load_dl_model_weightsT_load_dl_model_weightsLoadDlModelWeightsLoadDlModelWeightsload_dl_model_weights (Operator)

Name

load_dl_model_weightsT_load_dl_model_weightsLoadDlModelWeightsLoadDlModelWeightsload_dl_model_weights — Load the weights of a source model into a target model.

Signature

load_dl_model_weights( : : DLModelHandleSource, DLModelHandleTarget : ChangesByLayer)

Description

The operator load_dl_model_weightsload_dl_model_weightsLoadDlModelWeightsLoadDlModelWeightsLoadDlModelWeightsload_dl_model_weights loads weights of a source model DLModelHandleSourceDLModelHandleSourceDLModelHandleSourceDLModelHandleSourceDLModelHandleSourcedlmodel_handle_source into a target model DLModelHandleTargetDLModelHandleTargetDLModelHandleTargetDLModelHandleTargetDLModelHandleTargetdlmodel_handle_target. Thereby applies for every layer in the target model: Its weights are only changed if there is a layer in the source model having the same name and the same weight-shape. Note that DLModelHandleSourceDLModelHandleSourceDLModelHandleSourceDLModelHandleSourceDLModelHandleSourcedlmodel_handle_source must be different from DLModelHandleTargetDLModelHandleTargetDLModelHandleTargetDLModelHandleTargetDLModelHandleTargetdlmodel_handle_target, i.e., you cannot use the same model handle as source and target.

ChangesByLayerChangesByLayerChangesByLayerChangesByLayerchangesByLayerchanges_by_layer is a tuple indicating for every target layer how many weights changed. Its entries are sorted by ascending layer IDs. The layer IDs can be queried via the operator get_dl_model_paramget_dl_model_paramGetDlModelParamGetDlModelParamGetDlModelParamget_dl_model_param with the parameter 'summary'"summary""summary""summary""summary""summary".

Note, that 'weights' means all weights and biases for all layers which can have such values (e.g., convolutional layer, batch normalization layer, etc.).

Attention

The operator load_dl_model_weightsload_dl_model_weightsLoadDlModelWeightsLoadDlModelWeightsLoadDlModelWeightsload_dl_model_weights is only applicable to self-created networks. For networks delivered by HALCON, the operator is not working.

Execution Information

Multithreading type: reentrant (runs in parallel with non-exclusive operators).
Multithreading scope: global (may be called from any thread).
Processed without parallelization.

Parameters

DLModelHandleSourceDLModelHandleSourceDLModelHandleSourceDLModelHandleSourceDLModelHandleSourcedlmodel_handle_source (input_control) dl_model → (handle)

Handle of the source deep learning model.

DLModelHandleTargetDLModelHandleTargetDLModelHandleTargetDLModelHandleTargetDLModelHandleTargetdlmodel_handle_target (input_control) dl_model → (handle)

Handle of the target deep learning model.

ChangesByLayerChangesByLayerChangesByLayerChangesByLayerchangesByLayerchanges_by_layer (output_control) integer(-array) → (integer)

Indicates for every target layer how many weights changed.

Result

If the parameters are valid, the operator load_dl_model_weightsload_dl_model_weightsLoadDlModelWeightsLoadDlModelWeightsLoadDlModelWeightsload_dl_model_weights returns the value 2 ( H_MSG_TRUE) . If necessary, an exception is raised.

Module

Deep Learning Training

Operators