Stuck at the post step (GPU)


It gets stuck when I run the demo python code and post the family-dog.jpg to do object detection on deepquestai/deepstack:gpu. Yet I had succeed in CPU version. I use Ubuntu 18.04, nvidia RTX 3070 , the driver version is 455. There seems no logs and I cannot find where is wrong. I guess maybe the problem is the driver version and I change it to 460 but that not work, just stuck after “post”. Any way to locate the problem?


the new gpu version does not solve the problem,
after I use a new gpu version image DeepStack February 2021 Release - Fixes for Jetson, Windows Native and Docker Versions
stuck become timeout, and the response in python is {‘success’: False, ‘error’: ‘failed to process request before timeout’, ‘duration’: 0}

@YuY2019 What version of CUDA and cuDNN do you have installed?

I didn’t install cuda/cuDNN then – I followed the document and the step3 says "Install NVIDIA Container Toolkit " which doesn’t need cuda/cudnn.

Yet I have installed cuda 11.0.207 and cuDNN 8.05 and tensorflow 2.4 GPU to run ImageAI and that ImageAI works.

Just now I tried deepstack again and it still got stuck.

Does your question means that “NVIDIA Container Toolkit” also has a version of cuda/cuDNN ? and how do I find that?