They are working on a fix for the flux2 PiD models, (flux2_1024_4096_2606) INT8ConvRot quantizations for PiD decode upscale models PiD reformulates the latent-to-pixel decoder as a conditional pixel-space diffusion model, unifying decoding and upsampling into a single generative module. It denoises directly in high-resolution pixel space and produces a super-resolved image in one pass. This repository hosts the released decoder checkpoints, plus the encoder/decoder ("VAE") weights they depend on. BF16 models here <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/PixelDiT/tree/main/diffusion_models">https://huggingface.co/Comfy-Org/PixelDiT/tree/main/diffusion_models</a> Load your INT8 models with this custom node for speedup on RTX 30XX &amp; 20XX series GPU <a target="_blank" rel="ugc" href="https://github.com/BobJohnson24/ComfyUI-INT8-Fast">https://github.com/BobJohnson24/ComfyUI-INT8-Fast</a>

PiD INT8

Works with Zimage/Turbo and Flux1 (or any flux1 vae model)

pidINT8_flux15122048INT8.safetensors

flux1_512_2048_INT8

sdxl_1024_4096_INT8

Files with SHA256: 56b8337b574d00de44019883bc59c3a16645d6ae8ba8080daa99f4f8e149aab1