Multi-GPU deep learning coding framework based on PyTorch.
PyTorch DDP module provides the mean to efficiently train a deep learning model on multiple GPUs, but using this module requires the knowledge of multiprocessing programming which alone has a steep learning curve. This repo provides a coding framework making multiprocessing transparent, thus helping users unleash the computational power of more than 1 GPUs without additional efforts. Besides that, this framework is integrated with a visualization tools, for example tensorboard, to track the training progress.
To install
pip install git+https://github.com/vkola-lab/multi-GPU.git