It is costs large GPU resources.

guoyii / SACNN

Try to reproduce 2020_TMI_SACNN Self-Attention Convolutional Neural Network for Low-Dose CT Denoising with Self-supervised Perceptual Loss Network

MIT License

32 stars 5 forks source link

Open babbu3682 opened 2 years ago

babbu3682 commented 2 years ago

I have a 48GB single GPU and I used the 1 batch size and (64, 64, 3) patch size.

However, OOM shows up, when I train the SACNN.

How can you use the 20 batch size?

Cal_P_att and Cal_D_att function use too much GPU resource...