Skip to content

Lack of Computational resource consumption in paper #12

Open

Description

@greasebig
No description provided.

Activity

  1. zibojia commented on Oct 31, 2025

    @zibojia
    Owner

    The final version has these information.

  2. greasebig commented on Nov 3, 2025

    @greasebig
    Author

    Specifically, For the described two-stage training (CogVideoX-5B-I2V + ControlNet) with the given hyperparameters, what are the training duration (days/hours), VRAM usage, and GPU(s) used? Particularly interested in the details related to the Style Transfer task.

    We use CogVideoX-5B-I2V(Yang et al., 2024) as the base
    model and integrate it with ControlNet to leverage the edited
    first frame to guide the editing process. The batch size of
    the editing model’s training is 32, the learning rate is 1e-5,
    and weight decay is 1e-4. We train model for 2 epoch. We
    sample 33 frames of the videos to train with a resolution
    of 336×592 in first stage. Different from the first stage, we
    use higher resolution of 448×768 and batch size of 16 in
    stage two, fintuning with 1 epoch to help model edit high
    resolution
    
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