DataFrame column names are stripped when combining timeseries forecast output.

[felixcollins]

I'm running a timeseries forecast using the many models accelerator and the results come back in a plain text file with whitespace delimited values. There are no headers in the first row. The return value from each mini batch is a pandas DataFrame with columns names. These are being combined to make the output file but the column names are lost.

[cartacioS]

To add header in output automatically, please add below argument when creating the step.

Step = ParallelRunStep( … arguments=["--append_row_dataframe_header", True] … )

[deeptim123]

If you are using AutoMLPipelineBuilder.get_many_models_batch_inference_steps to build the parallel run step you can pass that as arguments as follows

inference_steps = AutoMLPipelineBuilder.get_many_models_batch_inference_steps( ... arguments=["--append_row_dataframe_header", True] ...)

[felixcollins]

Thanks I'll give it a shot and close the ticket when it is going.

[felixcollins]

That worked. Thanks!