Sagemaker issues with pip?

Hi all, just a really quick question about something weird I’m experiencing with Fast.AI on Sagemaker. I used the CloudFormation links on the Fast.AI setup page to get my notebook set up in Sagemaker, and I’m using the Lifecycle Config on startup. However, when I’m actually in my notebook, something weird is going on when I try to use pip. I cannot pip install any modules. I mean, I can use !pip install , I can use !pip list inside the notebook, and my modules that I just !pip installed are there, but when I try to import that module, it says the module can’t be found.

What’s weird is if I change my notebook kernel to another one, like the the default Sagemaker conda_pytorch_36 kernel, and run !pip install, then change back to the Python 3 kernel that the lifecycle config creates, then the modules are installed.

Is there something I’m not understanding when it comes to the notebook kernel and how sets it up? Thanks for the help!