get_dl_pruning_param🔗
Short description🔗
get_dl_pruning_param — Get information from a pruning data handle.
Signature🔗
get_dl_pruning_param( dl_pruning DLPruningHandle, attribute.name GenParamName, out attribute.name GenParamValue )
Description🔗
get_dl_pruning_param returns the parameter and pruning values
GenParamValue of GenParamName from the pruning data
handle DLPruningHandle.
The values of following parameters GenParamName can be retrieved:
-
'mode': Specifies the pruning method See
create_dl_pruningfor further information.Default: 'oracle'
-
'percentage': Determines how much will be removed in percents. E.g., in case of 'mode'='oracle' it specifies the relative amount of possibly removable kernels to be removed, whereby only removable kernels are considered.
Default: 10
Additionally, for 'mode'='oracle', the following pruning
values can be retrieved using GenParamName:
-
'prunable_conv_num': Number of prunable convolutional layers in the network. Whether a layer is prunable depends on the specifications of the layer itself as well as the network architecture.
-
'prunable_conv_layer_name': Names of the prunable convolutional layers in the network.
-
'prunable_kernel_num': Number of prunable convolution kernels in the network. It considers that a pruned convolution has to retain at least 1 convolution kernel.
-
'pruned_kernel_num': Number of pruned convolution kernels in the network after a call of
gen_dl_pruned_model. -
'pruned_percentage': Actual percentage of pruned convolution kernels in the network after a call of
gen_dl_pruned_model: the ratio 'pruned_kernel_num' versus 'prunable_kernel_num'.
The following table gives an overview, which parameters and values can be set
using set_dl_pruning_param and which ones can be retrieved
using get_dl_pruning_param.
| Parameters | set |
get |
|---|---|---|
| 'mode' | x |
x |
| 'percentage' | x |
x |
| 'prunable_conv_num' | x |
|
| 'prunable_conv_layer_name' | x |
|
| 'prunable_kernel_num' | x |
|
| 'pruned_kernel_num' | x |
|
| 'pruned_percentage' | x |
Execution information🔗
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🔗
DLPruningHandle (input_control) dl_pruning → (handle)
Pruning data handle.
GenParamName (input_control) attribute.name → (string)
Name of the generic parameter.
Default: 'percentage'
List of values: 'mode', 'percentage', 'prunable_conv_layer_name', 'prunable_conv_num', 'prunable_kernel_num', 'pruned_kernel_num', 'pruned_percentage'
GenParamValue (output_control) attribute.name(-array) → (integer / string / real)
Value of the generic parameter.
Combinations with other operators🔗
Module🔗
Deep Learning Professional