AffineLayoutProto#

Describes built-in affine encoding with tensor-valued scales and zero points. See Quantizes tensors into encoded values for encoded value representations.

class onnx_light.onnx.AffineLayoutProto(*args, **kwargs)#

Describes the built-in affine (linear) quantization layout of an EncodedValueProto. The branch is deliberately closed: other affine forms use a structured layout instead of extending this message.

ByteSize(self) → int#: Returns the serialized size in bytes, following the protobuf API.

Clear(self) → None#: Clears the object.

ClearField(self, field_name: str) → None#: Clears the field field_name, following the protobuf API.

CopyFrom(self, arg: onnx_light.onnx_py._onnxpyprotoop.AffineLayoutProto, /) → None#: Copies one instance into this one.

HasField(self, field_name: str) → bool#: Checks if a field is set, following the protobuf HasField API.

ParseFromFile(self, name: str, options: object | None = None, external_data_file: str = '') → None#: Parses a binary file to fill this instance.

ParseFromString(self, data: bytes, options: object | None = None) → None#

ParseFromString(self, data: str, options: object | None = None) → None

Overloaded function.

ParseFromString(self, data: bytes, options: object | None = None) -> None

Parses a sequence of bytes to fill this instance.

ParseFromString(self, data: str, options: object | None = None) -> None

Parses a string to fill this instance.

SerializeSize(self, options: object | None = None) → onnx_light.onnx_py._onnxpyprotoop.SerializeSizeResult#: Returns the size once serialized without serializing.

SerializeToFile(self, name: str, options: object | None = None, external_data_file: str = '') → None#: Serializes this instance into a file. If external_data_size is not empty, big weights are stored in this (depending on options.raw_data_threshold). When writing to two files, temporary external-data metadata is cleared so the in-memory model stays unchanged.

SerializeToFileDescriptor(self, fd: int, options: object | None = None) → None#: Serializes this instance into an open file descriptor without closing it.

SerializeToOStream(self, output: object) → None#: Alias for SerializeToOstream (capital S) for protobuf API compatibility. Serializes this instance to a Python file-like object that has a write() method.

SerializeToOstream(self, output: object) → None#: Serializes this instance to a Python file-like object that has a write() method.

SerializeToString(self, options: object | None = None) → bytes#: Serializes this instance into a sequence of bytes.

add_scale(self) → onnx_light.onnx_py._onnxpyprotoop.TensorProto#: Sets an empty value.

add_zero_point(self) → onnx_light.onnx_py._onnxpyprotoop.TensorProto#: Sets an empty value.

property axis#: Omitting axis selects per-tensor quantization; setting it selects per-axis parameters.

property block_size#: Blocked quantization along axis. It is valid only together with axis.

has_axis(self) → bool#: Tells if ‘axis’ has a value.

has_block_size(self) → bool#: Tells if ‘block_size’ has a value.

has_scale(self) → bool#: Tells if ‘scale’ has a value.

has_signed_storage(self) → bool#: Tells if ‘signed_storage’ has a value.

has_storage_type(self) → bool#

has_zero_point(self) → bool#: Tells if ‘zero_point’ has a value.

property scale#: Scalar or parameter tensor with a floating element type. This field MUST be present for a valid affine layout.

property signed_storage#

0 is unsigned and 1 is signed. The code width is inferred from the logical shape and payload size and must identify exactly one supported width.

Type:: Signedness of stored integer codes when storage_type is absent

property storage_type#: Element type of the stored codes. It MUST be one of INT8, UINT8, INT4, UINT4, INT2 or UINT2 and the payload holds only row-major codes packed like TensorProto.raw_data. Exactly one of storage_type and signed_storage MUST be present. When storage_type is absent, signed_storage determines signedness and the code width is inferred from the logical element count and payload byte length; the payload MUST identify exactly one of the supported 2-bit, 4-bit or 8-bit widths. N is the product of the concrete dimensions in EncodedValueProto.logical_type.tensor_type.shape. B is either EncodedValueProto.raw_data.size() for an inline payload or the numeric ‘length’ entry in EncodedValueProto.external_data when data_location is EXTERNAL. Two-bit storage matches when B == ceil(N / 4) == N / 4 + (N % 4 != 0), 4-bit storage matches when B == ceil(N / 2) == N / 2 + N % 2, and 8-bit storage matches when B == N. Exactly one equality MUST hold; signed_storage then selects the signed INT2, INT4 or INT8 type when equal to 1, and the corresponding unsigned type otherwise.

property zero_point#: Optional zero point with element type storage_type. It defaults to zero.