OpenVINO™ Model conversion

This Jupyter notebook can be launched on-line, opening an interactive environment in a browser window. You can also make a local installation. Choose one of the following options:

BinderGoogle ColabGithub

This notebook shows how to convert a model from original framework format to OpenVINO Intermediate Representation (IR).

Table of contents:

# Required imports. Please execute this cell first.
%pip install --upgrade pip
%pip install -q --extra-index-url https://download.pytorch.org/whl/cpu \
"openvino-dev>=2023.2.0" "requests" "tqdm" "transformers[onnx]>=4.21.1" "torch" "torchvision" "tensorflow_hub" "tensorflow"

Requirement already satisfied: pip in /opt/home/k8sworker/ci-ai/cibuilds/ov-notebook/OVNotebookOps-609/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages (24.0)

Note: you may need to restart the kernel to use updated packages.

Note: you may need to restart the kernel to use updated packages.

OpenVINO IR format

OpenVINO Intermediate Representation (IR) is the proprietary model format of OpenVINO. It is produced after converting a model with model conversion API. Model conversion API translates the frequently used deep learning operations to their respective similar representation in OpenVINO and tunes them with the associated weights and biases from the trained model. The resulting IR contains two files: an .xml file, containing information about network topology, and a .bin file, containing the weights and biases binary data.

There are two ways to convert a model from the original framework format to OpenVINO IR: Python conversion API and OVC command-line tool. You can choose one of them based on whichever is most convenient for you.

OpenVINO conversion API supports next model formats: PyTorch, TensorFlow, TensorFlow Lite, ONNX, and PaddlePaddle. These model formats can be read, compiled, and converted to OpenVINO IR, either automatically or explicitly.

For more details, refer to Model Preparation documentation.

# OVC CLI tool parameters description

! ovc --help

usage: ovc INPUT_MODEL... [-h] [--output_model OUTPUT_MODEL]
           [--compress_to_fp16 [True | False]] [--version] [--input INPUT]
           [--output OUTPUT] [--extension EXTENSION] [--verbose]

positional arguments:
  INPUT_MODEL           Input model file(s) from TensorFlow, ONNX,
                        PaddlePaddle. Use openvino.convert_model in Python to
                        convert models from PyTorch.

optional arguments:
  -h, --help            show this help message and exit
  --output_model OUTPUT_MODEL
                        This parameter is used to name output .xml/.bin files
                        of converted model. Model name or output directory can
                        be passed. If output directory is passed, the
                        resulting .xml/.bin files are named by original model
                        name.
  --compress_to_fp16 [True | False]
                        Compress weights in output OpenVINO model to FP16. To
                        turn off compression use "--compress_to_fp16=False"
                        command line parameter. Default value is True.
  --version             Print ovc version and exit.
  --input INPUT         Information of model input required for model
                        conversion. This is a comma separated list with
                        optional input names and shapes. The order of inputs
                        in converted model will match the order of specified
                        inputs. The shape is specified as comma-separated
                        list. Example, to set input_1 input with shape
                        [1,100] and sequence_len input with shape [1,?]:
                        "input_1[1,100],sequence_len[1,?]", where "?" is a
                        dynamic dimension, which means that such a dimension
                        can be specified later in the runtime. If the
                        dimension is set as an integer (like 100 in [1,100]),
                        such a dimension is not supposed to be changed later,
                        during a model conversion it is treated as a static
                        value. Example with unnamed inputs: "[1,100],[1,?]".
  --output OUTPUT       One or more comma-separated model outputs to be
                        preserved in the converted model. Other outputs are
                        removed. If output parameter is not specified then
                        all outputs from the original model are preserved. Do
                        not add :0 to the names for TensorFlow. The order of
                        outputs in the converted model is the same as the
                        order of specified names. Example: ovc model.onnx
                        output=out_1,out_2
  --extension EXTENSION
                        Paths or a comma-separated list of paths to libraries
                        (.so or .dll) with extensions. To disable all
                        extensions including those that are placed at the
                        default location, pass an empty string.
  --verbose             Print detailed information about conversion.

