tensorflow: [tflite][java] has_rank_one_input_condition error when use runForMultipleInputsOutputs()

I obtain an error when use two inputs in runForMultipleInputsOutputs().

System information

• Custom code

• Android API 29

• Android simulator

• Used libraries: – org.tensorflow:tensorflow-lite:2.1.0 – org.tensorflow:tensorflow-lite-gpu:2.1.0 – org.tensorflow:tensorflow-lite-support:0.0.0-nightly

• Used model information The model tolite file

[{‘name’: ‘input_1’, ‘index’: 0, ‘shape’: array([ 1, 10], dtype=int32), ‘dtype’: <class ‘numpy.float32’>, ‘quantization’: (0.0, 0)}, {‘name’: ‘input_2’, ‘index’: 1, ‘shape’: array([1, 2], dtype=int32), ‘dtype’: <class ‘numpy.float32’>, ‘quantization’: (0.0, 0)}]

__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_1 (InputLayer)            [(None, 10)]         0                                            
__________________________________________________________________________________________________
embedding_1 (Embedding)         (None, 10, 256)      15616       input_1[0][0]                    
__________________________________________________________________________________________________
input_2 (InputLayer)            [(None, 2)]          0                                            
__________________________________________________________________________________________________
lstm_1 (LSTM)                   (None, 256)          525312      embedding_1[0][0]                
__________________________________________________________________________________________________
concatenate (Concatenate)       (None, 258)          0           input_2[0][0]                    
                                                                 lstm_1[0][0]                     
__________________________________________________________________________________________________
dense_1 (Dense)                 (None, 61)           15799       concatenate[0][0]                
==================================================================================================

The error

java.lang.IllegalArgumentException: Internal error: Failed to run on the given Interpreter: tensorflow/lite/kernels/select.cc:80 data->has_rank_one_input_condition was not true.
Node number 23 (SELECT) failed to prepare.
Node number 19 (WHILE) failed to invoke.

The code

MappedByteBuffer tfliteModelCaption = FileUtil.loadMappedFile(activity, "classifier_caption.tflite");
Interpreter.Options options = new Interpreter.Options();
Interpreter tflite = new Interpreter(tfliteModelCaption, options);
float[][] inputString = new float[1][10];
inputString[0][9] = 4;
float[][] inputData = new float[1][2];
inputData[0][0] = 0;
inputData[0][1] = 1;
Object[] inputArray = {inputString, inputData};
Map<Integer, Object> outputMap = new HashMap<>();
outputMap.put(0, new float[1][61]);
tflite.runForMultipleInputsOutputs(inputArray, outputMap);
tflite.close();
float[] result = ((float[][]) outputMap.get(0))[0];
textViewAge.setText("Test: " + Arrays.toString(result));

The error seems to means the shape of the inputs does not fit the model input, but I checked it many time and for me it looks the same. I create a similar model with only one input (no input_2) and I can use it without error.