-
Notifications
You must be signed in to change notification settings - Fork 109
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
will you release the code for SSC? #91
Comments
Hi, thanks for your interest in our work. |
Thanks for your timely reply! Stay following! |
Thanks for your great work! I met up with a problem of too many query voxels that exceed the GPU memory limit and I am eager for your help! As is described by Fig.7 in your paper AdaPoinTr, I tried to add your method as a plug-in block to an existing SSC method called VoxFormer.
the forward method of VoxformerHead is as follows:
But I met up with the problem of too large memory requirement. In my case, the scene is voxelized into 3D volumes of [128, 128, 16]=262144, the number of which is almost two times as many as the number of voxels in your [60, 36, 60] case. the pre-defined query is [1, 262144, 384]. the coordinate is [1, 262144, 3]. the program tried to allocate 256G and led to "RuntimeError: CUDA out of memory".
If i set 'denoise_length' to 0, the error occurs at line 249, which tries to define a mask with shape [262144, 262144]. |
Thanks for your great work! I noted that you implemented this work to the SSC task in your paper ``AdaPoinTr: Diverse Point Cloud Completion with Adaptive Geometry-Aware Transformers''. Will you release the code for SSC?
The text was updated successfully, but these errors were encountered: