The output size issue of the ONNX exported model

[jgy0]

Why does the output of the four types of detectors I trained, after exporting to ONNX, result in [1, 8, 33600], which does not conform to 1: batch size 9: 4(bbox) + 1(obj_conf) + 4(cls_conf) = 9?

for example

from ultralytics import YOLO
model = YOLO('/home/guangyao/code/yolov13/runs/detect/echinoderm43/weights/best.pt')
print('Model classes:', model.names)

output is Model classes: {0: 'starfish', 1: 'holothurian', 2: 'echinus', 3: 'scallop'}

import onnx
model = onnx.load('/home/guangyao/code/yolov13/runs/detect/echinoderm43/weights/best.onnx')
for i, output in enumerate(model.graph.output):
shape = [dim.dim_value for dim in output.type.tensor_type.shape.dim]
print(f'Output {i}: {shape}')

Output 0: [1, 8, 33600]

why?what's mean 8 and 33600