Відео

so i managed to get the program to run, but i cant get it to work. i am using a cyberrealistic model that i use in stable diffusion flawlessly, but it immediately crashes my pc in comfy ui. ive tried with normal vr ram usage too but same issue. i have a radeon 7900xtx

You are a life saver bro, i was fkn pissed off watching tutorials that didn't work, God bless you. I thought I cant install it because I don't have NVIDIA.

top canal, top video, you are awesome.

Good: 1. It's still working. 2. After all is set and done the generation is faster than before. 3. ComfyUI does use the complete GPU RAM. Check your task manager. If not, see a comment down below. You need to find the mentioned python script, open it in your python editor, search for the line and type TRUE there, not true, it's case sensitive. Bad: 1. The video isn't complete enough for any artist who has no experience with such installations. For how to set the PATH in windows you need to Google. It's not difficult but a hurdle. 2. The performance boost you get is ridiculous. I am running a 6800XT with 16GB RAM on a system with 32GB RAM. The rendering of a 1024x512 image is now around 2 minutes and 30 seconds. With the CPU only it's around 12 minutes. That's 6 times faster but for any productive workflow it's still too slow. But better than before. A friend of mine has a system with an older Nvidia 1660 GTX. Maybe I will make a short test there to see the difference. If so, I'll report here. I wonder how fast a 4070 might be in comparison to my 6800.

When I execute the PIP install requirements text, I get an error that says, “sub process exited with error. Getting requirements to build the wheel did not run successfully exit code: 1 File not found error: [winError 2] the system cannot find the file specified [end of output]” I’m running on Windows 11 with a AMD 7900 XT 32 gigs of RAM 12th gen Intel i5 12600 K

if you are having issues with torch install through the requirements text make sure you are using python 3.10 like in the video

Thank you my good brother, it took 3 hours to follow your instructions and it was all worth it when the pictures came out, love from China!

Not working with AMD Radeon RX 7600 XT

Not very helpful. These torch.cuda.is_available() torch.cuda.device_count() torch.cuda.current_device() torch.cuda.device(0) torch.cuda.get_device_name(0) say they detected my gpu(rx 7600 xt) and are gonna use it, but when i ran a neural network, this was the output: Segmentation fault (core dumped). Currently no solution. rocm6.2 + pytorch 2.5.1 + python 3.9.

Your guide succeeded where others have failed. Many thanks - works perfectly.

I am trying this in my ubuntu with Ubuntu 24.04.1 LTS and Python 3.11 but got the errors: got prompt model weight dtype torch.float16, manual cast: None model_type EPS Segmentation fault (core dumped)

ERROR: Could not find a version that satisfies the requirement torch (from versions: none) ERROR: No matching distribution found for torch

Thanks, man.

@luinux-tech excuse me veery fast in rx 580 8gb only install linux mint kernel 5.19.0.50 rocm 5.5 amdgpu install version 5.5 with -- usecase = rocm,hiplibsdk,mlsdk GO github of viebrix repo pytorch-gfx803 with 2 archives pytorch and vision and you proces for all polaris ONLY UPDATE driver amdgpu more stability in etc /apt /souceslist.d /amdgpu list change line only of 5.5.0 to 6.2.3 today OK

I reached the part where I need to download files during the Git setup, but I found that version 2.46.0 is not included. So, can I download the latest one instead？

Hi there, I'm in the process of trying to install Applio the voice RVC, I thinks it's functionally less complex or demanding than Stable Diffusion but eh! I wondered if you could take a look at how to install it as the peeps at Al Hispano are not as clued up about Linux. I have an idea that if I follow this tutorial and adjust it then it should work...even so if you took a look at it you would get a lot of views for figuring it out, it's a very good RVC, also there is F5-TTS which doesn't have a lot of info on linux installation, I'm having a go but any help would be appreciated and a lot of people would watch you videos, I don't think it that difficult in principle but it's when you get errors things get tricky if you're not so familiar, thanks buddy, let me know what you think....any advice would be greatly appreciated. ♥

I had stability issues with SD forge, I just wanted to add that if you want to use ComfyUI the process is near exactly the same. Only three things to keep in mind is that ComfyUI's requirements file is named requirements.txt, that in order to start it you activate python and type main .py, and that you won't need the skip install command to launch it. Everything else follows this tutorial like a charm. (edited to prevent accidental link in comment. Make sure main and .py are one word when you use it to start ComfyUI)

thanks

Ryzen 5900X, RX 6800, Windows 11 - worked flawlessly, thanks.

my "dkms status" shows as: amdgpu/6.8.5-2041575.22.04: added what do I do?

This was the only tutorial I've followed that I managed to get working. Thanks.

The best tutorial bro, Thank you very much for the video

i keep running into this error "ERROR: Could not find a version that satisfies the requirement torch (from versions: none)" i'm using python 3.10.6

can it work with ollama?

I'm Chinese, I don't know English, but after seeing your video, it helped me a lot, although I need translation software to understand, thank you very much, and I have subscribed.

It worked just fine. 6650 XT. used lowvram. Thanks a lot!

Do you think having a pc with AMD gpu for Linux for gaming is a better and there's no driver issues?

on windows??

would love to see the tutorial for the sd forge directml fork i think would work fine with 5000 and polaris gpus even if not as fast

thanks for the video, unfortunately comfyui only sees 1gb vram on my 7900xt :(

Hey brother i just recently switched to atch linux and i have 2 gui kde plasma and hyprland, i have gtx 1650 and cpu i5 f, i got more than 100fps average in cs2 casuals where there are 10 players in one team, got approx 300 fps in Minecraft and this is without any texture packs i guess i would easily get 350 fps. So is everything good

Nice click bait miniature.

When I try to generate images with 16:9 resolutions, I get the error "Warning: Ran out of memory when regular VAE decoding, retrying with tiled VAE decoding." Any idea how to fix this ?

What does SDL or GUI mean? because some mods work differently depending on which version you have and I don't know what's mine

Bro i know this is dumb question but did u made it by using xformers for rocm?

im using ComfyUI, when I run "pip install -r . equirements.txt" I get this error "ERROR: Could not find a version that satisfies the requirement torch (from versions: none) ERROR: No matching distribution found for torch" GPU AMD Radeon RX 6700 XT CPU AMD Ryzen 7 3700X 8-Core Processor 8 Cores

For some reason evertthing works fine but when I try to execute a prompt it crashes with a blue screen. I have A 9700 xtx so I dont know why It crashed everytime

I am gonna try it tomorrow. Thanks man.

need rocm sir?

bro after typing the command py -m venu venu "No module named venu" i am getting this error

What about Windows?

Thanks

awesome!

Why u r using pip install torch instead of pip3.11 install torch?

RADV driver is a gold, but lack of lossless scaling or gamescope native frame generation made me switch back to windows. But whenever gamescope native frame gen comes, i will fully use linux again

it support on rx5700xt?

I tried to build the latest pytorch whl files from source to support my RX 570 GPU and failed so I ended up using your builded files on GitHub and it works, you sir are a genius

I followed your instructions on my newly installed Ubuntu Studio 22.04, and it worked. I had to adjust the power settings of my RX 570 GPU because it unexpectedly shuts down during image generation.

That is a cool new trick!

I'm not sure I ever commented on YT. Thanks for this video ; it isn't promoting artificial obsolescence or anti-poor tactics as it makes "playing" with AI accessible with old gpus and for that i'm very grateful. I 'll upgrade soon but it does give some perspective. Why didn't you use Linux ?