Roop not using gcolab GPUs

[cubargh]

Describe the bug
I have collab premium. Copied the notebook, ran it in a v100 hardware. No GPU utilization. Tried it in a100, same result

Help?

got this error on console
[E:onnxruntime:Default, provider_bridge_ort.cc:1480 TryGetProviderInfo_CUDA] /onnxruntime_src/onnxruntime/core/session/provider_bridge_ort.cc:1193 onnxruntime::Provider& onnxruntime::ProviderLibrary::Get() [ONNXRuntimeError] : 1 : FAIL : Failed to load library libonnxruntime_providers_cuda.so with error: libcufft.so.10: cannot open shared object file: No such file or director

[cubargh]

@C0untFloyd got any idea why is this happening?

[marat2509]

Google Colab has updated the Cuda version to version 12. ONNXRuntime does not support this version yet, but this feature will appear soon
See microsoft/onnxruntime#18850
Also see googlecolab/colabtools#4214

[cubargh]

For a temporary fix, go to the command palette (ctrl+shift+P) and select "Use fallback runtime version"

[C0untFloyd]

I was having problems on my local pc as well, suddenly not recognizing cuda but instead having "Azure" among the execution providers. I went back to installing onnxruntime-gpu v1.16.2 which seemed to have solved it. Weird...

[thany22]

For a temporary fix, go to the command palette (ctrl+shift+P) and select "Use fallback runtime version"

The previous runtime version isn't available anymore. Any solutions)

[marat2509]

For a temporary fix, go to the command palette (ctrl+shift+P) and select "Use fallback runtime version"

The previous runtime version isn't available anymore. Any solutions)

Install latest official onnxruntime-gpu

[Marv761125]

For a temporary fix, go to the command palette (ctrl+shift+P) and select "Use fallback runtime version"

The previous runtime version isn't available anymore. Any solutions)

Install latest official onnxruntime-gpu

Hi. Could you tell me how to do this or is there a link you used? Thx.

[marat2509]

how to do this

pip install --upgrade onnxruntime-gpu

[Marv761125]

how to do this

pip install --upgrade onnxruntime-gpu

Hi Marat. This may sound pretty stupid, but where do I have to put this codeline in? I use the Roop-Unleashed colab version. Thx.

[htmlcodepreview]

how to do this

pip install --upgrade onnxruntime-gpu

Tried this and several versions like 16.2 17 and 17.1, with no success.

[marat2509]

Setup (after cloning):

import os
!pip install pyyaml -r requirements.txt

if os.path.isfile("/usr/bin/nvidia-smi") or os.path.isfile("/opt/bin/nvidia-smi"):
    !pip uninstall -y onnxruntime onnxruntime-gpu
    !pip install onnxruntime-gpu==1.17.0 --extra-index-url https://pkgs.dev.azure.com/onnxruntime/onnxruntime/_packaging/onnxruntime-cuda-12/pypi/simple

Run:

import yaml
import os

roop_config = {
    'clear_output': True,
    'live_cam_start_active': False,
    'max_threads': 32,
    'output_image_format': 'png',
    'output_video_codec': 'libx264',
    'output_video_format': 'mp4',
    'video_quality': 14,
    'frame_buffer_size': 6,
    'selected_theme': 'Default',
    'server_name': '',
    'server_port': 0,
    'server_share': True
}

if os.path.isfile("/usr/bin/nvidia-smi") or os.path.isfile("/opt/bin/nvidia-smi"):
    roop_config["provider"] = "cuda"
    print("I: Using CUDA")
else:
    roop_config["provider"] = "cpu"

with open("./config.yaml", "w") as config_file:
    config_file.write(yaml.dump(roop_config))

!python run.py

[zachysaur]

wht about on kaggle?

[michidaeron]

how to do this

pip install --upgrade onnxruntime-gpu

Is there no way to avoid the message about disallowed code and cause it to be disconnected?
Using fallback runtime version it worked before but not now :C

[marat2509]

@michidaeron @htmlcodepreview @Marv761125 - me updated and tested new setup code. Please test

[htmlcodepreview]

it works for me now

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[github-actions]

This issue was closed because it has been stalled for 5 days with no activity.