-
Notifications
You must be signed in to change notification settings - Fork 81
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Mac Contributor Support Thread #106
Comments
In a previous version of blender/os/addon i managed to make it work once on macos. But it did not last long. After i always get this error when trying to install the dependencies , it aborts it works on windows, but can't get the dependencies to complete on macos. |
Is the error referring to a specific line/lib? |
As i am not a programmer and don't grasp everything i gather here i paste were it stops working on both an INtel and Arm Mac for me: Successfully built transformers During handling of the above exception, another exception occurred: Traceback (most recent call last): During handling of the above exception, another exception occurred: Traceback (most recent call last): |
Try to install Git as described in the installation instructions. |
I also have to point out i have both platforms at hand (windows and mac); i use windows mostly at home. BUT when at Work , i use mainly mac along with a bunch of creators that would really benefit from this working on mac |
it is installed already, i'll try updating it, but i doubt it is the problem since on the intel MBP it's a fresh install |
NotImplementedError: cannot instantiate 'WindowsPath' on your system |
|
I managed to get past the first error NotImplementedError: cannot instantiate 'WindowsPath' on your system i still had to manually install the rest of the depedencies manually with blender pip plugin, wich was tedious but worked... managed to make the generate button download the models for sdxl, but then i get something concerning the accelerate module not working when i try to use it... |
@limnetik Recently, I did some improvements in the dependencies' installation code, maybe it is working better for you now? |
I tested install using Blender 4.2 ... still giving error:
|
@enzyme69 Apparently, Mac disagrees with that version of soundfile lib. I've updated Pallaidium with another one. Try that. |
Thanks for the update, you might need to update Palladium Mac, but anyhow I use Master Repo. Dependencies installed, however still bug error:
|
@enzyme69 The important line is the last line: |
I will try resolving via Claude... After all, I also did use Blender Module Manager and install pillow module already.
|
Hello, after installing LTS 4.2.1 and deinstalling dependencies and reinstalling. File "/Applications/Blender.app/Contents/Resources/4.2/python/lib/python3.11/site-packages/torch/cuda/init.py", line 305, in _lazy_init If there could be a button to disable cuda and use cpu; on mac it would probably work then... i don't know about coding it to disable torch with cuda on mac... |
@limnetik The switch should happen automatic, however torch/cuda was used to find the available vram, which doesn't work on Mac. As I don't run Mac, I asked a chatGPT for some code to get the VRAM amount on Mac. Could you check if this both solves the problem and that the value is correct? |
The error seems shorter and precise: Error: Python: Traceback (most recent call last): so here it is: Thank you! |
ok same with a dependencies reinstall: |
Okay thank you. Feeling like I'm stumbling around in the dark here, since I have no Mac to test things on, so I'll have to rely on your feedback. Here one more attempt to get rid of CUDA: https://github.com/tin2tin/Pallaidium/archive/refs/heads/Pallaidium_MacOS.zip |
I cannot activate the plugin with this variant, in opposition to the main wich works; Python: 3.11.7 (main, Jun 26 2024, 17:15:05) [Clang 15.0.0 (clang-1500.1.0.2.5)] |
Okay, this means that chatgpt do not know how to get the vram on a mac. Try this instead: https://github.com/tin2tin/Pallaidium/archive/refs/heads/main.zip Torch apparently detects mps on your computer, and tries to use that. |
It activates but , similar message when trying rendering: Traceback (most recent call last): |
Is there something called mps on macs? If not, I can try to force it to cpu. |
i think mps does not apply since it's arm cpu and Automatic1111 forces cpu only on mac platform; so i think the answer is to force it to be in cpu mode. |
Okay, so now it is forcing cpu usage on mps detected hardware (I hope...): https://github.com/tin2tin/Pallaidium/archive/refs/heads/Pallaidium_MacOS.zip |
hmm loads, but still message on render: |
Is it the same thing with ex. OpenDalle? |
it is with OpenDalle that i get the message and it says the same with Stable Diffusion XL 1.0 |
Let's try another approach. Open the Text Editor in Blender, copy/paste this code into it. Change cuda to mps and/or CPU. Open the Blender system console. Run the script and see if it can run without errors (the dependencies should already be installed with Pallaidium):
The code is from here: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0 If it throws errors on torch and cuda, even though you add mps or cpu, you'll have to investigate what version of torch is useable on mac, and then you can try to install it with the pip add-on, untill you find something that works. |
seems to run fine only set with with cpu it give out errors and stall very long i don't get a resulting image but the procedure render something until the end with mps |
@limnetik That's great now we're making progress. Let's check how it is behaving on the various memory optimizations. First, check the code if it runs as is. If it runs, then try to enable each of the optimize-lines one by one, and note which ones work and which ones fail:
|
TEST 1- 2- 3- 4- 5- (same as 2?) |
Thank you. A black image could mean nsfw content. Could you try with #5 again? |
Try this version: https://github.com/tin2tin/Pallaidium/archive/refs/heads/Pallaidium_MacOS.zip |
A black image is the result |
Any attemps with XL or OpenDalle or else result in error, no images: |
Thanks, looks like every model needs additional hand-holding on MacOS, pretty hard to do, without having some hardware to test it on. What about the audio models? |
hmm: error sndfile module not installed and fails to install with python module manager |
I believe this is the one attempted to be installed on Mac: What did you do before getting that error? What is missing sndfile lib? |
This is the error i get with dependencies installed ok and i try to render audio Traceback (most recent call last): During handling of the above exception, another exception occurred: Traceback (most recent call last): |
Searching for the last part of the stdout, things like these show up: |
@limnetik I've replaced Soundfile(not working good on Mac) with Scripy, so maybe the audio models should work for you now? And maybe OpenDalle could work now? |
@limnetik Was there any improvements in how the latest version run om Mac? |
Unfortunately, last time i tried, i managed to install dependencies, but the same cuda error message appeared aborting the image generation. The only thing working was the small script experiments with mps processing the green astronaut on a horse thing. I am going to use my pc, using a mac for this is not an end in itself for me right now. If anyone else finds a way further, i will keep an eye out! |
Back to this thread, still kind of sad that Palladium is not used much in the macOS. It has potential. If you actually get Mac Mini M4, try the macOS and probably it is worth to test. Minimum 64 GB. What would work outside of Blender:
Ideally streamlining Blender for composition and layout, with prompt ability in VSE and adding and generating audio all fully open source. But your addon relies too much by generating inside Blender. Maybe let external app for AI generation and VSE to layout storyboards and prompts. |
@enzyme69 Since I don't have a Mac, I can't identify where it is choking. Someone with a Mac needs to run Pallaidium on it and tinker with the code until it works. It should be possible to get it working, but not without a Mac to run it on. |
As I do not personally use Mac, I can't offer support for it, but if anyone has problems with it, report the Mac-related bugs here, and let's hope some other Mac users can help out submitting fixes for those problems.
The text was updated successfully, but these errors were encountered: