Describe the bug
I am having trouble with data post-processing in AWS Sagemaker, where I need to split one large text file with predictions (~2-10 GB) into millions of small files (one file per user ~3-10KB).
I've been able to process a small dataset (32MB, 13540 records). When I try 1.2 million records (2.2 GB), ScriptProcessor successfully processes the input file and saves the output files to /opt/ml/processing/output, however it fails to put them in S3 with an error.
Describe the bug I am having trouble with data post-processing in AWS Sagemaker, where I need to split one large text file with predictions (~2-10 GB) into millions of small files (one file per user ~3-10KB).
I've been able to process a small dataset (32MB, 13540 records). When I try 1.2 million records (2.2 GB), ScriptProcessor successfully processes the input file and saves the output files to
/opt/ml/processing/output
, however it fails to put them in S3 with an error.To reproduce Jupyter notebook:
callable.py:
Dockerfile:
Expected behavior All files that I save to
/opt/ml/processing/output
should be saved to S3.Screenshots or logs
System information
Additional context See my stackoverflow question for more details.