pytorch_lightning.utilities.exceptions.MisconfigurationException: GPUAccelerator can not run on your system since the accelerator is not available. The following accelerator(s) is available and can be passed into accelerator argument of Trainer: ['cpu'].

[uiop2020]

I am receiving this error, but I have GPU on my device. what should I do to be able to use GPU?

pytorch_lightning.utilities.exceptions.MisconfigurationException: GPUAccelerator can not run on your system since the accelerator is not available. The following accelerator(s) is available and can be passed into accelerator argument of Trainer: ['cpu'].

[akihironitta]

Hi @uiop2020 PyTorch Lightning simply uses torch.cuda.device_count() to detect if any GPU(s) is available:
https://github.com/PyTorchLightning/pytorch-lightning/blob/bfa8b7be2d99b980afa62f5cb0433326bcfd2ef0/pytorch_lightning/accelerators/gpu.py#L90-L91

so, I believe that's an issue with your environment. To address your issue, you might want to read the threads linked in this comment.

[rehammsalah]

did you solve this problem?
cuz i have the same error

[akihironitta]

What is the output of torch.cuda.device_count() in your env?

[kaan-amani]

This is how it looks like normally

But after I run my training script, which was working fine until last week, gives this error and changes the output for device count

[Huiflorazhan]

What is the output of torch.cuda.device_count() in your env?

my output is 0. Could you help me with this problem?

[jxtngx]

can you please run the following command in terminal, and post a screenshot of the result here.

nvidia-smi

[Huiflorazhan]

The output is command not found. I'm using Mac.

[jxtngx]

okay, that explains it. when on mac, we need to set accelerator to mps instead of gpu. and devices should be 1.

[Huiflorazhan]

Could you please help me with this issue? Thanks~
scverse/scvi-tools#2246