Pow inference Error

[oneflyingfish]

Describe the issue

We find some error while inference by onnxruntime, such as Pow (onnx). We get Pow(3, 1) is 2.

To reproduce

To reproduce

import onnx
from typing import *
from onnx import helper
from onnx import TensorProto
import numpy as np
import onnxruntime as ort
from onnx import helper, TensorProto


def GetOnnx(shape=[1, 1, 64, 64], save: bool = False) -> onnx.ModelProto:
    # analyse shapes and generate test-onnx
    input1_tensor = helper.make_tensor_value_info(
        'input1_', TensorProto.INT32, shape)
    input2_tensor = helper.make_tensor_value_info(
        'input2_', TensorProto.INT32, shape)
    output_tensor = helper.make_tensor_value_info(
        'output', TensorProto.INT32, shape)

    # create op-Node
    pow_node = helper.make_node('Pow', inputs=['input1_', 'input2_'], outputs=[
                                'output'], name='pow_node')

    # create Graph
    graph_def = helper.make_graph(
        [pow_node],
        'pow_graph',
        [input1_tensor, input2_tensor],
        [output_tensor]
    )

    # create model and set versions
    model_def = helper.make_model(graph_def, producer_name='pow_model', ir_version=8, opset_imports=[
        helper.make_opsetid("", 14)])

    model_def = onnx.shape_inference.infer_shapes(model_def)
    if save:
        onnx.save(model_def, "power.onnx")

    return model_def


def test():
    shape = [6]
    model_def = GetOnnx(shape, True)
    onnx_session = ort.InferenceSession(
        "power.onnx", providers=['CUDAExecutionProvider'])

    # 创建输入数据
    input1 = np.ones(shape, np.int32)*3
    input2 = np.ones(shape, np.int32)

    # 进行推理
    # 进行推理
    output = onnx_session.run(
        None, {'input1_': input1, 'input2_': input2})

    print('Input1:', input1)
    print('Input2:', input2)
    print('output', output[0])


test()

Urgency

No response

Platform

Linux

OS Version

ubuntu 22.04

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.12.0

ONNX Runtime API

Python

Architecture

X64

Execution Provider

CUDA

Execution Provider Library Version

11.8

[yufenglee]

It is easy to repro @pranavsharma.

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.