Fetching example models

This notebook uses two models for conversion examples:

  • Distilbert NLP model from Hugging Face

  • Resnet50 CV classification model from torchvision

from pathlib import Path

# create a directory for models files
MODEL_DIRECTORY_PATH = Path("model")
MODEL_DIRECTORY_PATH.mkdir(exist_ok=True)

Fetch distilbert NLP model from Hugging Face and export it in ONNX format:

from transformers import AutoModelForSequenceClassification, AutoTokenizer
from transformers.onnx import export, FeaturesManager

ONNX_NLP_MODEL_PATH = MODEL_DIRECTORY_PATH / "distilbert.onnx"

# download model
hf_model = AutoModelForSequenceClassification.from_pretrained(
    "distilbert-base-uncased-finetuned-sst-2-english"
)
# initialize tokenizer
tokenizer = AutoTokenizer.from_pretrained(
    "distilbert-base-uncased-finetuned-sst-2-english"
)

# get model onnx config function for output feature format sequence-classification
model_kind, model_onnx_config = FeaturesManager.check_supported_model_or_raise(
    hf_model, feature="sequence-classification"
)
# fill onnx config based on pytorch model config
onnx_config = model_onnx_config(hf_model.config)

# export to onnx format
export(
    preprocessor=tokenizer,
    model=hf_model,
    config=onnx_config,
    opset=onnx_config.default_onnx_opset,
    output=ONNX_NLP_MODEL_PATH,
)

2024-02-09 23:07:37.212192: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
2024-02-09 23:07:37.247733: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-02-09 23:07:37.883428: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

/opt/home/k8sworker/ci-ai/cibuilds/ov-notebook/OVNotebookOps-609/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/transformers/models/distilbert/modeling_distilbert.py:246: TracerWarning: torch.tensor results are registered as constants in the trace. You can safely ignore this warning if you use this function to create tensors out of constant variables that would be the same every time you call this function. In any other case, this might cause the trace to be incorrect.
  mask, torch.tensor(torch.finfo(scores.dtype).min)

(['input_ids', 'attention_mask'], ['logits'])

Fetch Resnet50 CV classification model from torchvision:

from torchvision.models import resnet50, ResNet50_Weights

# create model object
pytorch_model = resnet50(weights=ResNet50_Weights.DEFAULT)
# switch model from training to inference mode
pytorch_model.eval()

ResNet(
  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (relu): ReLU(inplace=True)
  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  (layer1): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer2): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer3): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (4): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer4): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc): Linear(in_features=2048, out_features=1000, bias=True)
)

Convert PyTorch model to ONNX format:

import torch
import warnings

ONNX_CV_MODEL_PATH = MODEL_DIRECTORY_PATH / "resnet.onnx"

if ONNX_CV_MODEL_PATH.exists():
    print(f"ONNX model {ONNX_CV_MODEL_PATH} already exists.")
else:
    with warnings.catch_warnings():
        warnings.filterwarnings("ignore")
        torch.onnx.export(
            model=pytorch_model, args=torch.randn(1, 3, 224, 224), f=ONNX_CV_MODEL_PATH
        )
    print(f"ONNX model exported to {ONNX_CV_MODEL_PATH}")

ONNX model exported to model/resnet.onnx

Conversion

To convert a model to OpenVINO IR, use the following API:

import openvino as ov

# ov.convert_model returns an openvino.runtime.Model object
print(ONNX_NLP_MODEL_PATH)
ov_model = ov.convert_model(ONNX_NLP_MODEL_PATH)

# then model can be serialized to *.xml & *.bin files
ov.save_model(ov_model, MODEL_DIRECTORY_PATH / "distilbert.xml")

model/distilbert.onnx

! ovc model/distilbert.onnx --output_model model/distilbert.xml

huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
To disable this warning, you can either:
    - Avoid using tokenizers before the fork if possible
    - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
[ INFO ] Generated IR will be compressed to FP16. If you get lower accuracy, please consider disabling compression by removing argument "compress_to_fp16" or set it to false "compress_to_fp16=False".
Find more information about compression to FP16 at https://docs.openvino.ai/2023.0/openvino_docs_MO_DG_FP16_Compression.html

[ SUCCESS ] XML file: model/distilbert.xml
[ SUCCESS ] BIN file: model/distilbert.bin

Setting Input Shapes

Model conversion is supported for models with dynamic input shapes that contain undefined dimensions. However, if the shape of data is not going to change from one inference request to another, it is recommended to set up static shapes (when all dimensions are fully defined) for the inputs. Doing so at the model preparation stage, not at runtime, can be beneficial in terms of performance and memory consumption.

For more information refer to Setting Input Shapes documentation.

import openvino as ov

ov_model = ov.convert_model(
    ONNX_NLP_MODEL_PATH, input=[("input_ids", [1, 128]), ("attention_mask", [1, 128])]
)

! ovc model/distilbert.onnx --input input_ids[1,128],attention_mask[1,128] --output_model model/distilbert.xml

huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
To disable this warning, you can either:
    - Avoid using tokenizers before the fork if possible
    - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
[ INFO ] Generated IR will be compressed to FP16. If you get lower accuracy, please consider disabling compression by removing argument "compress_to_fp16" or set it to false "compress_to_fp16=False".
Find more information about compression to FP16 at https://docs.openvino.ai/2023.0/openvino_docs_MO_DG_FP16_Compression.html

[ SUCCESS ] XML file: model/distilbert.xml
[ SUCCESS ] BIN file: model/distilbert.bin

The input parameter allows overriding original input shapes if it is supported by the model topology. Shapes with dynamic dimensions in the original model can be replaced with static shapes for the converted model, and vice versa. The dynamic dimension can be marked in model conversion API parameter as -1 or ? when using ovc:

import openvino as ov

ov_model = ov.convert_model(
    ONNX_NLP_MODEL_PATH, input=[("input_ids", [1, -1]), ("attention_mask", [1, -1])]
)

! ovc model/distilbert.onnx --input "input_ids[1,?],attention_mask[1,?]" --output_model model/distilbert.xml

huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
To disable this warning, you can either:
    - Avoid using tokenizers before the fork if possible
    - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
[ INFO ] Generated IR will be compressed to FP16. If you get lower accuracy, please consider disabling compression by removing argument "compress_to_fp16" or set it to false "compress_to_fp16=False".
Find more information about compression to FP16 at https://docs.openvino.ai/2023.0/openvino_docs_MO_DG_FP16_Compression.html

[ SUCCESS ] XML file: model/distilbert.xml
[ SUCCESS ] BIN file: model/distilbert.bin

To optimize memory consumption for models with undefined dimensions in runtime, model conversion API provides the capability to define boundaries of dimensions. The boundaries of undefined dimension can be specified with ellipsis in the command line or with openvino.Dimension class in Python. For example, launch model conversion for the ONNX Bert model and specify a boundary for the sequence length dimension:

import openvino as ov


sequence_length_dim = ov.Dimension(10, 128)

ov_model = ov.convert_model(
    ONNX_NLP_MODEL_PATH, input=[("input_ids", [1, sequence_length_dim]), ("attention_mask", [1, sequence_length_dim])]
)

! ovc model/distilbert.onnx --input input_ids[1,10..128],attention_mask[1,10..128] --output_model model/distilbert.xml

huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
To disable this warning, you can either:
    - Avoid using tokenizers before the fork if possible
    - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
[ INFO ] Generated IR will be compressed to FP16. If you get lower accuracy, please consider disabling compression by removing argument "compress_to_fp16" or set it to false "compress_to_fp16=False".
Find more information about compression to FP16 at https://docs.openvino.ai/2023.0/openvino_docs_MO_DG_FP16_Compression.html

