AzureML model monitoring error

Question

I try to set up model monitoring as in Azure ML documentation. When I run the monitoring job, I get the following error:

No data found for the given time window: 2024-07-29 07:54:54.277000+00:00 to 2024-08-05 07:54:54.277000+00:00 in input wasbs://******.blob.core.windows.net/modelDataCollector/******/******/model_inputs. We expect folder pattern <root>/YYYY/MM/DD/HH/<your_log>.jsonl

I have the error for any kind of signal, and for any data source. Input data and output data is already stored in jsonl format by data Collector, and for data reference no format is specified in the Azure documentation (my data is in tabular CSV format).

Answer 1

Verify if your data is stored in the correct folder pattern as expected by Azure ML:

<root>/YYYY/MM/DD/HH/<your_log>.jsonl

Make sure that the time window specified in your monitoring job accurately reflects the timestamps of the data stored in your blob storage. The error suggests no data was found within the specified time window, so it’s important to ensure the timestamps in your data match the requested window.

Even though the documentation may not specify a format for data reference, it's important to adhere to the expected JSONL format for data collected via the data collector. For tabular CSV data, ensure proper conversion to JSONL if necessary.

If you need to convert CSV data to JSONL:

import csv

import json

csv_file_path = 'data.csv'

jsonl_file_path = 'data.jsonl'

with open(csv_file_path, 'r') as csv_file:

    csv_reader = csv.DictReader(csv_file)

    with open(jsonl_file_path, 'w') as jsonl_file:

        for row in csv_reader:

            json.dump(row, jsonl_file)

            jsonl_file.write('\n')

Answer 2

The solution has finally been to transform the labeled and training data as JSON Lines. There has been no way to make a pre-processing component work.