mdx.behavior_learning.model_management module

class ModelManagement(config)

Bases: object

Module to manage behavior-learning models.

Parameters:

config (dict) – configuration for the module

model_management = ModelManagement(config)

delete_past_versions(sensor)

Deletes the old version, retaining versionsToRetain number of versions. versionsToRetain is set using the config file.

Parameters:

sensor (string) – sensor for which to delete past versions of the model.

Returns:

Returns the number of versions deleted.

deleted_versions = model_management.delete_past_versions(sensor_name)

get_versions(sensor)

Returns a sorted list of model versions for the sensor.

Parameters:

sensor (string) – sensor for which model versions are to be returned.

Returns:

list of model versions (ints)

Return type:

python list

version_list = model_management.get_versions(sensor_name)

load_model(sensor)

Loads model into the Triton server. The version loaded is dependent on Triton configurations. Default is to always load the latest.

Parameters:

sensor (string) – name of the model to be loaded.

Returns:

None

model_management.load_model(model_name)

save_model(model, sensor)

Saves model by scripting and writing model to model_repository. Model_repository location is set using the config file. Currently only supports pyTorch models.

Parameters:

• model (nn.Module) – model to be saved.

• sensor (string) – sensor corresponding to the model.

Returns:

None

model_management.save_model(model, sensor_name)

unload_model(sensor)

Unloads the model from the triton server.

Parameters:

sensor (string) – name of the model to be unloaded.

Returns:

None

model_management.unload_model(model_name)

update_config(new_config)

Updates the config of the module with the user provided config.

Parameters:

new_config (dict) – new config for update

Returns:

None

model_management.update_config(new_config)

validate_model_management_config()