Skip to content

Configuration for ImageNet 512x512 dataset #24

@zheyuanhu01

Description

@zheyuanhu01

Dear Authors,

Thanks for your great work. You mentioned in the ICLR published paper that

We further extend our method to latent space and conducted experiments on ImageNet 512 × 512.
Our ECM-M (498M parameters) achieves FID of 6.88 for 2-step generation using a modest budget
of 12.8M training images, surpassing ADM-G (Dhariwal and Nichol, 2021)(559M parameters, FID
of 7.72 at NFE=250x2) on the ImageNet 512 × 512 dataset. The strong performance on ImageNet
512×512 also indicates that the insights from our scaling analysis generalize beyond the low-resolution
regime.

Could you kindly provide the Configuration for ImageNet 512x512 dataset?

Best regards,
Zheyuan

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