Skip to content

Issue with resuming training #31

Description

@cjchristopher

Hi team - a question about the following line of code:

mask_prob = self.args.mask_prob + (.8 - self.args.mask_prob) * (epoch - 1) / 20

Can I ask what function this serves, and why? If resuming training from a checkpoint model where the epoch count exceeds 28, this is greater than 1 and causes assertion fails. I note that if starting training from epoch 0 with no interruption, the mask_prob remains fixed at the initialized value (0.2), but with any interruption the training resumes with a higher mask prob.
Should this just be mask_prob = self.args.mask_prob ?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions