[Python] Issue with "IndexError: _Map_base::at" when convert a tensorflow model to tensorflow...

I want to test the performance of the quantized tensorflow model .First, I need to quantize the model using tensorflow lite,and got the .tflite model.

When I converter a tensorflow model .h5 to .tflite,print the error:

Exception has occurred: IndexError _Map_base::at File tflite_model = converter.convert() IndexError: _Map_base::at"
​

The tensorflow version is V2.4; ubuntu 22.04; RX3090

The code is as follows: original code

if __name__ == '__main__':

dpcrn = DPCRN_model(batch_size = 1,
length_in_s = 5,
lr = 1e-3,
block_len = 400,
block_shift = 200)
my_model = dpcrn.build_DPCRN_model()
converter = tf.lite.TFLiteConverter.from_keras_model(my_model)
converter.optimizations = [tf.lite.Optimize.DEFAULT]
converter.target_spec.supported_ops = [tf.lite.OpsSet.SELECT_TF_OPS,tf.lite.OpsSet.TFLITE_BUILTINS_INT8] #
converter.representative_dataset = representative_data_gen
tflite_model = converter.convert()

quant_model_path = './model_tflite.tflite'
with open(quant_model_path, 'wb') as f:
f.write(tflite_model)

Continue reading...