AWS Sagemaker unable to parse csv

Question

I'm trying to run a training job on AWS Sagemaker, but it keeps failing giving the following error:

ClientError: Unable to parse csv: rows 1-5000, file /opt/ml/input/data/train/KMeans_data.csv

I've selected 'text/csv' as the content type and my CSV file contains 5 columns with numerical content and text headers.

Can anyone point out what could be going wrong here?

Thanks!

Reservation

could you post a snippet of what the CSV looks like, and the method that you're using to to load and parse the file? – Dock

do you have a numeric label in your first column? – Bioplasm

Answer 1

From https://docs.aws.amazon.com/sagemaker/latest/dg/cdf-training.html CSV must not have headers:

Amazon SageMaker requires that a CSV file doesn't have a header record ...

Try removing the header row.

Reserpine

Just came here to say that if you have only 1 feature column (2 columns total), SageMaker doesn't like that either – Akira

Answer 2

Try to make sure that there are no other files other than the training file in the training folder in S3 bucket.

Rudd

Answer 3

Make sure to not have any new lines in the cells of .csv file.

Boycie