Reinterpret cast tensor from float type to string typ

Question

I am using an ML library which forces me to provide input as a numpy array with dtype float32. But on the Tensorflow side I want to input tensors which contain