Encountered “RuntimeError: Can Not Constant Fold Eltwise Node” Error When Inferencing with OpenVINO™

• Converted custom Keras model into Intermediate Representation (IR) with Model Optimizer:
  mo --use_new_frontend --framework tf --input_shape --saved_model_dir --output_dir
• Received "RuntimeError: Can not constant fold eltwise node" error when inferencing with the IR file.

The error is due to the missing proper support for the 64-bit type of transformation.

1. Change the precision of the model into 32-bit:
   # set precision to 64 bit
   - tf.keras.backend.set_floatx('float64')
   + tf.keras.backend.set_floatx('float32')
2. Generate the IR file where 64-bit is replaced with 32-bit.