SyncTrainingBatches

Document processing is carried out in several different projects or environments. Hence, images are accumulated, and models are trained in a specific project or environment. If a need to use the training results in a different environment arises, trained models will need to be transferred from one environment to another. The SyncTrainingBatches command is used to transfer training batches from one project or environment to another. Only the models that have a higher quality than the ones already present in an existing project or environment will be transferred. This command can also be used to transfer images used for training.

Parameters

/Source Required The project from which the training batches should be copied.
/Target Required The project to which the training batches should be copied.
/SourceTenant Optional The tenant containing the source project.
/TargetTenant Optional The tenant where the project should be updated.
/SourceU Optional User name.

If no user name is specified, Windows authentication will be used.
/SourceP Optional Password.
/TargetU

Optional

User name.

If no user name is specified, Windows authentication will be used.

/TargetP Optional Password.
/CopyTrainingBatches

Optional

Copies the training batches.

Possible values are All, LayoutOnly, ClassifierOnly or None. By default, this parameter is set to All.

/BatchIDs

Optional

The IDs of the batches that should be synchronized.

If no batch IDs are specified, all the training batches will be synchronized. If some batch IDs are specified, the /CopyTrainingBatches parameter should be set to either LayoutOnly or ClassifierOnly to avoid duplicate batch IDs as different collections may contain identical batch IDs.

/CopyTrainingSamples Flag

Copies the contents of a training batch.

If the source and target batches have not changed and /ForceSync is enabled (i.e. their GUIDs are identical), no batch contents will be copied.

The contents of a batch are copied only if the batch itself is copied.

/ForceSync Flag Disables the training quality check and copies a training batch even if the quality of the new model is worse than that of the existing model.

Example of use:

SyncTrainingBatches /Source="https://FCSRV/1/SingleEntryPoint" /Target=D:\Projects\SingleEntryPoint /SourceU=user /SourceP=password /ForceSync /BatchIDs=1,2,3,4

10.11.2020 12:08:03


Please leave your feedback about this article