[ SUCCESS ] XML file: model/distilbert.xml
[ SUCCESS ] BIN file: model/distilbert.bin

Compressing a Model to FP16

By default model weights compressed to FP16 format when saving OpenVINO model to IR. This saves up to 2x storage space for the model file and in most cases doesn’t sacrifice model accuracy. Weight compression can be disabled by setting compress_to_fp16 flag to False:

import openvino as ov

ov_model = ov.convert_model(ONNX_NLP_MODEL_PATH)
ov.save_model(ov_model, MODEL_DIRECTORY_PATH / 'distilbert.xml', compress_to_fp16=False)

! ovc model/distilbert.onnx --output_model model/distilbert.xml --compress_to_fp16=False

huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
To disable this warning, you can either:
    - Avoid using tokenizers before the fork if possible
    - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
[ SUCCESS ] XML file: model/distilbert.xml
[ SUCCESS ] BIN file: model/distilbert.bin

Convert Models from memory

Model conversion API supports passing original framework Python object directly. More details can be found in PyTorch, TensorFlow, PaddlePaddle frameworks conversion guides.

import openvino as ov
import torch

example_input = torch.rand(1, 3, 224, 224)

ov_model = ov.convert_model(pytorch_model, example_input=example_input, input=example_input.shape)

WARNING:tensorflow:Please fix your imports. Module tensorflow.python.training.tracking.base has been moved to tensorflow.python.trackable.base. The old module will be deleted in version 2.11.

import openvino as ov
import tensorflow_hub as hub

model = hub.load("https://www.kaggle.com/models/google/movenet/frameworks/TensorFlow2/variations/singlepose-lightning/versions/4")
movenet = model.signatures['serving_default']

ov_model = ov.convert_model(movenet)

2024-02-09 23:07:56.008045: E tensorflow/compiler/xla/stream_executor/cuda/cuda_driver.cc:266] failed call to cuInit: CUDA_ERROR_COMPAT_NOT_SUPPORTED_ON_DEVICE: forward compatibility was attempted on non supported HW
2024-02-09 23:07:56.008076: I tensorflow/compiler/xla/stream_executor/cuda/cuda_diagnostics.cc:168] retrieving CUDA diagnostic information for host: iotg-dev-workstation-07
2024-02-09 23:07:56.008081: I tensorflow/compiler/xla/stream_executor/cuda/cuda_diagnostics.cc:175] hostname: iotg-dev-workstation-07
2024-02-09 23:07:56.008267: I tensorflow/compiler/xla/stream_executor/cuda/cuda_diagnostics.cc:199] libcuda reported version is: 470.223.2
2024-02-09 23:07:56.008284: I tensorflow/compiler/xla/stream_executor/cuda/cuda_diagnostics.cc:203] kernel reported version is: 470.182.3
2024-02-09 23:07:56.008287: E tensorflow/compiler/xla/stream_executor/cuda/cuda_diagnostics.cc:312] kernel version 470.182.3 does not match DSO version 470.223.2 -- cannot find working devices in this configuration

Migration from Legacy conversion API

In the 2023.1 OpenVINO release OpenVINO Model Conversion API was introduced with the corresponding Python API: openvino.convert_model method. ovc and openvino.convert_model represent a lightweight alternative of mo and openvino.tools.mo.convert_model which are considered legacy API now. mo.convert_model() provides a wide range of preprocessing parameters. Most of these parameters have analogs in OVC or can be replaced with functionality from ov.PrePostProcessor class. Refer to Optimize Preprocessing notebook for more information about Preprocessing API. Here is the migration guide from legacy model preprocessing to Preprocessing API.

Specifying Layout

Layout defines the meaning of dimensions in a shape and can be specified for both inputs and outputs. Some preprocessing requires to set input layouts, for example, setting a batch, applying mean or scales, and reversing input channels (BGR<->RGB). For the layout syntax, check the Layout API overview. To specify the layout, you can use the layout option followed by the layout value.

