Module `furiosa.quantizer.furiosa_sdk_quantizer.frontend.onnx.utils.version_checker`

Expand source code

import onnx

from onnx import version_converter

from furiosa_sdk_quantizer.frontend.onnx import __OPSET_VERSION__
from furiosa_sdk_quantizer.frontend.onnx.utils.check_model import check_model
from furiosa_sdk_quantizer.frontend.onnx.transformer.utils import include_initializer_to_graph_input
from furiosa_sdk_quantizer.interfaces.transformer import Transformer


class CheckVersion(Transformer[onnx.ModelProto]):
    """Convert an ONNX model to ONNX opset 12"""

    def transform(self, model: onnx.ModelProto) -> onnx.ModelProto:
        # https://github.com/onnx/onnx/issues/2873#issuecomment-652541006
        #
        # > There is an underlying issue in version converter. It relies on the
        # > C++ IR which I believe has not been updated after IR v3. Because of
        # > this I think it expects initializers also be added as graph inputs.
        # > If you try to change the version of your model to IRv3 or create a
        # > model with initializers also as inputs then I think this will work.
        model = include_initializer_to_graph_input(model)

        version = int(model.opset_import[0].version)
        if version != __OPSET_VERSION__:
            try:
                model = version_converter.convert_version(model, __OPSET_VERSION__)
                check_model(model)
            except Exception:
                raise Exception(f"Can't convert the model (ONNX opset {version}) to ONNX opset {__OPSET_VERSION__}")
        return model

Classes

class CheckVersion (*args, **kwds)

Convert an ONNX model to ONNX opset 12

Expand source code

class CheckVersion(Transformer[onnx.ModelProto]):
    """Convert an ONNX model to ONNX opset 12"""

    def transform(self, model: onnx.ModelProto) -> onnx.ModelProto:
        # https://github.com/onnx/onnx/issues/2873#issuecomment-652541006
        #
        # > There is an underlying issue in version converter. It relies on the
        # > C++ IR which I believe has not been updated after IR v3. Because of
        # > this I think it expects initializers also be added as graph inputs.
        # > If you try to change the version of your model to IRv3 or create a
        # > model with initializers also as inputs then I think this will work.
        model = include_initializer_to_graph_input(model)

        version = int(model.opset_import[0].version)
        if version != __OPSET_VERSION__:
            try:
                model = version_converter.convert_version(model, __OPSET_VERSION__)
                check_model(model)
            except Exception:
                raise Exception(f"Can't convert the model (ONNX opset {version}) to ONNX opset {__OPSET_VERSION__}")
        return model

Ancestors

furiosa_sdk_quantizer.interfaces.transformer.Transformer
typing.Generic

Methods

def transform(self, model: onnx.onnx_ml_pb2.ModelProto) ‑> onnx.onnx_ml_pb2.ModelProto

Expand source code

def transform(self, model: onnx.ModelProto) -> onnx.ModelProto:
    # https://github.com/onnx/onnx/issues/2873#issuecomment-652541006
    #
    # > There is an underlying issue in version converter. It relies on the
    # > C++ IR which I believe has not been updated after IR v3. Because of
    # > this I think it expects initializers also be added as graph inputs.
    # > If you try to change the version of your model to IRv3 or create a
    # > model with initializers also as inputs then I think this will work.
    model = include_initializer_to_graph_input(model)

    version = int(model.opset_import[0].version)
    if version != __OPSET_VERSION__:
        try:
            model = version_converter.convert_version(model, __OPSET_VERSION__)
            check_model(model)
        except Exception:
            raise Exception(f"Can't convert the model (ONNX opset {version}) to ONNX opset {__OPSET_VERSION__}")
    return model