Description
Getting black images with my 5090 setup:
--- Starting Img2Img Generation --- Model: full (uncensored), Input Size: 1024x1024 Denoising: 0.8000000000000002, Steps: 50, CFG: 5.0, Shift: 0.0, Seed: 606634206760325 Ensuring pipe on: cuda:0 (Offload NOT enabled) Executing pipeline inference... Starting denoising from step 9/50 (strength: 0.8000000000000002) Using UniPC add_noise with timestep tensor([0], device='cuda:0') 100%|██████████████████████████████████████████████████████████████████████████████████| 41/41 [01:00<00:00, 1.49s/it] C:\AI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\image_processor.py:147: RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") Pipeline inference finished. --- Generation Complete --- Processing output image. Type: <class 'PIL.Image.Image'> pil2tensor: Image mode=RGB, size=(1024, 1024) Numpy array shape=(1024, 1024, 3), dtype=uint8 Final tensor shape=torch.Size([1, 1024, 1024, 3]) Output tensor shape: torch.Size([1, 1024, 1024, 3]) HiDream: Requesting ComfyUI memory cleanup... HiDream: Final VRAM usage: 25742.50 MB (Change: 0.00 MB) Prompt executed in 235.96 seconds