Compile pytorch complexYOLOv3 model to MLIR failed

Nobody has claimed this yet.

Hi, I am trying to use torch-mlir to compile my model (ComplexYOLOv3) -> https://github.com/ghimiredhikura/Complex-YOLOv3

I can get right result if i use default code test_detection.py in github

Here is my compile code

`
import numpy as np
import math
import os
import argparse
import cv2
import time
import torch

import utils.utils as utils
from models import *
import torch.utils.data as torch_data

import utils.kitti_utils as kitti_utils
import utils.kitti_aug_utils as aug_utils
import utils.kitti_bev_utils as bev_utils
from utils.kitti_yolo_dataset import KittiYOLODataset
import utils.config as cnf
import utils.mayavi_viewer as mview
from torch_mlir import fx

if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--model_def", type=str, default="config/complex_tiny_yolov3.cfg", help="path to model definition file")
parser.add_argument("--weights_path", type=str, default="checkpoints/tiny-yolov3_ckpt_epoch-220.pth", help="path to weights file")
parser.add_argument("--class_path", type=str, default="data/classes.names", help="path to class label file")
parser.add_argument("--conf_thres", type=float, default=0.5, help="object confidence threshold")
parser.add_argument("--nms_thres", type=float, default=0.5, help="iou thresshold for non-maximum suppression")
parser.add_argument("--img_size", type=int, default=cnf.BEV_WIDTH, help="size of each image dimension")
parser.add_argument("--split", type=str, default="valid", help="text file having image lists in dataset")
parser.add_argument("--folder", type=str, default="training", help="directory name that you downloaded all dataset")
opt = parser.parse_args()
print(opt)

classes = utils.load_classes(opt.class_path)
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device = "cpu"

print("Device: ", device)
# Set up model
model = Darknet(opt.model_def, img_size=opt.img_size).to(device)
# Load checkpoint weights
model.load_state_dict(torch.load(opt.weights_path))
# Eval mode

with torch.no_grad():
model.eval()

dummy_input = torch.randn(1, 3, 1024, 1024)

print("Exporting to torch-MLIR...")
m = fx.export_and_import(model, dummy_input, output_type="torch")
print("✓ Successfully exported to TOSA!")

# Save MLIR to file
output_file = "yolov3_tosa.mlir"
print(f"\nSaving MLIR to {output_file}...")
with open(output_file, "w") as f:
f.write(str(m))
print(f"✓ Successfully saved to {output_file}!")
`

Follow is Error Log
`(base) root@16db5de91610:/workspace# python test_detection.py --split=sample --folder=sampledata
/opt/conda/lib/python3.11/site-packages/traits/etsconfig/etsconfig.py:425: UserWarning: Environment variable "HOME" not set, setting home directory to /tmp
warn(
Namespace(model_def='config/complex_tiny_yolov3.cfg', weights_path='checkpoints/tiny-yolov3_ckpt_epoch-220.pth', class_path='data/classes.names', conf_thres=0.5, nms_thres=0.5, img_size=608, split='sample', folder='sampledata')
Device: cpu
Exporting to torch-MLIR...
Segmentation fault (core dumped)`

if the error log is more detail, i will try debugging but just say to me "segmentation fault"

Environment:
python 3.11.14
torch 2.9.1+cu126
torch-mlir 20260127.705
torchaudio 2.9.1+cu126
torchelastic 0.2.2
torchvision 0.24.1+cu126

Thank you

Reproduce the command in test_detection.py with the reported Python, PyTorch, and torch-mlir versions, then narrow the crash to the model setup or the fx.export_and_import(model, dummy_input, output_type="torch") call. Use the reported ComplexYOLOv3 configuration, checkpoint, and 1024x1024 dummy input; done means identifying a reproducible failing operation or obtaining a non-crashing export with actionable diagnostics.

Written by the indexing model from the issue text.