How to fix 'permissions denied' errors when installing NVIDIA driver for SageMaker Studio

Question

I'm using a SM studio instance (ml.g4dn.4xlarge) and getting permission denied error when trying to run the needed files. I first started[ here ](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/install-nvidia-driver.html) and selected Option 2. When I follow the [Quick start](https://docs.nvidia.com/datacenter/tesla/tesla-installation-notes/index.html)  for runfile installers I get the 'permission denied' error when I run 
```
 sudo sh NVIDIA-Linux-x86_64-$DRIVER_VERSION.run
```

I get the same error when I slow down and follow the [pre-installation actions ](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#pre-installation-actions)so I can't move forward

My SM role has SMFullAccess and EC2FullAccess

Answer

In SageMaker Studio, both `JupyterServer` and `KernelGateway` Apps are containers hosted on instances. We never install NVIDIA drivers in containers. They are controlled by docker runtime. If host system has NVIDIA driver and you start docker with gpu option, the driver will be available inside the container.

If you want to use GPU, all you need is CUDA. Studio started GPU instance with GPU docker option. This is nothing to do with IAM or something. It's a local Docker issue.

How to fix 'permissions denied' errors when installing NVIDIA driver for SageMaker Studio

相關內容