/torch2coreml

Torch7 -> CoreML

Primary LanguagePythonMIT LicenseMIT

Convert Torch7 models into Apple CoreML format.

Short tutorial

This tool helps convert Torch7 models into Apple CoreML format which can then be run on Apple devices.

fast-neural-style example app screenshot

Installation

pip install -U torch2coreml

In order to use this tool you need to have these installed:

  • Xcode 9
  • python 2.7

If you want to run tests, you need MacOS High Sierra 10.13 installed.

Dependencies

  • coremltools (0.6.2+)
  • PyTorch

How to use

Using this library you can implement converter for your own model types. An example of such a converter is located at "example/fast-neural-style/convert-fast-neural-style.py". To implement converters you should use single function "convert" from torch2coreml:

from torch2coreml import convert

This function is simple enough to be self-describing:

def convert(model,
            input_shapes,
            input_names=['input'],
            output_names=['output'],
            mode=None,
            image_input_names=[],
            preprocessing_args={},
            image_output_names=[],
            deprocessing_args={},
            class_labels=None,
            predicted_feature_name='classLabel',
            unknown_layer_converter_fn=None)

Parameters

model: Torch7 model (loaded with PyTorch) | str
A trained Torch7 model loaded in python using PyTorch or path to file with model (*.t7).

input_shapes: list of tuples Shapes of the input tensors.

mode: str ('classifier', 'regressor' or None)
Mode of the converted coreml model:
'classifier', a NeuralNetworkClassifier spec will be constructed.
'regressor', a NeuralNetworkRegressor spec will be constructed.

preprocessing_args: dict
'is_bgr', 'red_bias', 'green_bias', 'blue_bias', 'gray_bias', 'image_scale' keys with the same meaning as https://apple.github.io/coremltools/generated/coremltools.models.neural_network.html#coremltools.models.neural_network.NeuralNetworkBuilder.set_pre_processing_parameters

deprocessing_args: dict
Same as 'preprocessing_args' but for deprocessing.

class_labels: A string or list of strings.
As a string it represents the name of the file which contains the classification labels (one per line). As a list of strings it represents a list of categories that map the index of the output of a neural network to labels in a classifier.

predicted_feature_name: str
Name of the output feature for the class labels exposed in the Core ML model (applies to classifiers only). Defaults to 'classLabel'

unknown_layer_converter_fn: function with signature:
(builder, name, layer, input_names, output_names)
builder: object - instance of NeuralNetworkBuilder class
name: str - generated layer name
layer: object - PyTorch (python) object for corresponding layer
input_names: list of strings
output_names: list of strings
Returns: list of strings for layer output names
Callback function to handle unknown for torch2coreml layers

Returns

model: A coreml model.

Currently supported

Models

Only Torch7 "nn" module is supported now.

Layers

List of Torch7 layers that can be converted into their CoreML equivalent:

  1. Sequential
  2. ConcatTable
  3. SpatialConvolution
  4. ELU
  5. ReLU
  6. SpatialBatchNormalization
  7. Identity
  8. CAddTable
  9. SpatialFullConvolution
  10. SpatialSoftMax
  11. SpatialMaxPooling
  12. SpatialAveragePooling
  13. View
  14. Linear
  15. Tanh
  16. MulConstant
  17. SpatialZeroPadding
  18. SpatialReflectionPadding
  19. Narrow
  20. SpatialUpSamplingNearest
  21. SplitTable

License

Copyright (c) 2017 Prisma Labs, Inc. All rights reserved.

Use of this source code is governed by the MIT License that can be found in the LICENSE.txt file.