[Issue]: SDXL 1.0 Producing Black Images #1858
Comments
|
I'm using an Nvidia RTX 3070 |
|
At least try running with |
Ah yeah, I should have thought of that, that's the one thing I hadn't tried, I'll try that now and update you. |
|
So I tried --safe and got the same result |
|
I get the same result, both using |
|
SD.Next doesn't support running the SDXL models in Original backend mode (yet?). Try the diffusers backend. |
How do you do this? |
|
Read wiki. |
|
Im having same issue, running in diffusers mode. XL was working 2 hours ago with same settings. Now black images. I made no changes, but did not turn off git update. |
|
I have the same issue. |
|
Have you tried no half vae? |
|
Exact same problem |
Yes. Not worked. |
|
I got it working now by doing this to start it: Setting the diffusers in the GUI didn't do anything |
I just tested it. The refiner model gets loaded under these circumstances but causes this error: For base model still lead to this: |
|
I have same problem but different errors, I get errors with torch size, IDK what that is, I am not an expert at all, black squares all the time, tried different combinations in compute settings and nothing worked. This only happens when I run SDXL 1.0 and any models in SDXL. 16:03:51-692756 ERROR Error loading model weights: D:\SD\automatic\models\Stable-diffusion\sd_xl_base_1.0.safetensors
please help idk why this is happening, only with SDXL |
|
Same errors as @SubmitCodes. Using SDXL or any model that derives from SDXL from civitai. |
|
I was having this issue as well after following the steps in the wiki. I fixed it by downloading the model from Huggingface (using "Models > Huggingface > Download model" in the web UI) instead of using the .safetensors file. This only occurred with me when using the base SDXL base .safetensors file from Huggingface. I was able to generate just fine with the .safetensors file from DreamShaper XL on CivitAI. |
For me, both don't work, unfortunately.
I just tried that, no longer black images, but the images are super unusable, like at all. I tried with different sizes, steps and samplers but nothing worked. |
|
@SubmitCodes SDXL does not work with 512x512, try 1024x1024. |
I did
|
|
Had the same issue with automatic1111 in windows. Everything was up to date (manually updated xformers to 0.20) |
|
I got mine working again by changing the settings under Diffuser Settings > VAE upcasting to default, and only check Enable VAE slicing. Apply and don't just click server reset (this didn't work for me), but close and reopen the whole thing. --autolaunch is the only thing I'm starting it up with as well. Everything else is default settings (fresh install). Using sd_xl_base_1.0_0.9vae.safetensors. Edit: After playing around with settings a bit more, for me anyway, it is definitely the VAE upcasting option. If this is changed from default, I get black images. |
|
Unfortunately, this didn't help me either, even though I set up and restarted the program. So for now I'll wait until vlad fixes this, until then I'll use the auto1111 version. |
|
Have you tried the vae fix here? |
|
Yes, I did the VAE fix, I still get a black image. :( So I'm a bit tired of it. |
Unfortunately, this did not help. I use SDP instead of xformers, and start with just that many switches: |
|
black images are always platform specifc, there is no single magic bullet or otherwise it would have been implemented already. there are many users here reporting similar issues, but no info on what systems its happening:
|
|
I had the same issue, with a fresh install, according to https://github.com/vladmandic/automatic/wiki/Installation . Following Nyx01 steps as listed above resolved the issue. Set VAE upcasting to default and close down and start up again, using the restart sever option does not work. |
|
Maybe related |
|
The only thing that worked for me was adding --backend diffusers to the bash prompt. Changing it in the settings doesn't seem to make a difference. The --safe option made little difference as well as the no half settings. |
|
Why is everyone ignoring the ask I had two days ago? |
|
Sorry Ubuntu Mate 22.04.3 Float16, but changing the settings to disable half made little difference with the all black |
|
@kabloink your issue is not same, that's all covered in wiki:
and all that is covered in wiki. everyone else, i have no idea how much of this is applicable to your systems - purely as this thread has turned into noise without much relevant information. so again, please post exactly what's required so i can try to help. |
Ok, I deleted the whole configuration file a couple of days ago and I haven't reconfigured the system yet. What I know off the top of my head right now:
|
You're trying to load SDXL using wrong backend (you have not even enabled diffusers) and that does not work, thats not related to VAE black images at all. Read SDXL wiki. |
|
OS: Windows 10 Not sure if I should open a separate issue. I'm consistently hitting "DefaultCPUAllocator: not enough memory" whilst attempting to load SDXL models; diffusers with regular SD 1.5 models works, and I've been able to get comfy to work with said memory limitation. Any settings that could ease up on the memory usage? The wiki seems only to elaborate on vram, which is ample in my case. |
Definitely separate. Although you might want to ask community on discord for best low ram settings. |
|
Fresh install same issues. In Automatic1111's I had to add the no half vae -- however here, this did not fix it. Win11x64 Also getting these errors on model load: Calculating model hash: C:\Users\xxxx\Deep\automatic\models\Stable-diffusion\SDXL\sd_xl_base_1.0_0.9vae.safetensors 0… |
|
@bbecausereasonss |
|
i'm closing this issue as original author has not provided updates in several weeks and all recent comments are not related to original issue. if there is an update to original issue, i'll reopen. everyone else, having problems, a) please check wiki first, it really does solve almost everything noted here, b) if it doesn't help, create a new issue - adding to unrelated issues prevents me from helping. |
|
self.vae.to(torch.float32) |
Issue Description
When attempting to generate images with SDXL 1.0 all I get is a black square [EXAMPLE ATTACHED]
Version Platform Description
Windows 10 [64 bit]
Google Chrome
12:37:28-168928 INFO Starting SD.Next
12:37:28-172918 INFO Python 3.10.9 on Windows
12:37:28-226138 INFO Version: a32bf08 Fri Jul 28 12:15:25 2023 +0300
12:37:28-677988 DEBUG Setting environment tuning
12:37:28-679982 DEBUG Torch overrides: cuda=False rocm=False ipex=False diml=False
12:37:28-680980 DEBUG Torch allowed: cuda=True rocm=True ipex=True diml=True
12:37:28-684983 INFO nVidia CUDA toolkit detected
12:37:28-870251 DEBUG Repository update time: Fri Jul 28 10:15:25 2023
12:37:28-871223 DEBUG Previous setup time: Fri Jul 28 12:28:49 2023
12:37:28-873217 INFO Disabled extensions: []
12:37:28-875150 INFO Enabled extensions-builtin: ['a1111-sd-webui-lycoris', 'clip-interrogator-ext', 'LDSR', 'Lora',
'multidiffusion-upscaler-for-automatic1111', 'ScuNET', 'sd-dynamic-thresholding',
'sd-extension-aesthetic-scorer', 'sd-extension-steps-animation', 'sd-extension-system-info',
'sd-webui-agent-scheduler', 'sd-webui-controlnet', 'sd-webui-model-converter', 'seed_travel',
'stable-diffusion-webui-images-browser', 'stable-diffusion-webui-rembg', 'SwinIR']
12:37:28-883121 INFO Enabled extensions: ['a1111-sd-webui-tagcomplete', 'canvas-zoom', 'sd-webui-ar',
'sd-webui-aspect-ratio-helper', 'sd-webui-infinite-image-browsing',
'Stable-Diffusion-Webui-Civitai-Helper', 'ultimate-upscale-for-automatic1111']
12:37:28-888121 DEBUG Latest extensions time: Fri Jul 28 12:25:12 2023
12:37:28-889105 DEBUG Timestamps: version:1690535725 setup:1690543729 extension:1690543512
12:37:28-892108 INFO No changes detected: Quick launch active
12:37:28-894092 INFO Verifying requirements
12:37:28-919064 INFO Disabled extensions: []
12:37:28-921059 INFO Enabled extensions-builtin: ['a1111-sd-webui-lycoris', 'clip-interrogator-ext', 'LDSR', 'Lora',
'multidiffusion-upscaler-for-automatic1111', 'ScuNET', 'sd-dynamic-thresholding',
'sd-extension-aesthetic-scorer', 'sd-extension-steps-animation', 'sd-extension-system-info',
'sd-webui-agent-scheduler', 'sd-webui-controlnet', 'sd-webui-model-converter', 'seed_travel',
'stable-diffusion-webui-images-browser', 'stable-diffusion-webui-rembg', 'SwinIR']
12:37:28-930036 INFO Enabled extensions: ['a1111-sd-webui-tagcomplete', 'canvas-zoom', 'sd-webui-ar',
'sd-webui-aspect-ratio-helper', 'sd-webui-infinite-image-browsing',
'Stable-Diffusion-Webui-Civitai-Helper', 'ultimate-upscale-for-automatic1111']
Relevant log output
12:37:58-961923 DEBUG Model weights loaded: {'ram': {'used': 3.92, 'total': 15.94}, 'gpu': {'used': 1.08, 'total': 8.0}, 'retries': 0, 'oom': 0} 12:37:59-702036 DEBUG Model weights moved: {'ram': {'used': 1.94, 'total': 15.94}, 'gpu': {'used': 3.11, 'total': 8.0}, 'retries': 0, 'oom': 0} 12:37:59-713007 INFO Applying xformers cross attention optimization 12:37:59-766818 INFO Embeddings: loaded=3 skipped=1 12:37:59-777787 INFO Model loaded in 5.5s (load=1.0s config=0.1s create=2.2s apply=0.5s vae=0.9s move=0.7s) 12:38:00-046995 DEBUG gc: collected=423 device=cuda {'ram': {'used': 1.97, 'total': 15.94}, 'gpu': {'used': 3.11, 'total': 8.0}, 'retries': 0, 'oom': 0} 12:38:00-049974 INFO Model load finished: {'ram': {'used': 1.97, 'total': 15.94}, 'gpu': {'used': 3.11, 'total': 8.0}, 'retries': 0, 'oom': 0} cached=0 12:38:00-354070 DEBUG gc: collected=124 device=cuda {'ram': {'used': 1.13, 'total': 15.94}, 'gpu': {'used': 3.11, 'total': 8.0}, 'retries': 0, 'oom': 0} 12:38:00-357072 INFO Startup time: 31.4s (torch=6.8s gradio=1.0s libraries=2.7s vae=0.2s models=0.1s codeformer=0.1s scripts=6.3s onchange=0.1s ui-txt2img=0.3s ui-img2img=0.2s ui-settings=0.1s ui-extensions=5.4s ui-defaults=0.1s launch=0.3s app-started=0.7s checkpoint=6.9s) 12:38:00-361090 DEBUG Server alive=True Requests=2 memory used: 1.13 total: 15.94 12:40:00-457119 DEBUG Server alive=True Requests=107 memory used: 1.13 total: 15.94 12:40:03-727197 DEBUG Paste prompt: Kitten in space suit Steps: 20, Seed: 322289802, Sampler: Euler a, CFG scale: 6, Size: 1024x1024, Parser: Full parser, Model: SDXL_sd_xl_base_1.0, Model hash: 31e35c80fc, Version: a32bf08, Pipeline: Original, Token merging ratio: 0.5 12:40:18-319292 DEBUG gc: collected=2028 device=cuda {'ram': {'used': 1.13, 'total': 15.94}, 'gpu': {'used': 3.11, 'total': 8.0}, 'retries': 0, 'oom': 0} 12:40:18-322255 DEBUG txt2img: id_task=task(8in5a9tqo05c7b7)|prompt=Kitten in space suit|negative_prompt=|prompt_styles=[]|steps=20|sampler_index=3|latent_index=None|restore_faces =False|tiling=False|n_iter=1|batch_size=1|cfg_scale=6|clip_skip=1|seed=-1.0|subseed=-1.0|subsee d_strength=0|seed_resize_from_h=0|seed_resize_from_w=0||height=1024|width=1024|enable_hr=False| denoising_strength=0.7|hr_scale=2|hr_upscaler=Latent|hr_second_pass_steps=20|hr_resize_x=0|hr_r esize_y=0|image_cfg_scale=6|diffusers_guidance_rescale=0.7|refiner_start=0.8||refiner_prompt=|r efiner_negative=|override_settings_texts=[]args=(0, False, 'MultiDiffusion', False, True, 1024, 1024, 96, 96, 48, 1, 'None', 2, False, 10, 1, 1, 64, False, False, False, False, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, -1.0, False, 1536, 96, True, True, True, False, False, 7, 100, 'Constant', 0, 'Constant', 0, 4, False, 'x264', 'blend', 10, 0, 0, False, True, True, True, 'intermediate', 'animation', <scripts.controlnet_ui.controlnet_ui_group.UiControlNetUnit object at 0x00000234B4008160>, <scripts.controlnet_ui.controlnet_ui_group.UiControlNetUnit object at 0x00000234B204F760>, <scripts.controlnet_ui.controlnet_ui_group.UiControlNetUnit object at 0x00000234B204D090>, <scripts.controlnet_ui.controlnet_ui_group.UiControlNetUnit object at 0x00000234B204FEE0>, False, False, 'positive', 'comma', 0, False, False, '', 7, '', [], 0, '', [], 0, '', [], True, False, False, False, 0, False, None, None, False, None, None, False, None, None, False, None, None, False, 50, False, 4.0, '', 10.0, 'Linear', 3, False, 30.0, True, False, False, 0, 0.0, 'Lanczos', 1, True, 0, 0, 0.001, 75, 0.0, False, True) 12:40:18-347189 DEBUG Script process: ['Tiled Diffusion:0.0s', 'Tiled VAE:0.0s', 'Dynamic Thresholding (CFG Scale Fix):0.0s', 'Steps animation:0.0s', 'Agent Scheduler:0.0s', 'ControlNet:0.0s', 'Aspect Ratio picker:0.0s', 'Aspect Ratio Helper:0.0s'] 12:40:18-350181 DEBUG Script before-process-batch: ['Tiled Diffusion:0.0s', 'Tiled VAE:0.0s', 'Dynamic Thresholding (CFG Scale Fix):0.0s', 'Steps animation:0.0s', 'Agent Scheduler:0.0s', 'ControlNet:0.0s', 'Aspect Ratio picker:0.0s', 'Aspect Ratio Helper:0.0s'] 12:40:18-353172 DEBUG Script process-batch: ['Tiled Diffusion:0.0s', 'Tiled VAE:0.0s', 'Dynamic Thresholding (CFG Scale Fix):0.0s', 'Steps animation:0.0s', 'Agent Scheduler:0.0s', 'ControlNet:0.0s', 'Aspect Ratio picker:0.0s', 'Aspect Ratio Helper:0.0s'] 12:40:18-604499 DEBUG Sampler: Euler a {'uses_ensd': True} 100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:10<00:00, 1.98it/s] 12:40:44-668493 DEBUG Script postprocess-batch: ['Tiled Diffusion:0.0s', 'Tiled VAE:0.0s', 'Dynamic Thresholding (CFG Scale Fix):0.0s', 'Steps animation:0.0s', 'Agent Scheduler:0.0s', 'ControlNet:0.0s', 'Aspect Ratio picker:0.0s', 'Aspect Ratio Helper:0.0s'] 12:40:44-703377 DEBUG Script postprocess-image: ['Tiled Diffusion:0.0s', 'Tiled VAE:0.0s', 'Dynamic Thresholding (CFG Scale Fix):0.0s', 'Steps animation:0.0s', 'Agent Scheduler:0.0s', 'ControlNet:0.0s', 'Aspect Ratio picker:0.0s', 'Aspect Ratio Helper:0.0s'] 12:40:44-707366 DEBUG Saving image: PNG K:\Stable Diffusion\Local\stable-diffusion\Vlad\automatic\outputs/text\2023-07-28\00019-Kitten in space suit.png (1024, 1024) 12:40:45-062821 DEBUG gc: collected=592 device=cuda {'ram': {'used': 3.55, 'total': 15.94}, 'gpu': {'used': 3.41, 'total': 8.0}, 'retries': 0, 'oom': 0} 12:40:45-065814 DEBUG Script postprocess: ['Tiled Diffusion:0.0s', 'Tiled VAE:0.0s', 'Dynamic Thresholding (CFG Scale Fix):0.0s', 'Steps animation:0.0s', 'Agent Scheduler:0.0s', 'ControlNet:0.0s', 'Aspect Ratio picker:0.0s', 'Aspect Ratio Helper:0.0s'] 12:40:45-072823 DEBUG Processed: 1 Memory: {'ram': {'used': 3.54, 'total': 15.94}, 'gpu': {'used': 3.41, 'total': 8.0}, 'retries': 0, 'oom': 0} txt 12:40:45-347089 DEBUG gc: collected=220 device=cuda {'ram': {'used': 3.55, 'total': 15.94}, 'gpu': {'used': 3.16, 'total': 8.0}, 'retries': 0, 'oom': 0}Acknowledgements
The text was updated successfully, but these errors were encountered: