same

@6tZ Yes, in ComfyUI. I have tested versions back to V5, and seemingly from V1 up to V7 (and the GGUF versions, used with gguf loader) all work. from V8-10, all files, when loading as a checkpoint or diffusion model loader produce this error in ComfyUI (V10 used for example):

ERROR UNSUPPORTED DIFFUSION MODEL C:\ComfyUI\models\diffusion_models\fluxedUpFluxNSFW_100BF16.safetensors

!!! Exception during processing !!! ERROR: Could not detect model type of: C:\ComfyUI\models\diffusion_models\fluxedUpFluxNSFW_100BF16.safetensors

I can only assume your method for processing or doing something changed around this time and some necessary tagging information that is supposed to be applied to the completed file is not being done so, but your setup that created the files is supplying that info to the system from elsewhere somehow, so it seems normal on your end.

Hopefully that helps and can be figured out so they can work as V7 GGUF works great and the new version look like they are even better.

6tZ How would i get that? There is no copy other than to copy the text prompt used and that is not the entire workflow for me to import into my local ComfyUI instance. I cannot obtain the .json of the workflow to import it into my local instance that I am aware of.

My instance is a simple base level install of comfyUI, then i create a workflow based on the link your provide, then i load your file. That's it. The CLIP and VAE are correct. the fail comes when it tries to LOAD the model, not even to generation yet. Just trying to load the model to memory and to ComfyUI the file is missing some designators that it is supposed to have and it can;t tell this is a FLUX file it seems. and it fails on loading the model on any file newer than V7 for this.

So if by "workflow of the included images" you mean the provided clip prompt, that is useless since the CLIP loads fine. but NO model data can be loaded from your file. Like, if you read the log file of comfyui/the command prompt window when it is running. you know how when you load models is says things like
"got prompt

model weight dtype torch.bfloat16, manual cast: None

model_type FLUX"

and then loads the model? your files after V7 all say "got prompt

ERROR UNSUPPORTED DIFFUSION MODEL "

In other words, ComfyUI has no clue this is a FLUX file. WHATEVER (cause idk how all that works) is supposed to be set/created/imported to the file at time of exporting it for general use that sets the model type is failing and being left blank instead of being set as FLUX, it would seem.

I would hope that as the creator, you could just spin up a fresh local ComfyUI instance in a new folder, follow the instructions your link provides to make the same workflow and load in your files as they come downloaded from the website as most use cases would be. I would expect you to hit the same error in the stated case.

but if you try and check things in your existing instance (maybe it is even the one that created these instances) there may be things per-resolving that error for you so everything looks time and you don't know why it doesn't work.

@6tZ If it helps, using the DiffusionModelLoaderKJ where i can manually set more options (except the model type, which is what it is missing/needs) i get this error:

!!! Exception during processing !!! 'NoneType' object has no attribute 'model_options'

when it tried to read the model file. so maybe that helps identify where the missing designation info of "FLUX" needs to go.

@6tZ First of all, COOL! Thank You! Did not know that. Very Handy!

Secondly, i did that. and input the V10 download from here. and boom, same error. becasue IT IS NOT a output workflow error, IT IS a model Metadata error. These files (V8-V10 here AND Fux Capacity V6-V6.2) are literally missing the metadata that tells the loader it is a FLUX file (your case), or a SDXL file so the loader accesses it, looks for that metadata tat tells it what kind of model file it is, and sees a blank so it produces the error instead of the output "model FLUX" and continuing to load.

Certainly, I DO believe your version of the file called "FluxedUp_V10.0_BF16.safetensors" seen in your workflow works on your system and that when you run your workflow to "test" and it uses that file it works fine.

But the file that is being uploaded to this site as "fluxedUpFluxNSFW_100BF16.safetensors" is NOT that same file and somewhere along the creation pathway the metadata that says this is a FLUX model is being dropped on/by the final output .safetensors file creation.

I am not trying to be rude; I am trying to help fix the problem because otherwise you will continue to be working on an releasing these updates, but they will only ever work for you. I know I for one would LOVE to see what comes out of the latest versions but sadly can;t isn't the Metadata of the model type in the file is blank/unknown.

Hopefully you will believe me and take close look at your process, especially in any location where metadata is set for the output files because no amount of testing and retesting on my end will make a file that is missing essential metadata from the creator work.

@6tZ I HAVE FIGURED IT OUT!

Thank you for working through the problem, such a such was. Glad to know the problem is not with the file.

Sorry about all that before, once you start down the wrong road it can be hard to find any other. The reason for this error and why it worked for me using previous versions up to a certain point.

For those having the error using the latest files in their local ComfyUI:

UPDATE YOUR COMFYUI INSTANCE TO THE LATEST VERSION!

That was my problem. because the models were created on a more up to date install of ComfyUI, my old version (I was on v13; v18 is current as of writing) it could only read the ones created on the same or lower.

Once i completed the update (BACK UP you ComfyUI folder [minus models folder haha] and run the update as needed and the models should work now.

After running the update, i was able to successfully load the model (the key part that continuously failed with UNKNOWN model type designation) and then generate with the latest models of both fluxed up (v10) and fuxcapacity (v6.2) with success.

@6tZ Tjhanks for yor answer, I work now with ConfyUI, its works ;)