Is it possible to create FP8 GGUF?
Recently I've started creating GGUF, but the request that I had were for FP8 merged models, and I noticed that the noscript would turn FP8 to FP16.
I did some search and found that it is the weight that GGUF accepted, but then I saw this PR - https://github.com/ggml-org/llama.cpp/issues/14762 \- and would like to know if anyone was able to make this work or not?
The main issue at this moment, is the size of the GGUF vs the initial model, since it converts to FP16.
The other one, is that I don't know if it is making the model better, due to FP16, or even worst because of the noscript conversion.
https://redd.it/1oshg41
@rStableDiffusion
Recently I've started creating GGUF, but the request that I had were for FP8 merged models, and I noticed that the noscript would turn FP8 to FP16.
I did some search and found that it is the weight that GGUF accepted, but then I saw this PR - https://github.com/ggml-org/llama.cpp/issues/14762 \- and would like to know if anyone was able to make this work or not?
The main issue at this moment, is the size of the GGUF vs the initial model, since it converts to FP16.
The other one, is that I don't know if it is making the model better, due to FP16, or even worst because of the noscript conversion.
https://redd.it/1oshg41
@rStableDiffusion
GitHub
Feature Request: Direct FP8 conversion from convert_hf_to_gguf.py · Issue #14762 · ggml-org/llama.cpp
Prerequisites I am running the latest code. Mention the version if possible as well. I carefully followed the README.md. I searched using keywords relevant to my issue to make sure that I am creati...
QWEN IMAGE EDIT: MULTIPLE ANGLES IN COMFYUI IS MORE EASY
Innovation from the community: Dx8152 created a powerful LoRA model that enables advanced multi-angle camera control for image editing. To make it even more accessible, Lorenzo Mercu (mercu-lore) developed a custom node for ComfyUI that generates camera control prompts using intuitive sliders.
Together, they offer a seamless way to create dynamic perspectives and cinematic compositions — no manual prompt writing needed. Perfect for creators who want precision and ease!
https://preview.redd.it/s35kgwxer80g1.png?width=806&format=png&auto=webp&s=91e2aa1ce264de0ebe866a815931dddcc71b89af
Link for Lora by Dx8152: dx8152/Qwen-Edit-2509-Multiple-angles · Hugging Face
Link for the Custom Node by Mercu-lore: https://github.com/mercu-lore/-Multiple-Angle-Camera-Control.git
https://redd.it/1oskosi
@rStableDiffusion
Innovation from the community: Dx8152 created a powerful LoRA model that enables advanced multi-angle camera control for image editing. To make it even more accessible, Lorenzo Mercu (mercu-lore) developed a custom node for ComfyUI that generates camera control prompts using intuitive sliders.
Together, they offer a seamless way to create dynamic perspectives and cinematic compositions — no manual prompt writing needed. Perfect for creators who want precision and ease!
https://preview.redd.it/s35kgwxer80g1.png?width=806&format=png&auto=webp&s=91e2aa1ce264de0ebe866a815931dddcc71b89af
Link for Lora by Dx8152: dx8152/Qwen-Edit-2509-Multiple-angles · Hugging Face
Link for the Custom Node by Mercu-lore: https://github.com/mercu-lore/-Multiple-Angle-Camera-Control.git
https://redd.it/1oskosi
@rStableDiffusion