# How to define the shape of hidden and meory state in Numpy, keras?

I was very confused about the concept " number of units ". Now I clearly understand it. It means the number of LSTM cell in each time-step. I thought it means the number of neurons in each gate. I went through Numpy LSTM codes to understand LSTM model. And I have two question:

1. Each gate includes neural network layer (yellow Color). It means that we have to define the size of network. As you can see below, the shape of hidden (a) and memory (c) is (n_a, m). In this code, m means the number of training examples. But what is the n_a? I assume that it is the number of neurons in the each gate (neural network). If it is correct, why we don't define this prarameter in keras?

2. As you know that, each gate includes neural network. Lets assume that number of units is 1. It means that we have only one LSTM cell at each time step. In this case, Can I say that each gates in this LSTM cell include neural network which has only one neuron? In this one neuron, input will be multiplied with weight and added with bias.

Thank you