Skip to content

[bug]: waifu-diffusion version or description is wrong #3160

@przemoc

Description

@przemoc

Is there an existing issue for this?

  • I have searched the existing issues

OS

Windows

GPU

cuda

VRAM

6GB

What version did you experience this issue on?

2.3.4

What happened?

Existing description for waifu-diffusion-1.4:

An SD-1.5 model trained on 680k anime/manga-style images (2.13 GB)

is incorrect, based on

instead of:

You can read there that:

The Waifu Diffusion 1.4 Anime model is a Stable Diffusion v2 model that has been finetuned from Stable Diffusion v2.1 Base.

The data used for finetuning Waifu Diffusion 1.4 Anime was 5,468,025 text-image samples that had been downloaded through an image board that provides high-quality tagging and original sources to the artworks themselves that are uploaded to the site.

But

What is installed by InvokeAI from:

is somehow not WD v1.4 that lives in branch main (since 2022-12-31), but WD v1.3 that lives in branch fp16.

So

What you may want to do is:

  • Give proper name to this model, i.e. waifu-diffusion-1.3.

WD v1.3, as SD v1.5 based, may be regarded by some as easier to work with prompt-wise, so I think it's good to have it regardless of potential WD v1.4 availability, which is SD v2.1 based.

  • Introduce proper waifu-diffusion-1.4 model.

Above observation leads also to following questions:

  • Does existing .yaml has sufficient configurability?
  • Does huggingface downloader in InvokeAI works properly and has sufficient configurability?

Side note

Let's check what is the HEAD branch in WD:

przemoc@NUC11PHKi7C002:~$ mkdir ttt
przemoc@NUC11PHKi7C002:~$ cd ttt
przemoc@NUC11PHKi7C002:~/ttt$ git init .
Initialized empty Git repository in /home/przemoc/ttt/.git/
przemoc@NUC11PHKi7C002:~/ttt$ git remote add origin https://huggingface.co/hakurei/waifu-diffusion
przemoc@NUC11PHKi7C002:~/ttt$ git remote show origin
* remote origin
  Fetch URL: https://huggingface.co/hakurei/waifu-diffusion
  Push  URL: https://huggingface.co/hakurei/waifu-diffusion
  HEAD branch: main
  Remote branches:
    fp16 new (next fetch will store in remotes/origin)
    main new (next fetch will store in remotes/origin)

It is actually main, so I'm not sure why fp16 is used, but I haven't investigated InvokeAI source code. There may be some bug or intentional logic related to favoring fp16.

Screenshots

No response

Additional context

I'm using WSL2, but it shouldn't matter in this case.

PS> wsl --version
WSL version: 1.1.6.0
Kernel version: 5.15.90.1
WSLg version: 1.0.50
MSRDC version: 1.2.3770
Direct3D version: 1.608.2-61064218
DXCore version: 10.0.25131.1002-220531-1700.rs-onecore-base2-hyp
Windows version: 10.0.22621.1413
PS> wsl --list --verbose
  NAME            STATE           VERSION
* Ubuntu-22.04    Running         2
$ lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 22.04.2 LTS
Release:	22.04
Codename:	jammy

I'm using manually installed InvokeAI from v2.3 branch in its own venv, that has been updated today. I didn't have WD installed before.

$ git log --oneline -1
2af511c98aae (HEAD -> v2.3, tag: v2.3.4, origin/v2.3) release 2.3.4
$ pip install -e .[xformers] --use-pep517 --extra-index-url https://download.pytorch.org/whl/cu117
$ invokeai-configure --root "${INVOKEAI_ROOT}"

Contact Details

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions