It's the exact same model as FLUX.1-dev
I typed "An anime girl sitting on a throne with a smug grin, legs crossed" with the base model of flux-dev
And then I used the same prompt but with UnfilteredAI.safetensors
Base model used is flux schnell and this is trained on 76k uncensored images
Base model used is flux schnell and this is trained on 76k uncensored images
I don't understand, the hashes are the same too
And the following files also have the same hashes than FLUX.1-dev
- text_encoder/model.safetensors
- text_encoder_2/*.safetensors
- tokenizer_2/spiece.model
- transformer/*.safetensors
- vae/diffusion_pytorch_model.safetensors
- ae.safetensors
Also, the python script that you provide gives ugly images, but if I increase num_inference_steps from 4 to 25 they are perfectly fine
What am I doing wrong please ?
I think it is Lora merged with Flux model
I think it is Lora merged with Flux model
Without changing the hash? I admit that I'm ignorant of the magic that makes these models work, but surely a merge changes the file in some capacity!
Without changing the hash? I admit that I'm ignorant of the magic that makes these models work, but surely a merge changes the file in some capacity!
I'm also surprised! A merge should definitely result in some change to the file, including the hash. This seems unusual
@Aarifkhan are u sure base model was flux schnell?
how long did it take to train the 76k images? or how much did it cost im looking to train about 100k images and wondering how much would it cost me
WAIT! Okay so so this is complete bullshit and actually just the exact same model just renamed? Didn't some tool get called out hard recently doing something similar with an LLM text to text model?
Can anyone confirm this is not a fake, with an explanation of identical hashes? NO, yes? It's a BS?
I typed "An anime girl sitting on a throne with a smug grin, legs crossed" with the base model of flux-dev
And then I used the same prompt but with UnfilteredAI.safetensors
WAIT! Okay so so this is complete bullshit and actually just the exact same model just renamed? Didn't some tool get called out hard recently doing something similar with an LLM text to text model?
Can anyone confirm this is not a fake, with an explanation of identical hashes? NO, yes? It's a BS?
dude got caught