Hi,
I am using the following code to export resnet18 pth file to onnx file.
I found that running_mean and running_var of batch_norm in onnx are different with those of pth state_dict.
Is there anything wrogn with my exporing code? or is there any reason why running_mean and running_var of batch_norm are not correctly exported?
torch.onnx.export(resnet18, dummy_input, onnx_path, opset_version=11, training=torch.onnx.TrainingMode.TRAINING, do_constant_folding=False)