Resnet batch prediction

My apologies if I am missing something obvious.
I wanted to classify batch of 10 images using resnet34v2.onnx downloaded [here].(https://github.com/onnx/models/tree/master/vision/classification/resnet)
I have simplified the example so instead of images lets classify zeros.

batch_size = 10
arr = np.zeros((batch_size, 3, 224, 224), dtype=np.float32)
sess = rt.InferenceSession("resnet34v2.onnx")
input_name = sess.get_inputs()[0].name
result = sess.run(None, {input_name: arr })

The error produced.

onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Got invalid dimensions for input: data for the following indices
 index: 0 Got: 10 Expected: 1

From this post I gathered that you can specify the input shape during conversion. In this case it should have been (None, 3, 224, 224), but was (1, 3, 224, 224).
Any ideas how to allow batch prediction for already converted model?

Start by running the provided Python reproduction with the linked resnet34v2.onnx model and inspect its declared input shape. Investigate how the model was converted and whether this repository contains the source or conversion process. Done means establishing a supported way for the model to accept a batch of 10, or clearly documenting that the existing model cannot do so.