Too much VRAM cost and ckpt file volume compared to the paper

[yryun]

Thanks for great work!

I followed the example provided with cfill dataset. -> https://github.com/sled-group/CycleNet/blob/main/docs/train.md

I found 2 differences with the original works.

1. Training with batch size = 1 cost over 30GB VRAM, which is different in paper (batch size = 4 cost 27.9GB VRAM).

2. My ckpt file (saved in training) is 13.7 GB, which is different in provided ckpt (summer2winter.ckpt = 6.7 GB)

Would you explain how to reproduce the original works?

[beautifulchoi]

seems image size is different. paper said "a single NVIDIA A40 GPU under 256 × 256 with batch size of 4" but that dataset size is 512 x 512