The following example specifies the NCHW layout for a Pytorch Resnet50 model that was exported to the ONNX format:

# Converter API
import openvino as ov

ov_model = ov.convert_model(ONNX_CV_MODEL_PATH)

prep = ov.preprocess.PrePostProcessor(ov_model)
prep.input('input.1').model().set_layout(ov.Layout("nchw"))
ov_model = prep.build()

# Legacy Model Optimizer API
from openvino.tools import mo

ov_model = mo.convert_model(ONNX_CV_MODEL_PATH, layout="nchw")

huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
To disable this warning, you can either:
    - Avoid using tokenizers before the fork if possible
    - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)

Changing Model Layout

Transposing of matrices/tensors is a typical operation in Deep Learning - you may have a BMP image 640x480, which is an array of {480, 640, 3} elements, but Deep Learning model can require input with shape {1, 3, 480, 640}.

Conversion can be done implicitly, using the layout of a user’s tensor and the layout of an original model:

# Converter API
import openvino as ov

ov_model = ov.convert_model(ONNX_CV_MODEL_PATH)

prep = ov.preprocess.PrePostProcessor(ov_model)
prep.input('input.1').tensor().set_layout(ov.Layout("nhwc"))
prep.input('input.1').model().set_layout(ov.Layout("nchw"))
ov_model = prep.build()

# Legacy Model Optimizer API
from openvino.tools import mo

ov_model = mo.convert_model(ONNX_CV_MODEL_PATH, layout="nchw->nhwc")

# alternatively use source_layout and target_layout parameters
ov_model = mo.convert_model(
    ONNX_CV_MODEL_PATH, source_layout="nchw", target_layout="nhwc"
)

Specifying Mean and Scale Values

Using Preprocessing API mean and scale values can be set. Using these API, model embeds the corresponding preprocessing block for mean-value normalization of the input data and optimizes this block. Refer to Optimize Preprocessing notebook for more examples.

# Converter API
import openvino as ov

ov_model = ov.convert_model(ONNX_CV_MODEL_PATH)

prep = ov.preprocess.PrePostProcessor(ov_model)
prep.input("input.1").tensor().set_layout(ov.Layout("nchw"))
prep.input("input.1").preprocess().mean([255 * x for x in [0.485, 0.456, 0.406]])
prep.input("input.1").preprocess().scale([255 * x for x in [0.229, 0.224, 0.225]])

ov_model = prep.build()

# Legacy Model Optimizer API
from openvino.tools import mo


ov_model = mo.convert_model(
    ONNX_CV_MODEL_PATH,
    mean_values=[255 * x for x in [0.485, 0.456, 0.406]],
    scale_values=[255 * x for x in [0.229, 0.224, 0.225]],
)

Reversing Input Channels

Sometimes, input images for your application can be of the RGB (or BGR) format, and the model is trained on images of the BGR (or RGB) format, which is in the opposite order of color channels. In this case, it is important to preprocess the input images by reverting the color channels before inference.

# Converter API
import openvino as ov

ov_model = ov.convert_model(ONNX_CV_MODEL_PATH)

prep = ov.preprocess.PrePostProcessor(ov_model)
prep.input('input.1').tensor().set_layout(ov.Layout("nchw"))
prep.input('input.1').preprocess().reverse_channels()
ov_model = prep.build()

# Legacy Model Optimizer API
from openvino.tools import mo

ov_model = mo.convert_model(ONNX_CV_MODEL_PATH, reverse_input_channels=True)

Cutting Off Parts of a Model

Cutting model inputs and outputs from a model is no longer available in the new conversion API. Instead, we recommend performing the cut in the original framework. Examples of model cutting of TensorFlow protobuf, TensorFlow SavedModel, and ONNX formats with tools provided by the Tensorflow and ONNX frameworks can be found in documentation guide. For PyTorch, TensorFlow 2 Keras, and PaddlePaddle, we recommend changing the original model code to perform the model cut.