Manu B.N
I have a 4090 GPU. And a batch_size=1 gives OOM out of memory How to train then ?
I made some modifications to gen_crop_data.py to create smaller patches and now I'm able to train on my 4090 GPU
I could not fix it. So I gave up !!
Training code please !!
If you have found it, can you please provide the link ?
parser.add_argument("--model-type", type=str, default="vit_l", help="The type of model to load, in ['vit_h', 'vit_l', 'vit_b']") has no option for a light HQ based on ViT-tiny. So how to get the Light HQ...
I tried & it worked well !! the psnr was better than other networks solely meant for super resolution
Yes but for a single step these parameters have no real effect ! So no improvements. If I set the VAE & UNet to trainable it's not showing any improvements...
Training VAE separately, and then putting it back into the CCSR improved the performance a little bit . I think if I want to use single step, the performance gains...
Hello, any ideas to predict images in batches ?