Cannot set tensor: Dimension mismatch. Got 3 but expected 4 for input 0

Question

This is probably going to be a stupid question but I am new to deep learning and TensorFlow. Here I have converted my deep learning model to TF-lite, after that when I am trying to do the inference with the TF lite model (Here I am dealing with audio files),I am getting this error - Cannot set tensor: Dimension mismatch.

Any help will be greatly appreciable. Thanks in Advance!

Code: '''

    import soundfile as sf
    import numpy as np
    import tflite_runtime.interpreter as tflite
    import time
    
    
    
    ##########################
    # the values are fixed, if you need other values, you have to retrain.
    # The sampling rate of 16k is also fix.
    block_len = 512
    block_shift = 128
    # load models
    interpreter_1 = tflite.Interpreter(model_path='./models_DTLN_model/models_DTLN_model_1.tflite')
    interpreter_1.allocate_tensors()
    interpreter_2 = tflite.Interpreter(model_path='./models_DTLN_model/models_DTLN_model_2.tflite')
    interpreter_2.allocate_tensors()
    
    # Get input and output tensors.
    input_details_1 = interpreter_1.get_input_details()
    output_details_1 = interpreter_1.get_output_details()
    
    input_details_2 = interpreter_2.get_input_details()
    output_details_2 = interpreter_2.get_output_details()
    # create states for the lstms
    states_1 = np.zeros(input_details_1[1]['shape']).astype('float32')
    states_2 = np.zeros(input_details_2[1]['shape']).astype('float32')
    # load audio file at 16k fs (please change)
    audio,fs = sf.read('./models_DTLN_model/input/fileid_3.wav')
    # check for sampling rate
    if fs != 16000:
        raise ValueError('This model only supports 16k sampling rate.')
    # preallocate output audio
    out_file = np.zeros((len(audio)))
    # create buffer
    in_buffer = np.zeros((block_len)).astype('float32')
    out_buffer = np.zeros((block_len)).astype('float32')
    # calculate number of blocks
    num_blocks = (audio.shape[0] - (block_len-block_shift)) // block_shift
    time_array = []      
    # iterate over the number of blcoks  
    for idx in range(num_blocks):
        start_time = time.time()
        # shift values and write to buffer
        in_buffer[:-block_shift] = in_buffer[block_shift:]
        in_buffer[-block_shift:] = audio[idx*block_shift:(idx*block_shift)+block_shift]
        # calculate fft of input block
        in_block_fft = np.fft.rfft(in_buffer)
        in_mag = np.abs(in_block_fft)
        in_phase = np.angle(in_block_fft)
        # reshape magnitude to input dimensions
        in_mag = np.reshape(in_mag, (1,1,-1)).astype('float32')
        # set tensors to the first model
        interpreter_1.set_tensor(input_details_1[1]['index'], states_1)
        interpreter_1.set_tensor(input_details_1[0]['index'], in_mag)
        # run calculation
        interpreter_1.invoke()
        # get the output of the first block
        out_mask = interpreter_1.get_tensor(output_details_1[0]['index']) 
        states_1 = interpreter_1.get_tensor(output_details_1[1]['index'])
# calculate the ifft
    estimated_complex = in_mag * out_mask * np.exp(1j * in_phase)
    estimated_block = np.fft.irfft(estimated_complex)
    # reshape the time domain block
    estimated_block = np.reshape(estimated_block, (1,1,-1)).astype('float32')
    # set tensors to the second block
    interpreter_2.set_tensor(input_details_2[1]['index'], states_2)
    interpreter_2.set_tensor(input_details_2[0]['index'], estimated_block)
    # run calculation
    interpreter_2.invoke()
    # get output tensors
    out_block = interpreter_2.get_tensor(output_details_2[0]['index']) 
    states_2 = interpreter_2.get_tensor(output_details_2[1]['index']) 
    
    
    # shift values and write to buffer
    out_buffer[:-block_shift] = out_buffer[block_shift:]
    out_buffer[-block_shift:] = np.zeros((block_shift))
    out_buffer  += np.squeeze(out_block)
    # write block to output file
    out_file[idx*block_shift:(idx*block_shift)+block_shift] = out_buffer[:block_shift]
    time_array.append(time.time()-start_time)
    
# write to .wav file 
sf.write('out.wav', out_file, fs) 
print('Processing Time [ms]:')
print(np.mean(np.stack(time_array))*1000)
print('Processing finished.')

'''

Error: I got the following Dimension mismatch error.

'''

Traceback (most recent call last):
  File "real_time_processing_tf_lite.py", line 69, in <module>
    interpreter_1.set_tensor(input_details_1[0]['index'], in_mag)
  File "/usr/local/lib/python3.7/site-packages/tflite_runtime/interpreter.py", line 698, in set_tensor
    self._interpreter.SetTensor(tensor_index, value)
ValueError: Cannot set tensor: Dimension mismatch. Got 3 but expected 4 for input 0.

''' This is what shape suppose to be:

print(interpreter_1.get_input_details())

>>name': 'serving_default_input_3:0', 'index': 0, 'shape': array([ 1, 2, 128, 2], dtype=int32), 'shape_signature': array([ 1, 2, 128, 2], dtype=int32), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}, {'name': 'serving_default_input_2:0', 'index': 1, 'shape': array([ 1, 1, 257], dtype=int32), 'shape_signature': array([ 1, 1, 257], dtype=int32), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}]

And when I am adding another dimension in the in_mag, getting this error:

in_mag = np.reshape(in_mag, (-1,1,1,-1)).astype('float32')


>>ValueError: can only specify one unknown dimension 

And for other values getting this error:

>>cannot reshape array of size 257 into shape (1,128,newaxis)

Answer 1

You're setting the shape here to a tensor of size 3:

in_mag = np.reshape(in_mag, (1,1,-1)).astype('float32')

You need to add another dimension. Not familiar with what you're trying to achieve, so this is only a suggestion:

in_mag = np.reshape(in_mag, (-1,1,1,-1)).astype('float32')