Baked in ML

ML Component does not Start

Verify that all the pre-requisites listed in the ‘Albie Basic/ML Prerequisite’ section of ‘Installation’ are met.

Check below log files for any errors:

<Container_Home>\data\Command\ML.log
<Container_Home>\data\ML\modelgenerator.log
<Container_Home>\data\ML\predictor.log
<Container_Home>\data\ML\queueforecast.log
Enable DEBUG log level in below configuration files to get debug level logs:
<Container_Home>\app\ML\modelgenerator\config.py
<Container_Home>\app\ML\predictor\api\config.py

<Container_Home>\app\ML\queueforecast\config.py

NOTE:

ML service restart is required for any configuration change to reflect.

Verify configurations inside <Container_Home>\app\Admin\config\config.yml

Set mlPrediction property to “true” for –

ML training data collection for generation/re-training the ML model for a process.

Prediction of the execution time of every RPA transaction

NOTE:

Prediction will not happen for a transaction unless ML model for that process has been generated

Set the minimum number of successful transactions for a process considered for calculation of Average Execution Time.

Verify configurations inside <Container_Home>\app\Vanguard\Monitor\PeriodicXml\MonitorAlerts.xml

Turn AutoscaleModelRetrain job Mode “on”. This is to start collecting training data for ML model generation/retrain.

NOTE:

ML training data will not be collected for retraining/creation of ML models unless this setting is turned “on”.

“RepeatInterval” property is set to 1440 minutes (1 day). Update this value in case required.

NOTE:

Do not touch any other property inside this section

Verify configurations inside <Container_Home>\app\Vanguard\appsettings.json
- This is the directory where ML Training data will get stored. In case of distributed or cluster/HA setup; update this property with a shared network drive location. Refer ‘Albie Basic/ML Prerequisite’ section of ‘Installation’ for details to mount network drive
- This is the minimum number of successful transactions for a process; after which ML training data will be generated for the creation of ML model for that process.
- This is the maximum number of successful transactions for a process; which will be taken into consideration for ML training.
- This is the ML training interval in days for a process. Once the model is trained; it will be re-trained after this interval occurs.
Check below log files for any errors:
- <Container_Home>\data\Command\<hostname>_Vanguard<YYYYMMDD>.log
Update below configuration inside <Container_Home>\app\Vanguard\appsettings.json to get debug level logs:
- Set the default value to “Verbose”
Verify configurations inside <Container_Home>\runtime\ logstash\logstash-7.10.2\config\ae\automationRequest.conf
- Set property “[@metadata][mlPrediction]" to “true”
Verify configurations inside <Container_Home>\runtime\ \logstash\logstash-7.10.2\Config\ae\transaction.conf
- Set property “[@metadata][mlPrediction]" to “true”

Verify that training data csv has got generated under configured location. Refer ‘ML component does not start section’ for details.
Verify below configuration inside <Container_Home>\app\ML\modelgenerator\config.py
- Verify the below configured directory where generated ML model will get stored. In case of distributed or cluster/HA setup; update this property with a shared network drive location. Refer ‘Albie Basic/ML Prerequisite’ section of ‘Installation’ for details to mount network drive.
- Verify the below configuration to set the time interval in which the Model Generator checks for any new ML Model is to be generated.

Check the modelgenerator logs for any errors. Refer ‘ML component does not start’ section for more details.

Verify that the ML model has got generated under configured location.
Verify configurations inside <Container_Home>\app\ app\Vanguard\Monitor\PeriodicXml \MonitorAlerts.xml
- Make sure the below jobs are turned “on”
Make sure the SMTP server is configured correctly. For more information, see ‘Server Properties’ section in ‘Installation’.