o Xi, @sdZddlmZddgZddlZddlmZmZddlZ ddl Z ddl m Z m Z mZmZmZmZejr=ddlmZddl Z   d1d2ddZ d3dddddddddd d4d&dZ  d5dddd'd6d/d0ZdS)7z(Convenience constructors for IR objects.) annotationstensornodeN)MappingSequence) _convenience_core_enums _protocolsserdetensor_adaptersvalueKnpt.ArrayLike | onnx.TensorProto | ir.DLPackCompatible | ir.ArrayCompatibledtypeir.DataType | Nonename str | None doc_stringreturn_protocols.TensorProtocolcCs6t|tjr|dur||jkrtd|d|jd|St|tjrMt|}|dur/||_ |dur6||_ |durK||jkrKtd|d|jd|St t |dkr]t j|||d St|tjtjfrotj||||d S|durt|tjstd ||}nTt|tr|std t|trt|tsttj}n8t|trttj}n,t|tr|rtd d|Drttj}ntdd|Drttj}nd}nd}tj||d}t|t st|tr |r tdd|Dr tj j!|dd}tj"|t#|j$||dStj||t#|j$||dS)acCreate a tensor value from an ArrayLike object or a TensorProto. The dtype must match the value. Reinterpretation of the value is not supported, unless if the value is a plain Python object, in which case it is converted to a numpy array with the given dtype. ``value`` can be a numpy array, a plain Python object, or a TensorProto. .. warning:: For 4bit dtypes, the value must be unpacked. Use :class:`~onnx_ir.PackedTensor` to create a tensor with packed data. Example:: >>> import onnx_ir as ir >>> import numpy as np >>> import ml_dtypes >>> import onnx >>> ir.tensor(np.array([1, 2, 3], dtype=np.int16)) Tensor(array([1, 2, 3], dtype=int16), name=None) >>> ir.tensor([1, 2, 3], dtype=ir.DataType.BFLOAT16) Tensor(array([1, 2, 3], dtype=bfloat16), name=None) >>> tp_tensor = ir.tensor(onnx.helper.make_tensor("tensor", onnx.TensorProto.FLOAT, dims=[], vals=[0.5])) >>> tp_tensor.numpy() array(0.5, dtype=float32) >>> import torch >>> ir.tensor(torch.tensor([1.0, 2.0]), name="torch_tensor") TorchTensor(tensor([1., 2.]), name='torch_tensor') Args: value: The numpy array to create the tensor from. dtype: The data type of the tensor. name: The name of the tensor. doc_string: The documentation string of the tensor. Returns: A tensor value. Raises: ValueError: If the dtype does not match the value when value is not a plain Python object like ``list[int]``. Nz=The dtype must match the value when value is a Tensor. dtype=z, value.dtype=z>. You do not have to specify the dtype when value is a Tensor.zBThe dtype must match the value when value is a TensorProto. dtype=z, value.data_type=zAYou do not have to specify the dtype when value is a TensorProto.z)rr)rrrz-dtype must be an instance of DataType. dtype=z8dtype must be specified when value is an empty sequence.css&|]}t|tot|t VqdSN) isinstanceintbool.0elemrV/home/ubuntu/.local/lib/python3.10/site-packages/onnx_ir/_convenience/_constructors.py rs$ztensor..cs|]}t|tVqdSr)rfloatrrrrrt)rcsr r)rstrrrrrrr"zutf-8)encoding)shaperr)rr%rr)%rr TensorProtocolr ValueErroronnx TensorProtor deserialize_tensorrrr#typer TorchTensorDLPackCompatibleArrayCompatiblerTensorr DataType TypeErrornumpyrrrnpint64r!float32allarraystringsencode StringTensorShaper%)r rrrtensor_ numpy_dtyper7rrrrsx 0        ) domainoverload num_outputsoutputsversiongraphrrmetadata_propsop_typer#inputsSequence[ir.Value | None] attributes4Mapping[str, _convenience.SupportedAttrTypes] | Noner?r@rA int | NonerBSequence[ir.Value] | NonerCrDir.Graph | NonerEdict[str, str] | Noneir.Nodec Cs:|durd} nt|} tj|||| |||||| | | d S)aCreate a :class:`~onnx_ir.Node`. This is a convenience constructor for creating a Node that supports Python objects as attributes. Example:: >>> import onnx_ir as ir >>> input_a = ir.val("A", shape=[1, 2], type=ir.TensorType(ir.DataType.INT32)) >>> input_b = ir.val("B", shape=[1, 2], type=ir.TensorType(ir.DataType.INT32)) >>> node = ir.node( ... "SomeOp", ... inputs=[input_a, input_b], ... attributes={"alpha": 1.0, "some_list": [1, 2, 3]}, ... domain="some.domain", ... name="node_name" ... ) >>> node.op_type 'SomeOp' Args: op_type: The name of the operator. inputs: The input values. When an input is None, it is an empty input. attributes: The attributes. RefAttr can be used only when the node is defined in a Function. overload: The overload name when the node is invoking a function. domain: The domain of the operator. For onnx operators, this is an empty string. num_outputs: The number of outputs of the node. If not specified, the number is 1. outputs: The output values. If None, the outputs are created during initialization. version: The version of the operator. If None, the version is unspecified and will follow that of the graph. graph: The graph that the node belongs to. If None, the node is not added to any graph. A `Node` must belong to zero or one graph. name: The name of the node. If None, the node is anonymous. doc_string: The documentation string. metadata_props: The metadata properties. Returns: A node with the given op_type and inputs. Nr) r?rFrGrIr@rArBrCrDrrrE)rconvert_attributesrNode) rFrGrIr?r@rArBrCrDrrrEattrsrrrrs"5 )r+ const_valuerEr%,ir.Shape | Sequence[int | str | None] | Noner+ir.TypeProtocol | NonerSir.TensorProtocol | Noneir.ValuecCs|durXt|j}|dur||krtd|d|d|dur2||jkr2td|d|jd|durJt||jkrJtd|d |jd tj||t|j||d S|dure|duret|}|durtt|tjstt|}tj||||d S) a8Create a :class:`~onnx_ir.Value` with the given name and type. This is a convenience constructor for creating a Value that allows you to specify dtype and shape in a more relaxed manner. Whereas to create a Value directly, you need to create a :class:`~onnx_ir.TypeProtocol` and :class:`~onnx_ir.Shape` object first, this function allows you to specify dtype as a :class:`~onnx_ir.DataType` and shape as a sequence of integers or symbolic dimensions. Example:: >>> import onnx_ir as ir >>> t = ir.val("x", ir.DataType.FLOAT, ["N", 42, 3]) >>> t.name 'x' >>> t.type Tensor(FLOAT) >>> t.shape Shape([SymbolicDim(N), 42, 3]) .. versionadded:: 0.1.9 Args: name: The name of the value. dtype: The data type of the TensorType of the value. This is used only when type is None. shape: The shape of the value. type: The type of the value. Only one of dtype and type can be specified. const_value: The constant tensor that initializes the value. Supply this argument when you want to create an initializer. The type and shape can be obtained from the tensor. metadata_props: The metadata properties that will be serialized to the ONNX proto. Returns: A Value object. Nz.The type does not match the const_value. type=z but const_value has type zC. You do not have to specify the type when const_value is provided.z0The dtype does not match the const_value. dtype=z but const_value has dtype zD. You do not have to specify the dtype when const_value is provided.z0The shape does not match the const_value. shape=z but const_value has shape zD. You do not have to specify the shape when const_value is provided.)rr+r%rSrE)rr+r%rE)r TensorTyperr'r;r%Valuer)rrr%r+rSrEconst_tensor_typerrrvals4*    r[)NNN) r rrrrrrrrrr)rFr#rGrHrIrJr?r#r@r#rArKrBrLrCrKrDrMrrrrrErNrrO)NN)rrrrr%rTr+rUrSrVrErNrrW)__doc__ __future__r__all__typingcollections.abcrrr2r3r(onnx_irrrr r r r TYPE_CHECKING numpy.typingnptirrrr[rrrrsH    |K