create_dl_layer_softmaxT_create_dl_layer_softmaxCreateDlLayerSoftmaxCreateDlLayerSoftmaxcreate_dl_layer_softmax (Operator)

Name

create_dl_layer_softmaxT_create_dl_layer_softmaxCreateDlLayerSoftmaxCreateDlLayerSoftmaxcreate_dl_layer_softmax — Create a softmax layer.

Signature

Description

The operator create_dl_layer_softmaxcreate_dl_layer_softmaxCreateDlLayerSoftmaxCreateDlLayerSoftmaxCreateDlLayerSoftmaxcreate_dl_layer_softmax creates a softmax layer whose handle is returned in DLLayerSoftMaxDLLayerSoftMaxDLLayerSoftMaxDLLayerSoftMaxDLLayerSoftMaxdllayer_soft_max.

The parameter DLLayerInputDLLayerInputDLLayerInputDLLayerInputDLLayerInputdllayer_input determines the feeding input layer and expects the layer handle as value.

The parameter LayerNameLayerNameLayerNameLayerNamelayerNamelayer_name sets an individual layer name. Note that if creating a model using create_dl_modelcreate_dl_modelCreateDlModelCreateDlModelCreateDlModelcreate_dl_model each layer of the created network must have an unique name.

The softmax layer applies the softmax function which is defined for each input as follows: where is the number of inputs. During training, the result of the softmax function is additionally shifted and transformed by a logarithm function, such that the values are suitable as input to e.g., a cross entropy loss layer. This behavior can be changed by setting the generic parameter 'output_mode'"output_mode""output_mode""output_mode""output_mode""output_mode", see below.

The following generic parameters GenParamNameGenParamNameGenParamNameGenParamNamegenParamNamegen_param_name and the corresponding values GenParamValueGenParamValueGenParamValueGenParamValuegenParamValuegen_param_value are supported:

'output_mode'"output_mode""output_mode""output_mode""output_mode""output_mode":

This parameter determines if and in which case the output is transformed by a logarithm function:

• 'default'"default""default""default""default""default": During inference, the result of the softmax function is returned as output while during training, the softmax is further transformed by a logarithm function.

• 'no_log_training'"no_log_training""no_log_training""no_log_training""no_log_training""no_log_training": The logarithm transformation is omitted during training and the result of the softmax function is directly returned.

• 'log_inference'"log_inference""log_inference""log_inference""log_inference""log_inference": The logarithm of the softmax is calculated during inference in the same way as during training.

Default: 'default'"default""default""default""default""default".

'is_inference_output'"is_inference_output""is_inference_output""is_inference_output""is_inference_output""is_inference_output":

Determines whether apply_dl_modelapply_dl_modelApplyDlModelApplyDlModelApplyDlModelapply_dl_model will include the output of this layer in the dictionary DLResultBatchDLResultBatchDLResultBatchDLResultBatchDLResultBatchdlresult_batch even without specifying this layer in OutputsOutputsOutputsOutputsoutputsoutputs ('true'"true""true""true""true""true") or not ('false'"false""false""false""false""false").

Default: 'false'"false""false""false""false""false"

'num_trainable_params'"num_trainable_params""num_trainable_params""num_trainable_params""num_trainable_params""num_trainable_params":

Number of trainable parameters (weights and biases) of the layer.

Certain parameters of layers created using this operator create_dl_layer_softmaxcreate_dl_layer_softmaxCreateDlLayerSoftmaxCreateDlLayerSoftmaxCreateDlLayerSoftmaxcreate_dl_layer_softmax can be set and retrieved using further operators. The following tables give an overview, which parameters can be set using set_dl_model_layer_paramset_dl_model_layer_paramSetDlModelLayerParamSetDlModelLayerParamSetDlModelLayerParamset_dl_model_layer_param and which ones can be retrieved using get_dl_model_layer_paramget_dl_model_layer_paramGetDlModelLayerParamGetDlModelLayerParamGetDlModelLayerParamget_dl_model_layer_param or get_dl_layer_paramget_dl_layer_paramGetDlLayerParamGetDlLayerParamGetDlLayerParamget_dl_layer_param. Note, the operators set_dl_model_layer_paramset_dl_model_layer_paramSetDlModelLayerParamSetDlModelLayerParamSetDlModelLayerParamset_dl_model_layer_param and get_dl_model_layer_paramget_dl_model_layer_paramGetDlModelLayerParamGetDlModelLayerParamGetDlModelLayerParamget_dl_model_layer_param require a model created by create_dl_modelcreate_dl_modelCreateDlModelCreateDlModelCreateDlModelcreate_dl_model.

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

Module

Deep Learning Training