joetorch 0.0.4

Useful code for testing new Deep Learning algorithms.

JoeTorch

A Python library containing useful utilities for testing new Deep Learning algorithms.

Installation

Install using pip:

pip install joetorch

Features

• Dataset Utilities

  • MNIST dataset loader with validation split and augmentation options
  • PreloadedDataset class for efficient data handling
  • Support for custom datasets

• Neural Network Components

  • MLP (Multi-Layer Perceptron) module
  • Convolutional blocks (Encoder/Decoder)
  • Auto-encoder architectures

• Training Utilities

  • Learning rate schedulers (Cosine, Step, Flat)
  • Mixed precision training support
  • TensorBoard logging integration
  • Optimized weight decay handling

• Loss Functions

  • MSE reconstruction loss
  • BCE reconstruction loss
  • KL divergence loss
  • Smooth L1 loss
  • Negative cosine similarity

• Feature Analysis

  • Feature correlation analysis
  • Feature standard deviation metrics
  • Representation analysis tools

Example Usage

from joetorch.datasets import MNIST
from joetorch.nn import MNIST_AE
from joetorch.optim import get_optimiser, train

# Load MNIST dataset
train_dataset = MNIST(root='datasets/', split='train', val_ratio=0.1, 
                     augment=True, device='cuda')
val_dataset = MNIST(root='datasets/', split='val', val_ratio=0.1, 
                   device='cuda')

# Create model and optimizer
model = MNIST_AE(out_dim=20, mode='cnn').to('cuda')
optimizer = get_optimiser(model, optim='AdamW')

# Train the model
train(model, train_dataset, val_dataset, optimizer, 
      num_epochs=50, batch_size=256)

Requirements

• Python >= 3.7
• PyTorch >= 1.19.2
• NumPy >= 1.19.2

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

Joe Griffith ([e-mail])