trying comfyui

[colertyui]

Hello everyone,
I am very new to anything and everything coding related, only starting to mess arround with generative AI, so I apologize in advance if my question sounds really stupid. I wanted to try and run comfy ui on Ubuntu 24.04.1 (kernel 6.8.0-48-generic), with ROCm in my rx5700 (non xt) which I got working with this tool. I wanted to know if I needed to do anything different from the standard installation to not f-up, Im kinda scared to just mess arround and find out since I already had to reinstall from scratch once due to a mistake in my part. Thanks to everyone in advance, and sorry for the question ;) I hope yall have wonderful days

[daniandtheweb]

I've been using ComfyUI for a while on the RX5700 XT (it's the same card as yours, just overclocked).
In order to setup and run ComfyUI that's what you can do:

1. Open the ComfyUI folder with your terminal.

2. Run the following commands:

   source /opt/rocm_sdk_612/bin/env_rocm.sh

   python -m venv venv

   source venv/bin/activate

   Remember to source the venv after the rocm_sdk otherwise you'll install every package directly in the rocm_sdk directory (this can potentially cause problems).

   pip install --upgrade pip

   The following command installs all the pytorch packages to the virtual environment.

   pip install your_rocm_sdk_builder_directory/packages/whl/torch*

   pip install -r requirements.txt

After the installation is complete you can just run the program each time running:

source /opt/rocm_sdk_612/bin/env_rocm.sh

source venv/bin/activate

Same advice as before, the order is important, source the venv after env_rocm.sh.

python main.py --fp16-unet --fp16-vae --fp16-text-enc --use-pytorch-cross-attention --preview-method taesd --normalvram --listen

That's the launch command I personally use (edit it as you like).

I hope this can be helpful for you.

[colertyui]

heya, did exctly ehat you wrote here, and ended up with an error

Traceback (most recent call last):
File "/media/phlq/HDD/ComfyUI/main.py", line 91, in
import execution
File "/media/phlq/HDD/ComfyUI/execution.py", line 13, in
import nodes
File "/media/phlq/HDD/ComfyUI/nodes.py", line 21, in
import comfy.diffusers_load
File "/media/phlq/HDD/ComfyUI/comfy/diffusers_load.py", line 3, in
import comfy.sd
File "/media/phlq/HDD/ComfyUI/comfy/sd.py", line 5, in
from comfy import model_management
File "/media/phlq/HDD/ComfyUI/comfy/model_management.py", line 143, in
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
^^^^^^^^^^^^^^^^^^
File "/media/phlq/HDD/ComfyUI/comfy/model_management.py", line 112, in get_torch_device
return torch.device(torch.cuda.current_device())
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/media/phlq/HDD/ComfyUI/venv/lib/python3.11/site-packages/torch/cuda/init.py", line 940, in current_device
_lazy_init()
File "/media/phlq/HDD/ComfyUI/venv/lib/python3.11/site-packages/torch/cuda/init.py", line 319, in _lazy_init
torch._C._cuda_init()
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx

also tried with the --directml vrible and got

Traceback (most recent call last):
File "/media/phlq/HDD/ComfyUI/main.py", line 91, in
import execution
File "/media/phlq/HDD/ComfyUI/execution.py", line 13, in
import nodes
File "/media/phlq/HDD/ComfyUI/nodes.py", line 21, in
import comfy.diffusers_load
File "/media/phlq/HDD/ComfyUI/comfy/diffusers_load.py", line 3, in
import comfy.sd
File "/media/phlq/HDD/ComfyUI/comfy/sd.py", line 5, in
from comfy import model_management
File "/media/phlq/HDD/ComfyUI/comfy/model_management.py", line 62, in
import torch_directml
ModuleNotFoundError: No module named 'torch_directml'

[daniandtheweb]

Directml doesn't work on Linux, if the tool redirects you to download the nvidia drivers there seems to be something wrong with it.
Try reinstalling the pip packages from ROCm_SDK _builder.