Operator Reference
create_dl_pruning (Operator)
create_dl_pruning
— Create a pruning data handle.
Signature
create_dl_pruning( : : DLModelHandle, Mode, GenParam : DLPruningHandle)
Description
The operator create_dl_pruning
creates a handle
DLPruningHandle
.
This handle is used to pass information when pruning a deep learning model.
Mode
specifies the pruning method.
The mode implies what will be removed and also which scores are needed.
Currently only the following 'mode' is implemented:
- 'oracle' :
Kernels of convolution layers are removed. In order to do so, a score is calculated for every (potentially removable) kernel, indicating its importance within the given network for the images used. See the given reference for details.
GenParam
is a dictionary for setting generic parameters.
Currently no generic parameters are supported.
For an explanation of the concept of deep learning see the introduction of chapter Deep Learning.
Execution Information
- Multithreading type: reentrant (runs in parallel with non-exclusive operators).
- Multithreading scope: global (may be called from any thread).
- Processed without parallelization.
This operator returns a handle. Note that the state of an instance of this handle type may be changed by specific operators even though the handle is used as an input parameter by those operators.
Parameters
DLModelHandle
(input_control) dl_model →
(handle)
Handle of a deep learning model.
Mode
(input_control) string →
(string)
Pruning method.
Default: 'oracle'
List of values: 'oracle'
GenParam
(input_control) dict →
(handle)
Dictionary with generic parameters.
Default: []
DLPruningHandle
(output_control) dl_pruning →
(handle)
Pruning data handle.
Possible Predecessors
Possible Successors
get_dl_pruning_param
,
set_dl_model_param
References
Pavlo Molchanov, Stephen Tyree, Tero Karras, Timo Aila, Jan Kautz "Pruning Convolutional Neural Networks for Resource Efficient Inference", ICLR 2017, 5th International Conference on Learning Representations, Toulon, France.
Module
Deep Learning Professional