SoftNeuro command list¶
The SoftNeuro command line tool has a variety of uses, as listed below.
It is possible to check all available commands by running the softneuro command without any arguments.
General Operation Commands¶
- help: Shows basic command help.
- version: Shows version information.
- license: Handle/display license information.
DNN File Operation Commands¶
Model Information¶
- attr: Show DNN file information.
- view: Show detailed information on the model structure and layers.
- plot: Save a DNN file model architecture as an image.
Model Editing¶
- extract: Extract weights and model data from a DNN file.
- archive: Archive weights and model data into a DNN file.
- refine: Optimize a DNN file for faster inference.
- compress: Reduce the DNN file weights data size.
- cipher: Encrypt a DNN file.
- calibrate : Calibrate a DNN file for static quantization.
- decompose : Decompose net of a DNN file.
Model Conversion¶
- import-onnx: Convert an ONNX model into a DNN file.
- import-tensorflow: Convert a TensorFlow or Keras model into a DNN file.
- import-ver3: Convert a SoftNeuro V3 DNN file into a V4 DNN file.
Execution, Profiling and Tuning Commands¶
- run: Run inference using a DNN file.
- init: Init profiling data.
- add: Add layer routine settings to profiling data.
- rm: Remove routine information from profiling data.
- reset: Reset profiling information.
- status: Show profiling results information.
- profile: Run profiling using profiling data to generate profiling information.
- tune: Tune a DNN file.
Layer/Routine Operation Commands¶
- plugins: Show available plugins, such as AVX and CUDA.
- layers: Shows available layers.
- lparams: Shows the parameters for a given layer.
- routine: Shows available routines.
- rparams: Shows the parameters for a given routine.
- test: Test runs a given layer.