Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Appropriate GPU memory #16

Open
TLB-MISS opened this issue Nov 6, 2023 · 4 comments
Open

Appropriate GPU memory #16

TLB-MISS opened this issue Nov 6, 2023 · 4 comments

Comments

@TLB-MISS
Copy link

TLB-MISS commented Nov 6, 2023

How much GPU memory is needed for MVDream's soft-shading, batch8, 512 resolution setting (i.e. the paper setting)?
You said A100 GPU in git, but it's not clear because there are two types of A100, 40GB and 80GB.

Thanks

@murlock1000
Copy link

murlock1000 commented Dec 12, 2023

For reference, I was able to train a model without soft-shading for 5000 iterations with an RTX 3080 Ti (12GB).

The settings I had to change were:
batch_size: [8,4] -> [4,4]

@TLB-MISS
Copy link
Author

@murlock1000
Thanks. But.. If you change n_view from 4 to 1, won't the meaning of 'multi view' disappear?

@murlock1000
Copy link

murlock1000 commented Dec 13, 2023

You're right. I've changed it back to 4. This did not affect the memory usage but increased the run time. Interestingly, the 3D model still converged without multi-view support, but it took significantly more iterations to converge to a 3D model and did not provide highly detailed features.

@yanjk3
Copy link

yanjk3 commented Feb 2, 2024

A100-80G is required for running the soft-shading configuration.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants