Feat: add Flux 1 Lite 8B (Freepik) support #474
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Tested this ontop of my flash attention pr, and the results are good 🚀 . cuda (rtx 2070 mobile / 8gb vram)
Good stuff, the compute buffer sizes are, unsurprisingly, the same as for dev/schnell and the speed faster. |
|
With this model, the flash attention pr and vae-tiling, you can do some obscene stuff like q4_k 2048x1024 on my 8gig vram. |
|
Flash Attention is still not supported on Vulkan sadly. I might try merging in ggml-org/llama.cpp#10206 later to see how it goes. |
|
Thank you for your contribution. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Diffusion model weights: https://huggingface.co/Freepik/flux.1-lite-8B-alpha/blob/main/flux.1-lite-8B-alpha.safetensors.
It's basically Flux 1 [Dev] with most double blocks removed. Flux 1 Lite Q4_k is smaller than Flux 1 Dev Q3_k, while delivering better image quality (in my subjective opinion). It's also about 25% faster during image generation.
.\build\bin\Release\sd.exe --diffusion-model ..\ComfyUI\models\unet\flux.1-lite-8B-alpha-q4_k.gguf --vae ..\ComfyUI\models\vae\ae.q8_0.gguf --clip_l ..\ComfyUI\models\clip\clip_l.q8_0.gguf --t5xxl ..\ComfyUI\models\clip\t5xxl_q4_k.gguf -p "a lovely cat holding a sign says 'flux.cpp'" --cfg-scale 1.0 --sampling-method euler -t 24 --vae-tiling --color -v -W 1024 -H 1024