CODEWITHSUNDEEP
amazon-web-servicestensorflowamazon-sagemakeraws-deeplens
Justin P
Justin P

Reputation: 71

AWS Sagemaker PermissionError when starting training job?

I'm following the AWS DeepLens documentation: Build and Run the Head Pose Detection Project, and facing problems in the section Train a Head Pose Detection Model in SageMaker, specifically running estimator.fit(dataset_location) to start the training job.

These are the code cells that I've run, and this is the error I receive: PermissionError: [Errno 13] Permission denied: '/home/ec2-user/SageMaker/lost+found'.

Honestly, I'm not sure what the lost+found folder is. It seems to be a hidden folder in the Jupyter Notebook instance. I've seen it appear in video tutorials on YouTube, but I don't see it in my directory; however, it definitely exists as when I try to create a new folder named 'lost+found', I get this error: An error occurred while renaming "Untitled Folder" to "lost+found". File already exists: lost+found.

The get_execution_role() returns me the correct role, with the following permissions:

I've also tried upgrading to Tensorflow 2 and Python 3.6 instead of the original TF 1.4 and Python 2.7 (the tutorial was written in 2018) but still get the same error. My IAM account has been given admin rights so that shouldn't be the issue either. No one else seems to have this problem, I've searched everywhere so I'm at a loss for what else to try.

Would greatly appreciate input from the community, thanks!

Upvotes: 4

Views: 2062

Answers (1)

Justin P
Justin P

Reputation: 71

Found the issue - I wasn't able to see the 'lost+found' folder (i.e. hidden) because I didn't have the right permissions.

Fix: In the Jupyter notebook, run !sudo chmod 777 lost+found.

Upvotes: 3

Related Questions