Convert DAB-deformable-DETR to ONNX

[sazani]

I am trying to convert the generated model that I trained and also your pretrained model to ONNX but unfortunately I faced the following error message:

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0, and cpu! (when checking argument for argument index in method wrapper__index_select)
By the way I have used static and dynamic input and I have used the following code:

import torch.onnx
import os, sys
import torch
import numpy as np

from models import build_dab_deformable_detr
from util.slconfig import SLConfig
import torchvision
import torchvision.transforms.functional as TF

from PIL import Image
import transforms as T

import cv2
import argparse
device = torch.device('cuda:0' )

if name == "main":
parser = argparse.ArgumentParser()
parser.add_argument('--model_checkpoint_path', help="change the path of the model checkpoint.",
default="./Checkpoints/checkpoint.pt")
parser.add_argument('--model_config_path', help="change the path of the model config file",
default="./Checkpoints/config.json")
args = parser.parse_args()
model_config_path = args.model_config_path
model_checkpoint_path = args.model_checkpoint_path
args_config = SLConfig.fromfile(model_config_path)
model, criterion, post_processors = build_dab_deformable_detr(args_config)
checkpoint = torch.load(model_checkpoint_path, map_location=device)
model.load_state_dict(checkpoint['model'])
model = model.to(device)
img_size =[1080,1920]
input = torch.zeros(1, 3, *img_size)
input = input.to(device)
model.eval()
results =model(input)
torch.onnx.export(
model,
input,
"test.onnx",
input_names=["input"],
output_names=["output"],
export_params=True,
opset_version=11, # I have also tried version 12,13,14,15
# dynamic_axes={'images': {0: 'batch', 2: 'height', 3: 'width'}, # shape(1,3,640,640)
# 'output': {0: 'batch', 1: 'anchors'} # shape(1,25200,85)
# } ,#if dynamic else None
dynamic_axes = None,
)

[Zalways]

have you solve this problem? i try to export the model into torchscript,(device is cpu),i met some problem: I finally got the model through the trace method, but there was a problem with this model,and it cann't work!
RuntimeError: The size of tensor a (32) must match the size of tensor b (237) at non-singleton dimension 1
i need your help!
so have you met the error like me?appreciate if you could give me some advice!thanks!