Unet (without clips ans vae) bnb-nf4 version of the merge of Flux.1-schnell and Flux.1-Dev
Comfyui node for unet nf4:
https://github.com/DenkingOfficial/ComfyUI_UNet_bitsandbytes_NF4
I used the tip from comfy.org for merging just the Flux.1-Dev double_blocks (MM-DiT) onto Flux.1-Schnell, giving you a higher quality model that still runs in just 4 steps!
https://blog.comfy.org/august-2024-flux-support-new-frontend-for-loops-and-more/
All in one:
Clips(t5xxlfp16, L) and vae are included, you can put it in comfyui checkpoint folder and use comfyui default WF
Work with Forge 2.0.1
Description
This is the Unet (without clips ans vae) bnb-nf4 version of the merge of Flux.1-schnell and Flux.1-Dev
FAQ
Comments (26)
Your previews look incredible for the UNET only model (flux1SchnellMergedWith_unetBnbNf4.safetensors), but I keep getting error when trying to load it in ComfyUI: ERROR: Could not detect model type
I've tried all the different checkpoint loaders (including CheckpointLoaderNF4), using separate DualCLIPLoader and Load VAE nodes. Keep getting that same error regardless. FYI, I'm able to successfully load/run other Flux Schnell and Dev models and merges without issue, so I'm not sure what the issue is. Could you maybe provide a simple ComfyUI workflow that you've verified works for you?
not tested on comfyui, working good with Forge
@barry9000 I just looked, the comfyui nf4 loader does not support the unet version without clips and vae, only the all in one (clips and vae included) nf4 version, sorry
Open a issue to comfyui github.
https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4/issues/44
@jice Understood, thanks for the update.
@jice I've got Forge installed, I just rarely use it. Loaded your model successfully in Forge and results are amazing. Thanks for sharing the model.
A GGUF version is possible?
if have BnB in the filename its mean (vae, clip-l,t5xxl) is baked and without Bnb means only model Unet, right?
@dongxiat No, BitsAndBytes (bnb) is a library optimized for manipulating model weights with low precision, particularly for deep learning. It allows models to be run using more compact representations (such as 8-bit or 4-bit) without significantly sacrificing model performance. So it has nothing to do with whether the clips are present or not.
I keep getting this: ERROR: Could not detect model type
Updated and everything, got the other flux models working except for this.
@jojaporms Look at description, i give the link for the node
I'm looking for like this checkpoint for flux!!!
Thanks!!
bro please remove login only download i am downloading it from cmd and i need it often
bro can you make nf4 of flux d lite alpha , i tried but it did not work
Hello :D Does anyone know how I can make the loras work?
I got problem ForgeUI "OperationalError: no such column: "size" - should this be a string literal in single-quotes?"
the problem is not the checkpoint: https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/2669#issuecomment-2670889693 A1111 has the same probleme, is a libsqlite 3.49.1 issue, go back to 3.47.2
based on SDXL 1.0? or another?
Based on Flux.1-Schnell and Flux.1-Dev
it works great on 6Gb VRAM and forge and is even faster than SDXL. but I've noticed the images have some sort of artifacts and are slightly less defined than your examples. I've narrowed down the issue to the fact that I believe I have a different T5 text encoder. Do you still have the link to the text encoder you are using?
I can't get it to work. 😪
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.