Openjourney-v4
Trained on +124k Midjourney v4 images, by PromptHero
Trained on Stable Diffusion v1.5 using +124000 images, 12400 steps, 4 epochs +32 training hours.
Pss... "mdjrny-v4 style" is not necessary anymore (yay!)
Description
Trained on Stable Diffusion v1.5 using +124000 images, 12400 steps, 4 epochs +32 training hours.
FAQ
Comments (17)
Just a curiosity: in what the new openjourney v4 is different from the old mdjrny-v4 version ? Just license problems ?
the last it was juts a dreambooth, this is a fine tuned model so no need to include mdjrny-v4 style
@PromptHero聽Thanks.
Safetensor, please
why and how?
Apparently OP forgot they already uploaded the safetensor version:
https://huggingface.co/prompthero/openjourney/blob/main/mdjrny-v4.safetensors
@PromptHero聽Because unlike pickled checkpoints they can't contain malware in the binary.
@PromptHero聽is this the safetensor of this model? https://huggingface.co/prompthero/openjourney/blob/main/safety_checker/model.safetensors
Do I need to apply a VAE or another extra file to run this model correctly? I see in Huggingface that there are many files like VAE, JSON files, etc, apart from the cpkt model itself.
Im quite new at using Stable Diffusion :)
Thanks in advance!
Will you train openjourney v5 with data from mj v5
Can you train a model based on DALLE-2?
bro dalle-2 is kinda trash compared to whats coming out of here
Will the dataset be made available? Openjourney doesn't seem very "open" as it is right now.
haha fu, why u need the dataset? make your own dataset by doing the hardwork yourself and not try to grab free labor of others
@foqatitu381聽the whole point of making something "open" is so that the whole community can contribute and collaborate on improving it. Look at all the people who are asking for a v5. Others could join in to keep adding onto it and make a V5 if the dataset were actually open. This model hasn't been updated for over 3 months now, because it's not actually open as the name would imply. It probably shouldn't be called "openjourney" at this point. More like "alternate-journey".
And to your point of "make your own dataset by doing the hardwork yourself and not try to grab free labor of others", what do you think the original author did? You think they made those 124k Midjourney images themselves? LOL. Your argument is ridiculous and you're a blatant hypocrite. You're creating images from the "free labor of others" with every image you make in Stable Diffusion. Even the Stable Diffusion software itself, that you use to make those images is created by the "free labor of others". Did you pay for those models that you used in Stable Diffusion to create those images, or were they provided by the "free labor of others"? None of this would exist without an open community and people contributing their work to eachother.
is this the safetensor of this model? https://huggingface.co/prompthero/openjourney/blob/main/safety_checker/model.safetensors
I'm not having any luck generating images with this model for the past few days here. :( I've created some great images in the past however. Did something change that I'm not aware of?