search

JCruan519 / VM-UNet

(ARXIV24) This is the official code repository for "VM-UNet: Vision Mamba UNet for Medical Image Segmentation".
Apache License 2.0
471 stars 22 forks source link

why does the memory explode when I'm training #24

Open hjklearn opened 6 months ago

hjklearn commented 6 months ago

Hello, why does the memory explode when I'm training The details are as follows: Using batchsize=4, image_size=512, using a server with 2080Ti Only when I set batchsize to 1 and image_size=224, it works, why is this? Is there something wrong with my model debugging?

JCruan519 commented 6 months ago

@hjklearn Hello, our experiments were conducted on a single A6000 GPU (48G). Following the settings in our original paper (bs=32, imgsize=256), at least 18GB of VRAM is required to run the experiments. Additionally, I have just tried using the settings bs=8, imgsize=256, and only 8GB of VRAM is required to run it. I suggest adjusting the experimental configuration according to your actual VRAM availability.

hjklearn commented 6 months ago

Thanks for you