Skip to content

Questions about training Q-UNet Simple and Q-Dense #2

Description

@pratch

Hello, I am trying to train the Q-UNet, Q-Dense on MNIST 8x8 by following the configurations in the table 2 from the supplementary. I have a few questions:

  • Is the model for Q-UNet Simple "UNetUndirectedS" ?
  • According to table 2, the model shape is "Channel count 2,4,8, 8 Layers each". Does this translate to depth=8, start_channels=2, qdepth=8 in the code?
  • What configuration (model name and parameters) of Q-Dense is used to generate the digits in Fig. 1? I tried the default values QDenseUndirected 55 8 and the generated images turned out quite mushy.

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

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions