openvino/docs/MO_DG/Deep_Learning_Model_Optimizer_DevGuide.md

Converting Models with Model Optimizer

@sphinxdirective

.. _deep learning model optimizer:

.. toctree:: :maxdepth: 1 :hidden:

openvino_docs_MO_DG_prepare_model_convert_model_Converting_Model openvino_docs_MO_DG_prepare_model_Model_Optimization_Techniques openvino_docs_MO_DG_prepare_model_convert_model_Cutting_Model openvino_docs_MO_DG_Additional_Optimization_Use_Cases openvino_docs_MO_DG_FP16_Compression openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_TensorFlow openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_ONNX openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_PyTorch openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_Paddle openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_MxNet openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_Caffe openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_Kaldi openvino_docs_MO_DG_prepare_model_convert_model_tutorials openvino_docs_MO_DG_prepare_model_Model_Optimizer_FAQ

@endsphinxdirective

Model Optimizer is a cross-platform command-line tool that facilitates the transition between training and deployment environments, performs static model analysis, and adjusts deep learning models for optimal execution on end-point target devices.

To use it, you need a pre-trained deep learning model in one of the supported formats: TensorFlow, PyTorch, PaddlePaddle, MXNet, Caffe, Kaldi, or ONNX. Model Optimizer converts the model to the OpenVINO Intermediate Representation format (IR), which you can infer later with OpenVINO™ Runtime.

Note that Model Optimizer does not infer models.

The figure below illustrates the typical workflow for deploying a trained deep learning model:

where IR is a pair of files describing the model:

• .xml - Describes the network topology.

• .bin - Contains the weights and biases binary data.

The generated IR can be additionally optimized for inference by Post-training optimization

that applies post-training quantization methods.

Tip

: You can also work with Model Optimizer in OpenVINO™ Deep Learning Workbench (DL Workbench), which is a web-based tool with GUI for optimizing, fine-tuning, analyzing, visualizing, and comparing performance of deep learning models.

How to Run Model Optimizer

To convert a model to IR, you can run Model Optimizer by using the following command:

mo --input_model INPUT_MODEL

If the out-of-the-box conversion (only the --input_model parameter is specified) is not successful, use the parameters mentioned below to override input shapes and cut the model:

• Model Optimizer provides two parameters to override original input shapes for model conversion: --input and --input_shape. For more information about these parameters, refer to the Setting Input Shapes guide.

• To cut off unwanted parts of a model (such as unsupported operations and training sub-graphs), use the --input and --output parameters to define new inputs and outputs of the converted model. For a more detailed description, refer to the Cutting Off Parts of a Model guide.

You can also insert additional input pre-processing sub-graphs into the converted model by using the --mean_values, scales_values, --layout, and other parameters described in the Embedding Preprocessing Computation article.

The --data_type compression parameter in Model Optimizer allows generating IR of the FP16 data type. For more details, refer to the Compression of a Model to FP16 guide.

To get the full list of conversion parameters available in Model Optimizer, run the following command:

mo --help

Examples of CLI Commands

Below is a list of separate examples for different frameworks and Model Optimizer parameters:

1. Launch Model Optimizer for a TensorFlow MobileNet model in the binary protobuf format:

mo --input_model MobileNet.pb

Launch Model Optimizer for a TensorFlow BERT model in the SavedModel format with three inputs. Specify input shapes explicitly where the batch size and the sequence length equal 2 and 30 respectively:

mo --saved_model_dir BERT --input mask,word_ids,type_ids --input_shape [2,30],[2,30],[2,30]

For more information, refer to the Converting a TensorFlow Model guide.

2. Launch Model Optimizer for an ONNX OCR model and specify new output explicitly:

mo --input_model ocr.onnx --output probabilities

For more information, refer to the [Converting an ONNX Model (prepare_model/convert_model/Convert_Model_From_ONNX.md) guide.

Note

: PyTorch models must be exported to the ONNX format before conversion into IR. More information can be found in Converting a PyTorch Model.

3. Launch Model Optimizer for a PaddlePaddle UNet model and apply mean-scale normalization to the input:

mo --input_model unet.pdmodel --mean_values [123,117,104] --scale 255

For more information, refer to the Converting a PaddlePaddle Model guide.

4. Launch Model Optimizer for an Apache MXNet SSD Inception V3 model and specify first-channel layout for the input:

mo --input_model ssd_inception_v3-0000.params --layout NCHW

For more information, refer to the Converting an Apache MXNet Model guide.

5. Launch Model Optimizer for a Caffe AlexNet model with input channels in the RGB format which needs to be reversed:

mo --input_model alexnet.caffemodel --reverse_input_channels

For more information, refer to the Converting a Caffe Model guide.

6. Launch Model Optimizer for a Kaldi LibriSpeech nnet2 model:

mo --input_model librispeech_nnet2.mdl --input_shape [1,140]

For more information, refer to the Converting a Kaldi Model guide.

• To get conversion recipes for specific TensorFlow, ONNX, PyTorch, Apache MXNet, and Kaldi models, refer to the Model Conversion Tutorials.
• For more information about IR, see Deep Learning Network Intermediate Representation and Operation Sets in OpenVINO™.