search

StanfordMIMI / DDM2

[ICLR2023] Official repository of DDM2: Self-Supervised Diffusion MRI Denoising with Generative Diffusion Models
125 stars 20 forks source link

Computational Requirements #3

Closed asif-hanif closed 1 year ago

asif-hanif commented 1 year ago

Hi @tiangexiang Thanks for great work.

Could you provide information about computational requirements of the model training from Stage-I through Stage-III i.e. number of GPUs used and approximate total training time?

tiangexiang commented 1 year ago

Hi, thanks for your interest in our work! Most of our experiments (from Stage-I through Stage-III) are able to run on GPUs with 11GB memory. The training of Stage-I may take up to 5~6 hours, while the training of Stage-III may take up to a day, depending on the computing resources you have. Thanks!

asif-hanif commented 1 year ago

Thanks for sharing the information. I will close the issue now.

tiangexiang commented 1 year ago

Hi, I have looked into this further and double-checked the training log, the running time stats are as follows:

Stage I: ~52 s for 1,000 steps, ~90 mins for all 100,000 steps training. Stage II: ~1-2 hours, depending on the size of the 4D MRI. Stage III: ~140 s for 1,000 steps, ~23-24 hours for all 100,000 steps training.

We ran all of the experiments on a single RTX 2080-Ti GPU with 11 GB memory.

asif-hanif commented 1 year ago

Hi @tiangexiang I really appreciate the provision of detailed time stats. Thank you.