search

sangmin-git / LMC-Memory

Official PyTorch implementation of "Video Prediction Recalling Long-term Motion Context via Memory Alignment Learning" (CVPR 2021 Oral)
Other
86 stars 22 forks source link

questions on training #3

Closed yuyudiandian closed 3 years ago

yuyudiandian commented 3 years ago

Thank you for this good job！But when i run train.py,the following errors happened. image It seems some dimensions didn't match. I try to modify the model.py , image but this errors still happen, can you give me some advice? image

sangmin-git commented 3 years ago

I think you used a different dataset than Moving-MNIST or KTH. I assume that used dataset has img_size 28 which is not divided by 16 in memory encoder. I recommend resizing the input to 32. (results are not guaranteed)