Skip to content

Questions about implementations #8

Description

@aoru45

In your code, you set self.use_phys = False;self.use_proj_grad = False;self.use_interprior = False.
I wanna reproduce the results of your paper and I found some misalignment between the code and your paper.
The Vel_Loss is always zero and the result after training is not smooth, I think this may affect the smoothness of predictions.
When setting self.use_phys = True, the loss becomes NAN when training.

I hope you can help me solve these problems and provide a stable code that can train and reproduce the results.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions