After the last update black images after few steps on Mac mini M2 pro with Sonoma 14.5

[Superacf]

How wrote in the subject, almost every workflows not working, I see the preview, but after few steps the image goes black. After many test I've discovered that "Karras" scheduler was causing this problem.

[tsevis]

I have the exact same issue here. Small number of steps is ok but anything above 10 steps isn't working. Started since Friday night. Any help is very appreciated.

[comfyanonymous]

Can you tell me which checkpoint and workflow you are using?

[Superacf]

Happen with all checkpoint

[comfyanonymous]

If you run it with --force-upcast-attention does that fix the issue?

[Superacf]

Here in Italy is 01:00 night, so I've made only one test and it works with that fix, but Comfyui was very very slow to start, but it is a begin to discover a better solution.

[lovelyesz]

I have same issue on mac m3 pro sonoma. but use --force-upcast-attention can fix the issue

[tsevis]

Hi everyone, Using --force-upcast-attention resolved the problem for me. I hope it won't change. :-) Thank you!

[comfyanonymous]

If you have the issue can you post your pytorch version here, you can print it with: python -c 'import torch;print(torch.version.__version__)'

[tsevis]

2.4.0.dev20240518

[comfyanonymous]

If you install stable 2.3 is the issue present?

[tsevis]

I am not sure it's appropriate to do that, as Apple recommends the nightly version. I will probably try but so far, using --force-upcast-attention has resolved the issue. I will follow your advise if I will face the problem again. Thanks for your support.

[Superacf]

If you have the issue can you post your pytorch version here, you can print it with: python -c 'import torch;print(torch.version.__version__)'

Mine is :2.4.0.dev20240413 The nightly version when I installed ComfyUI

[MattaTatTat76]

Noob here. Where do I enter --force-upcast-attention ? I'm on a Macbook. Do I enter it in terminal by itself or in addition to python3 main.py ?

[lquesada]

Possibly related: https://github.com/comfyanonymous/ComfyUI/issues/3523

Will try the suggested fix tomorrow

[tsevis]

Noob here. Where do I enter --force-upcast-attention ? I'm on a Macbook. Do I enter it in terminal by itself or in addition to python3 main.py ?

you are adding --force-upcast-attention after python3 main.py

[BannyLon]

为什么我的是：main.py: error: unrecognized arguments: --force-upcast-attention

[nataliyakarat]

2.4.0.dev20240521 on M3

dpmpp samplers (with and without karras) caused a black image. The --force-upcast-attention option solved it

[nataliyakarat]

为什么我的是：main.py: error: unrecognized arguments: --force-upcast-attention

update comfyui?

[BannyLon]

为什么我的是：main.py: error: unrecognized arguments: --force-upcast-attention

update comfyui?

没有update comfyui，因为更新到最新版后采样器几乎全部失效，而且用ComfyUI Manager 这个插件，只要选择将Preview method:更改为Auto，结果运行生图就会出现以下错误，python意外退出。Python(5424,0x16b8ef000) malloc: Incorrect checksum for freed object 0x33c1a4c00: probably modified after being freed. Corrupt value: 0x385229353f2d3444 Python(5424,0x16b8ef000) malloc: *** set a breakpoint in malloc_error_break to debug zsh: abort python3 main.py habhy@192 ComfyUI % /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/multiprocessing/resource_tracker.py:254: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown warnings.warn('resource_tracker: There appear to be %d '

[udiai]

Same issue, almost all workflows produce black frames. (OS 14.5 on M3 Max) Starting comfyUI with python3 main.py --cpu seems to solve it, but it's very very slow

[Creative-comfyUI]

The problem is not only on Sonoma, it is also on Ventura. It starts in May with some ksampler before being extended to all.

Same issue, almost all workflows produce black frames. (OS 14.5 on M3 Max) Starting comfyUI with python3 main.py --cpu seems to solve it, but it's very very slow

Exactly Karras now seems to be very slow now

If you run it with --force-upcast-attention does that fix the issue?

Read what I wrote in the issue https://github.com/comfyanonymous/ComfyUI/issues/3567 I explain the bug in detail, it will help you to understand what is happening. It seems that the origin of the problem is not the sampler.

[comfyanonymous]

If you update ComfyUI the issue should be solved.

[Creative-comfyUI]

If you update ComfyUI the issue should be solved.

Black Image again. --force-upcast-attention solve the problem yesterday but now the problem is here again after the update

[Creative-comfyUI]

If you update ComfyUI the issue should be solved.

The problem arrive if the batch size is > 1

with --force-upcast-attention everything is ok

[firestar222]

I am on a Mac M2 Max with 64 gig ram. Same issue.

--force-upcast-attention doesn't do anything to help.

Comfy has always been very reliable and stable on my machine, now I can't even generate using SD1.5 models. It's either black or gobbledegook depending on the sampler I choose.

[udiai]

Thank you [comfyanonymous], I'm on MacOS 14.5 with M3Max. --cpu did not generate black frames but was painfully slow. Non workable. I tried to load an archived folder of ComfyUI, before my IPAdapter update, but it didn't work. I tried removing the entire ComfyUI directory and reinstalling fresh ComfyUI, also didn't work. With the help of ChatGPT4o, I uninstalled home-brew and removed all versions of python I could find. Re-installed Home-brew, Python 3.11.9 (also verifying that this is what's active and it's not calling an earlier version when I start Comfy) and latest Pytorch Nighly. Took me 3 days to troubleshoot, but now everything is working as it did before with my full ComfyUI directory.

[firestar222]

Update- I had luck by updating torch to 2.3, now everything seems to be working again?

[tsevis]

If you update ComfyUI the issue should be solved.

Everything seems to work smoothly again. A bit slower though... Thank you!

[Creative-comfyUI]

If you update ComfyUI the issue should be solved.

It is repair thanks. No need --force-upcast-attention

[Creative-comfyUI]

If you update ComfyUI the issue should be solved.

It is repair thanks. No need --force-upcast-attention

The problem starts again. I don't explain why.

[thanhan-a17]

After update, any execution took almost x2 time than before. When steps >= 8, black image appear again. --force-upcast-attention still fix the issue with the cost of speed...

Torch ver: 2.4.0.dev20240526 Comfyui ver: 221116a493 Macbook M3 pro, 18Gb

[wanderingstan]

FWIW, same issues for me.

--force-upcast-attention doesn't seem to help, but --cpu does work but at super slow speed.

Model: Macbook Pro, 14-inch, 2023 Chip: Apple M2 Pro Memory: 32 GB OS: macOS 14.4.1 (23E224) Sonoma Torch: 2.3.0.dev20231215

$ python -c 'import torch;print(torch.version.__version__)'
2.3.0.dev20231215

[treeshark12]

Unexpected value in cast, It seems to be linked to denoise and seed number. Persisted after update but deleting samplers from earlier workflows and replacing them seemed to fix it for now. No SD1.5 models have been working and very few SDXL ones. I updated everything today PyTorch etc.

[treeshark12]

Further to my last comment for me it only happens on smallish (768 or 1024) square images, Ultimate upscale fails every time with black tiles everywhere.

[Creative-comfyUI]

Unexpected value in cast, It seems to be linked to denoise and seed number. Persisted after update but deleting samplers from earlier workflows and replacing them seemed to fix it for now. No SD1.5 models have been working and very few SDXL ones. I updated everything today PyTorch etc.

The first problem you mention seems to occur before the black screen for me too. It looks like Ksampler is sometimes corrupted we have to replace it.
The black screen problem doesn't appear when the batch size is one for me. Have a look at what I wrote here: https://github.com/comfyanonymous/ComfyUI/issues/3567 Your problem seems to confirm that there is something wrong with Ksampler. The problem is with all Ksamplers and in my opinion the problem is not really the black image but the way the information is treated by the Ksampler. If we read all the comments and I look at all the tests I have done, the black image is not always due to the same operation but always in relation to the Ksampler. @comfyanonymous

[treeshark12]

It happens partway through the sampling, at about step 6 or whatever it goes black. It doesn't always report the cast error which is very strange. By changing denoise and steps I can get it to work again and once going it won't happen again on that seed.

[abumolly]

my error is ComfyUI/custom_nodes/ComfyUI-VideoHelperSuite/videohelpersuite/nodes.py:95: RuntimeWarning: invalid value encountered in cast return tensor_to_int(tensor, 8).astype(np.uint8) anyone met the same situation? I used mac M3 and seem only animatediff working flow will cause the error. that's really suffer.

[treeshark12]

It's an odd bug, a 1024 or 768 square will often fail, but I never get a black square at 1280 and above. I'm on Sonoma 14.2 so it isn't only 14.5

[abumolly]

but the workflow i test is the original text to video one, with model of SD1.5, so the latent should be 512*512. acturally, everything was ok before I followed chatgpt 4o and installed tensor and pytorch, reinstalled my comfyui …… sometimes, chatgpt was not creditable, all was because he told me tensor and pytorch plus mps will enhance the efficience of rendering in mac...

[kokotnow]

my error is ComfyUI/custom_nodes/ComfyUI-VideoHelperSuite/videohelpersuite/nodes.py:95: RuntimeWarning: invalid value encountered in cast return tensor_to_int(tensor, 8).astype(np.uint8) anyone met the same situation? I used mac M3 and seem only animatediff working flow will cause the error. that's really suffer.

Yes, Im getting the same thing, same message. Everything was working fine, but like 10 days I did update and it started output either noise or black images only. I reinstalled the whole Comfy but the problem persists.

[abumolly]

my error is ComfyUI/custom_nodes/ComfyUI-VideoHelperSuite/videohelpersuite/nodes.py:95: RuntimeWarning: invalid value encountered in cast return tensor_to_int(tensor, 8).astype(np.uint8) anyone met the same situation? I used mac M3 and seem only animatediff working flow will cause the error. that's really suffer.

Yes, Im getting the same thing, same message. Everything was working fine, but like 10 days I did update and it started output either noise or black images only. I reinstalled the whole Comfy but the problem persists.

my working flow were fine except animatediff ones, after I updated all in comfyui manager, the error changed to the mix lab nodes...custom_nodes/comfyui-mixlab-nodes/nodes/Video.py:502: RuntimeWarning: invalid value encountered in cast img = Image.fromarray(np.clip(img, 0, 255).astype(np.uint8))

[abumolly]

my error is ComfyUI/custom_nodes/ComfyUI-VideoHelperSuite/videohelpersuite/nodes.py:95: RuntimeWarning: invalid value encountered in cast return tensor_to_int(tensor, 8).astype(np.uint8) anyone met the same situation? I used mac M3 and seem only animatediff working flow will cause the error. that's really suffer.

Yes, Im getting the same thing, same message. Everything was working fine, but like 10 days I did update and it started output either noise or black images only. I reinstalled the whole Comfy but the problem persists.

hi, I changed the sampler to dpmpp_2m_sde, and down the denoise value to 0.3, it can generated some images. not sure why but seem problems solved...you can have a try btw: denoise 0.5 will be ok, but 0.8, it turns to black ...

[kokotnow]

Hi, yes, same here, all workflows are fine but the AnimateDiff ones. I tried to lower the denoise value as you suggested but not working, still getting the black images.

[Creative-comfyUI]

@comfyanonymous the black screen problem is caused by the transformers in miniconda on mac. It needs to be updated to version 4.38. This was not a problem until some other components were updated, making the transformers out of date.

[tsevis]

@comfyanonymous the black screen problem is caused by the transformers in miniconda on mac. It needs to be updated to version 4.38. This was not a problem until some other components were updated, making the transformers out of date.

I have updated miniconda, PyTorch and dependencies. Everything runs smoothly so far. Thank you.

[kokotnow]

Hi, thanks for the instructions! I dont know hot find out on which version I am. Does it say somewhere? Anyway, I have updated everything from the Manager but still get black images.

[kokotnow]

@comfyanonymous the black screen problem is caused by the transformers in miniconda on mac. It needs to be updated to version 4.38. This was not a problem until some other components were updated, making the transformers out of date.

I have updated miniconda, PyTorch and dependencies. Everything runs smoothly so far. Thank you.

Is there some command for this in Terminal? I did pip install torch torchvision torchaudio but it says command not found. Any link please how to update miniconda and dependencies? Thank you. Sorry, I'm a complete beginner at this

[tsevis]

@comfyanonymous the black screen problem is caused by the transformers in miniconda on mac. It needs to be updated to version 4.38. This was not a problem until some other components were updated, making the transformers out of date.

I have updated miniconda, PyTorch and dependencies. Everything runs smoothly so far. Thank you.

Is there some command for this in Terminal? I did pip install torch torchvision torchaudio but it says command not found. Any link please how to update miniconda and dependencies? Thank you. Sorry, I'm a complete beginner at this

Hi! Newbie here as well. What I did was updating initially the anaconda from here: https://docs.conda.io/projects/conda/en/latest/user-guide/install/macos.html Then I followed the instructions on ComfyUI's GitHub page: https://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#installing I ran all the Apple instructions from here: https://developer.apple.com/metal/pytorch/ And at the end I ran the dependencies command: https://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#dependencies Everything works smoothly. Unfortunately I don't know how to find the miniconda version. But I know that I have the latest one. Hope some more experienced user could help you more.

[kokotnow]

Ok, thank you for your help, I will try to do all these

[takacsmark]

I can confirm that reinstalling ComfyUI solved the issue (M1 Max, Sonoma 14.5)

The root cause in my case was misalignment of package versions in the venv, I did some manual bug fixing and created a mess. I experienced the black images appear during sampling after a couple of steps.

I reinstalled ComfyUI into a new directory, created a new venv, installed the dependencies and re-installed the custom nodes one by one manually. I used the exact installation steps from github and used pytorch nightly. The clean install works perfectly.