Instructions to use Comfy-Org/PixelDiT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusion Single File
How to use Comfy-Org/PixelDiT with Diffusion Single File:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
What is the inference step count for the non‑4‑step models?
#8
by makisekurisu-jp - opened
pid_1.5_flux1_1024_to_4096_bf16.safetensors
pid_1.5_flux2_1024_to_4096_bf16.safetensors
pid_1.5_qwenimage_1024_to_4096_bf16.safetensors
diffenrent color change i think
Looking at the original code, closest ComfyUI equivalent for undistilled model would be:
- dpmpp_2m/simple
- 25 steps
- shift 8
- CFG 5.0
Kijai is back
Looking at the original code, closest ComfyUI equivalent for undistilled model would be:
- dpmpp_2m/simple
- 25 steps
- shift 8
- CFG 5.0
yo, i get this error when i use a CFG above 1.0 in the upscale sampler:
ValueError: PidNet requires lq_latent — attach via PiDConditioning
when i go back to cfg 1.0 the error is gone, and PiD works as expected.
~~
edith says comfyui implementation appears to be buggy for now: