Train model if my input is 32X32

[Shikhar-Soni]

trafficstars

Thanks for the code. I want to know what parameters in the config file must be changed to train a model to upscale a 32X32 image to 256X256.

[nfer-tazik]

I haven't done it myself yet, but will venture an educated guess based on https://github.com/Janspiry/Image-Super-Resolution-via-Iterative-Refinement/blob/feda9d468fa3e923249551afaa7339574cfb3a8e/config/sample_ddpm_128.json

Change the l_resolution and r_resolution params for both train and val under datasets.

Change image_size to 256

Let me know if that helps point you in the right direction

[Shikhar-Soni]

Thanks for your response. I tried it and while everything seems to work, the model doesn't converge and I wonder if changing the learning rate or similar can help here. Let me know if you have any pointers for this problem.