torch-swan 0.0.11

SWAN is a library for training and deploying state-of-the-art deep learning models in PyTorch.

# SWAN: Switchable Activation Networks

Overview

Swan is fun

Requirements

• Python 3.8+
• PyTorch
• Torchvision
• NumPy

Installation

Clone the repository and install the required dependencies:

git clone <repository-url>
cd SWAN

python -m pip uninstall -y torch-swan torch_swan
python -m pip install -e .

Or install via pip:

pip install torch_swan

import torch_swan as swan

Examples

python main.py mnist --device mps --epochs 5

nohup python main.py vgg16 --epochs 50 --device npu  --threshold 0.6 --resize 224 --batch_size 200  > train.log 2>&1 &

nohup python main.py vgg16 --epochs 50 --device gpu  --threshold 0.6 --resize 224 --batch_size 256 --pretrained False > train.log 2>&1 &

```</file>

<file name=pyproject.toml>[tool.hatch.build.targets.wheel]
packages = ["src/torch_swan"]
include = ["logo.png"]

[tool.hatch.build.targets.sdist]
include = [
  "src/torch_swan",
  "README.md",
  "LICENSE",
  
]
</file>