Model Export

nexport.export()

This function is the meat of nexport. It allows users to pass in a model compiled with either PyTorch or TensorFlow and export it to any supported file type.

Usage

import nexport

nexport.export(model=model,
               filetype="json"
               filename="model",
               indent=4,
               verbose=1,
               include_metadata=True,
               model_name="Example Model",
               model_author="Jordan Welsman",
               activation_function="sigmoid",
               uses_skip_connections=False
)

Arguments

Argument	Data type	Default	Description
model	object		Instantiated PyTorch or TensorFlow model
filetype	string		Intended export filetype extension
filename	string	model	Name to be used for output file, excluding extension
indent	integer	4	Number of spaces to be used to indent JSON object
verbose	integer	1	Degree of status output for model export
include_metadata	boolean	False	Whether to include arguments as metadata header in output file
model_name	string	My Model	Name of model to be used in metadata header
model_author	string	None	Name of author of model to be used in metadata header
activation_function	string	None	Activation function reference to be used in metadata header
uses_skip_connections	boolean	False	Whether the model uses skip connections

indent

This is the number of spaces each line indentation should take up. This only applies to exporting to .json, and .xml. Other filetypes will ignore this parameter.

verbose

This is the degree of verbosity the function should output to when reporting the export status. There are currently four levels of verbosity:

• 0: Status reporting disabled. No output to console when exporting.
• 1: Status reported when model has been successfully exported.
• 2: Status reported when parameters have been extracted from passes model, model object has been created, and export time reported, in addition to all updates in 1.
• 3: Status reported at start of layer creation and after each layer has been compiled, in addition to all updates in 2.

include_metadata

This is whether to include model_name, model_author, activation_function, uses_skip_connections and the compilation date in the header of the exported file:

• True: Tells the function to compile a metadata object and include it when writing the exported file.
• False: Tells the function to discard any passed values for model_name and model_author and to not compile a metadata object.

Accepted arguments

Argument	Required	Accepted values
model	✅	Any fully-connected PyTorch or TensorFlow model
filetype	✅	txt, json, csv, xml
filename	❌	Any string
indent	❌	Any integer
verbose	❌	0, 1, 2, 3
include_metadata	❌	True, False
model_name	❌	Any string
model_author	❌	Any string
activation_function	❌	Any string
uses_skip_connections	❌	True, False

Returns

Note This function does not currently return anything, but a success boolean return is planned.

Side-effects

This function exports the passed model to a file created in the directory nexport is called from. In most cases, this is where the calling file is located. If you are using Jupyter, the file will be created where the notebook is located.

This function also outputs status updates to the terminal depending on the degree of verbosity passed.