UGATIT-pytorch-colab

The Examples for training UGATIT in Google Colab with using PyTorch

About UGATIT

OVERVIEW

• colab/ - Sample ipynb files for training or mobile model conversion
  • train_example.ipynb - The template of training UGATIT models
• models/ - UGATIT networks (used in Colab files)

USAGE

Training

1. Copy train_example.ipynb to your Google Drive storage.
2. Prepare datasets and put them in the storage too.
3. Open train_example.ipynb and fill in necessary informations (ex. the path to datasets in your storage).
4. Modify others (ex. hyper parameters) if you need.
5. Execute blocks of code in order and start training (trained models are saved in the storage).

Mobile model conversion (pytorch-mobile/coreML)

There are two examples for PyTorch-Mobile(android) and coreML(ios).

1. Copy the sample file for mobile model conversion to your Google Drive storage.
2. Open the file and fill in necessary informations (ex. the path to trained models in your storage).
3. Modify others if necessary and execute blocks of code in order.

Quantization (only coreML)

1. Copy the sample file for quantization to your Google Drive storage.
2. Open the file and fill in necessary informations (ex. the path to the mobile model in your storage).
3. Modify others if necessary and execute blocks of code in order.

NOTICE

PyTorch objects in models/ has been modified from https://github.com/znxlwm/UGATIT-pytorch/blob/master/networks.py.