The model monitoring job is not executed.

[YoshinaoMori]

I followed the URL below to deploy the model to the endpoint.

https://github.com/GoogleCloudPlatform/vertex-ai-samples/blob/main/notebooks/official/pipelines/goog...

Manually I enabled monitoring for the model, but the monitoring job did not run and I got the message "The first run of the monitoring job is pending." 

I tried both Training-serving skew detection and Prediction drift detection, but they did not run.

How do I run a monitoring job?

I requested 20 predictions after deploying. Need more requests?

○Settings

Monitoring window length: 1hour

Sampling rate: 100%

Prediction input schema: not specified

[kvandres · Accepted solution]

Good day @YoshinaoMori,

Welcome to Google Cloud Community!

You need to provide the schema of your table dataset in order for it to correctly parse the input payload. To provide your own input schema, you must create a YAML file. you can refer to this link for an example: https://cloud.google.com/vertex-ai/docs/model-monitoring/schemas#custom-input-schemas-online
If you are trying to use automatic schema parsing where the model monitoring automatically parses the input schema, you might need 1,000 input requests in order for the model monitoring to determine its schema. It will work best if the input requests are in key-value pairs. Here is an example:

{"Name":"Mike", "age":"30", "gender":"M", "ethnicity":"latin american"}

You can check this link for more information: https://cloud.google.com/vertex-ai/docs/model-monitoring/schemas#automatic-schema-parsing-online

You can also check this guide on how to turn on monitoring models: https://cloud.google.com/blog/topics/developers-practitioners/monitor-models-training-serving-skew-v...

Hope this helps!

[YoshinaoMori]

Hi @kvandres

Thanks for answering.

Since we are still in the development stage, we will try to define the schema instead of making nearly 1000 requests.

[YoshinaoMori]

Hi @kvandres 

Once the schema was defined, the monitoring job was executed! Thanks!