job.yml

$schema: https://azuremlschemas.azureedge.net/latest/commandJob.schema.json
code: 
  local_path: src
command: >-
  python startDask.py
  --script prep-nyctaxi.py 
  --nyc_taxi_dataset ${{inputs.nyc_taxi_dataset}}
  --output_folder ${{outputs.output_folder}}
inputs:
  nyc_taxi_dataset:
    folder: wasbs://datasets@azuremlexamples.blob.core.windows.net/nyctaxi/
    mode: ro_mount
outputs:
  output_folder:
environment: 
  image: mcr.microsoft.com/azureml/openmpi4.1.0-ubuntu20.04
  conda_file: conda.yml
compute: azureml:cpu-cluster-lg
resources:
  instance_count: 4
distribution:
  type: pytorch
display_name: dask-nyctaxi-example
experiment_name: dask-nyctaxi-example
description: This sample shows how to run a distributed DASK job on AzureML. The 24GB NYC Taxi dataset is read in CSV format by a 4 node DASK cluster, processed and then written as job output in parquet format.