"Bad input. 'keyed_values' property is not specified" error when calling "score" method

[damnko]

Hi, I have successfully deployed a model by following this tutorial: https://github.com/pmservice/wml-sample-models/blob/master/tensorflow/import/tensorflow-import-python.ipynb

My model is slightly different because instead of doing the inference with

test_results = session.run( classification, feed_dict = { X : X_test } )

I do it with

tensor_input_ids = self.graph.get_tensor_by_name('input_ids_1:0')
tensor_input_mask = self.graph.get_tensor_by_name('input_mask_1:0')
tensor_label_ids = self.graph.get_tensor_by_name('unique_ids_1:0')
tensor_segment_ids = self.graph.get_tensor_by_name('segment_ids_1:0')
tensor_outputs = [self.graph.get_tensor_by_name('unstack:0'), 
                          self.graph.get_tensor_by_name('unstack:1')]

result = self.sess.run(tensor_outputs, feed_dict={
                tensor_input_ids: np.array(input_ids).reshape(-1, self.max_seq_length).tolist(),
                tensor_input_mask: np.array(input_mask).reshape(-1, self.max_seq_length).tolist(),
                tensor_label_ids: np.array(unique_ids).tolist(),
                tensor_segment_ids: np.array(segment_ids).reshape(-1, self.max_seq_length).tolist(),
            })

so my input of the model is quite different from the one of the jupyter notebook demo.

I have tried to format my input data in different ways, and among those I also tried this mock data

payload = {
    "fields": ["unique_ids", "segment_ids", "input_mask", "input_ids"],
    "values": [['one', 'two', 'three', 'four']]
}

but when I call the function client.deployments.score( model_endpoint_url, payload ) I always get the following error

Status code: 400, body: {"trace": "25d202cc18bf01c0e3de4dc56f090168", "errors": [{"code": "invalid_input_data", "message": "Bad input. 'keyed_values' property is not specified."}], "status_code": 400}

Do you have any idea on why this is happening? I am not able to find documentation on that error anywhere.

Thank you