Could not find an implementation for Conv for opset 11 and data type "double"

### Describe the issue

Can not run import model (exported by pytorch.onnx.export) by onnx run time method ort.InferenceSession.

Below error is generated:


### To reproduce

The code to export pytorch model:

import torch
import network.herbs.timm as timm
from network.herbs.models.pim_module.pim_module import PluginMoodel
from export_config import export_config_file

def export_herbs_onnx(input_model_path: str, output_model_path: str):
inputs = (1, 3, 384, 384)

num_selects = {'layer1': 256, 'layer2': 128, 'layer3': 64, 'layer4': 32}
num_class = 200

backbone = timm.create_model('swin_large_patch4_window12_384_in22k', pretrained=False)
model = PluginMoodel(backbone=backbone,
return_nodes=None,
img_size=384,
use_fpn=True,
fpn_size=1536,
proj_type="Linear",
upsample_type="Conv",
use_selection=True,
num_classes=num_class,
num_selects=num_selects,
use_combiner=True,
comb_proj_size=None)

checkpoint = torch.load(input_model_path, map_location=torch.device('cpu'))
model.load_state_dict(checkpoint['model'])
model.eval()

output_names = ['FPN1_layer4', 'FPN1_layer3', 'FPN1_layer2', 'FPN1_layer1', 'layer1', 'layer2', 'layer3', 'layer4',
'select_layer1', 'drop_layer1', 'select_layer2', 'drop_layer2', 'select_layer3', 'drop_layer3',
'select_layer3', 'drop_layer3',
'comb_outs']

model.double()
inputs = inputs.double()

torch.onnx.export(model, inputs, output_model_path,
input_names=["input"], output_names=output_names,
export_params=True,
opset_version=15)

--------------------------------------------------------------------------------------

The code to import onnx model:

import onnxruntime as ort

sess = ort.InferenceSession(model_path,providers=['CPUExecutionProvider'])

### Urgency

_No response_

### Platform

Windows

### OS Version

windows 11

### ONNX Runtime Installation

Built from Source

### ONNX Runtime Version or Commit ID

onnxruntime-1.17.1

### ONNX Runtime API

Python

### Architecture

X64

### Execution Provider

Default CPU

### Execution Provider Library Version

_No response_

Start by running the supplied PyTorch export and ort.InferenceSession reproduction with ONNX Runtime 1.17.1 and the CPU provider, checking the exported model's actual opset and Conv data type. The screenshot contains the only detailed error, so confirm the failure from a minimal model; done means determining whether Conv with double is supported and adding the appropriate fix or regression coverage if a code change is required.

Written by the indexing model from the issue text.