search

zsyOAOA / ResShift

ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting (NeurIPS@2023 Spotlight, TPAMI@2024)
Other
944 stars 50 forks source link

train error #10

Closed lmvgjp closed 1 year ago

lmvgjp commented 1 year ago

hello, i am not able to start training and get this error could you please help me? greetings Screenshot from 2023-08-21 15-39-42

zsyOAOA commented 1 year ago

You can firstly to train on one single GPU. If you can successfully training using one GPU, and then attempt to train with multiple GPU settings.

Additionally, you may need to adjust the micro batch size in the config file according to your GPU number. You would better to read the training code.

lmvgjp commented 1 year ago

thank you so much for your reply!

i can not make it work for micro batch = 1 and one GPU...and my dataset is extremely low

Screenshot from 2023-08-22 09-29-57 Screenshot from 2023-08-22 09-30-02

zsyOAOA commented 1 year ago

It seems that the dataloader failed to read a image. I can't reproduce this error. Please have a debug around Line 113-114 of the file basicsr/data/realesrgan_dataset.py.

lmvgjp commented 1 year ago

the problem was the path of images in the meta info txt was not complete!!! thank you so much, it started the training process now...now i can make my experiments of training on my own dataset

you are very helpful, i appreciate a lot!! greetings