What does this repo contains:

In Benchmark:

Benchmark1.ipynb, Benchmark2.ipynb.

Two Jupyter notebooks, contain two benchmark example with different datasets. Dataset information will be list in section [Dataset Using](## Dataset-using).
bm.py

A pytorch script version of benchmark2, allow you to pass arguments to choose different model, batch size, etc. See details in README.md inside Benchmark folder.
remote_image_folder.py

A custom class inherits VisionDataset of PyTorch, used for Benchmark2.ipnb

In doc:

UsingpyTorchwithPelican.md

Tutorial guides you through setting up and using PyTorch with Pelican for efficient data management and processing.
UsingpyTorchwithPelicanandHTCondor.md

An integrated tutorial for using pytorch with Pelican and HTCondor.

In Others:

Datasets:

For ImageNet standard data, train file is under /train.

Under this path, there are 1000 directories named with corresponding classes of these images inside the directories.

Val and Test follow the same rule.

ImageNetMini is a subset for convenient testing. It only have 10 classes. About 1000 images in each class.