mask rcnn model change onnx model that no had ops.caffe2.op(any) ,what shuld I do ?

[XUDINGYI312]

I have build model

torch_model = Trainer.build_model(cfg)
DetectionCheckpointer(torch_model).resume_or_load(cfg.MODEL.WEIGHTS)
torch_model.eval()

data_loader = build_detection_test_loader(cfg, cfg.DATASETS.TEST[0])
first_batch = next(iter(data_loader))
with torch.no_grad():
torch.onnx.export(
torch_model,
(first_batch ,),
"my.onnx",
# operator_export_type=OperatorExportTypes.ONNX,
#verbose=True, # NOTE: uncomment this for debugging
export_params=True,
training=False,
#keep_initializers_as_inputs=True,
)

the error is that :
RuntimeError: Only tuples, lists and Variables supported as JIT inputs/outputs. Dictionaries and strings are also accepted but their usage is not recommended. But got unsupported type ImageList
what shuld I do that how to change input ?