The manifest file contains JSON-formatted data that maps the input file to the prediction results file, as follows:
{S3 location of the batch prediction input file.csv : S3 location of the prediction results file}
In our context, the manifest file contains the following line:
{"s3://aml.packt/data/ch6_titanic_heldout.csv":"s3://aml.packt/batch-prediction/result/bp-yTDNSArMqa6-ch6_titanic_heldout.csv.gz"}
Multiple input files: If your input data is split into several files, all stored in the same S3 location s3://examplebucket/input/, all the input files will be considered by the batch prediction job. The manifest file will then contain the mapping from the different input files to the associated results files. ...