production-model --template advanced

Neurenix project 'production-model' initialized successfully with advanced template.

Initialize with a dataset

neurenix init --name image-classifier --template advanced --dataset cifar10

Downloading dataset from cifar10...
Saving dataset to image-classifier/data/dataset.csv...
Dataset saved successfully.
Neurenix project 'image-classifier' initialized successfully with advanced template.

Overwrite existing directory

neurenix init --name existing-project --force

Neurenix project 'existing-project' initialized successfully with basic template.

Generated Files

config.json (Basic)

{
  "model": {
    "type": "mlp",
    "layers": [128, 64, 10],
    "activation": "relu"
  },
  "training": {
    "batch_size": 32,
    "epochs": 10,
    "learning_rate": 0.001,
    "optimizer": "adam"
  },
  "hardware": {
    "device": "auto",
    "precision": "float32"
  }
}

train.py (Basic)

import neurenix
from neurenix.nn import Module
from neurenix.optim import Adam

config = neurenix.load_config("config.json")

train_data, val_data = neurenix.load_dataset("data/dataset.csv", split=0.8)

model = neurenix.create_model(config["model"])

optimizer = Adam(model.parameters(), lr=config["training"]["learning_rate"])
neurenix.train(
    model, 
    train_data, 
    val_data, 
    optimizer=optimizer,
    batch_size=config["training"]["batch_size"],
    epochs=config["training"]["epochs"]
)

neurenix.save_model(model, "models/model.nrx")

Error Handling

Directory already exists

neurenix init --name existing-directory

Error: Directory 'existing-directory' already exists. Use --force to overwrite.

Dataset download failure

neurenix init --name project --dataset invalid-url

Warning: Failed to download dataset: [error details]
Neurenix project 'project' initialized successfully with basic template.

Next Steps

After initializing your project:

1. Navigate to the project directory:

   cd my-first-project
   

2. Customize the configuration in config.json

3. Add your dataset to the data/ directory

4. Run the training script:

   neurenix run train.py
   

See Also

• Run command - Train models
• Dataset command - Manage datasets