Computer shuts down during inference time

1

My computer shuts down when running PyTorch. I have the following specs:

Ubuntu 20.04
GPU: GTX1080
CUDA 10.2
Python 3.7
PyTorch 1.6.0

I’m running some model using the image from the webcam or video, but the CPU heats up and then shuts down. First I thought it might be the model/code that causing this, but I’ve run several different models, but the issue persists.

2

Are you pushing your model, inputs etc. to cuda?

3

Yes, I’m. it’s using ‘cuda:0’ as a device.

4

This sounds like a root cause and I’m not sure if this is still a PyTorch question.
Could you make sure your CPU is properly cooled?

5

It only happens with PyTorch (I’ve tried several pytorch implementation of human pose estimation techniques). I’ve been running other ml stuff (TF, scikit-learn); they run just fine.

Currently, I’m trying to run AlphaPose. Same issue.

6

If you cannot provide proper cooling, you could try to artificially slow down the code e.g. by adding sleeps to the code or reducing the number of used threads (via torch.set_num_threads and torch.set_num_interop_threads).