Skip to content

How to train the DiffCast? #19

@BlackRab

Description

@BlackRab

Hello! When I train the DiffCast on my own dataset, the following error occurred:

NotImplementedError: We are sorry that we do not supportthe training process for now because of business limitation

This error was thrown by the predict function of GaussianDiffusion because the compute_loss parameter was set to True during training.

def predict(self, frames_in, compute_loss=False, **kwargs):
T_out = default(kwargs.get('T_out'), 20)
pred, mu, y = self.sample(frames_in=frames_in, T_out=T_out)
if compute_loss:
raise NotImplementedError("We are sorry that we do not support training process for now because of business limitation ")
return pred, None

So how can I train the model correctly? Thanks!

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