-
Notifications
You must be signed in to change notification settings - Fork 653
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Unable to Load Stable Diffusion 2.0 Custom Model ckpt (768-v-ema.ckpt) #347
Comments
Looks like it's encountering some error while running this script: https://github.com/divamgupta/diffusionbee-stable-diffusion-ui/blob/master/backends/model_converter/convert_model.py Edit: |
I have a similar issue, cannot import a custom model
|
@theorhythm hopefully the next version fixes most imports of 1.4 and 1.5 based models. @eshack94 the importer makes sure that the model is as expected. If you disable that check it will create the converted model but then Diffusion Bee will crash if you try to use it (at least the MPS version, I didn't try the TF version, maybe I should). PS for any code spelunking, the linked script is not the version in DBee 1.5.1, you have to go back a commit. The logic in question is the same, though. |
Same issue when trying to import Anything v3.0. |
For Anything v3.0, I was able to import the fp16 pruned version. Try that one if you haven't. However, I'm having problems importing any model that I have merged in AUTOMATIC1111 -- and the models I'm merging do import just fine on their own, such as Anything v3/hitokomoro and EimisAnime/Spiderverse. The error I get with it is on line 28 in module KeyError: A few other models I've downloaded from Huggingfaces also didn't import for me. I don't remember which ones, however. |
Same issue. I found a workaround by using this tool here: I ran a quick merge with 2 models which themselves imported previously on their own. It imported no problem, whereas prior trying AUTOMATIC1111 merging would result in the same error. However using models that didn't import before resulting 'state_dict' cause the same error to appear using the linked merge script. |
Thanks for the tip, @MDMAchine. As you said, the merge-models utility worked just fine. It serves as a good workaround for those who are comfortable using git and python from the terminal, but probably most DiffusionBee users are not. I'm glad to have it though. 😄 |
Had the same issue when trying, within settings, to run V2.1 of Stable Diffusion as a "custom model". If the merge utility works, that could be a workaround. How would this work with the 2.1 checkpoint? |
@metaphorz Edit: |
Same issue. Trying to load Protogen models. It's throwing this error:
|
@Bangkokian PR #319 fixes this, just waiting for a new version of the app to include it. |
In the meantime , I’ve been using the automatic1111 webui which has many options and will load multiple SD checkpoints |
The mac app is 1.5.1, from what I understand it's still the version with the bug. Do you have any date to publish the fix? |
On the GitHub page, under releases you can find 1.5.2 which includes the updated importer which should import the 1.4/1.5 based models that were failing. |
@CrudeDiatribe I understood. the fix only happens for 1.4/1.5 versions. So it is currently not possible to import the 2.1 model? In model v2.1-768px I am having a similar error from this issue:
More info: Apple M1 - 16 GB |
Correct. |
And this is still the case unfortunately 👎 |
Are you using the latest release from Github? SD2.x support was added per the release notes. |
Error output thrown to UI:
Trying to load the new Stable Diffusion model as a Custom Model.
More info:
Has anyone seen this, and is there a workaround for this? Or is this a known problem that requires code changes to add full support for?
The text was updated successfully, but these errors were encountered: