### Inference in Code: an example
The task is to find out in a coffee roasting scenario, given *enough time* and *temperature* whether it result to a good coffee (properly roasted) or a bad coffee (overcooked or undercooked.
![[Pasted image 20250606224146.png]]
Now in order to implement it in TensorFlow, we could suppose that `x` is an array of two numbers (enough time and temperature):
```python
x = np.array([[200, 17]])
layer_1 = Dense(units=3, activation="sigmoid")
a1 = layer_1(x)
```
Here we calculated the output of the first layer `a1` by the `Dense` function which is a *kind* of neural network function in TensorFlow.
```python
layer_2 = Dense(units=1, activation="sigmoid")
a2 = layer_2(a1)
```
Now `a2` is the answer of our algorithm. If we want to get the result, we have to check if it's more than 0.5 or not.
> [!tip]
> You can define a `model` in TensorFlow and chain out each layer by `Sequential` function:
> ```python
> model = Sequential([
> Dense(units=3, activation="sigmoid"),
> Dense(units=1, activation="sigmoid")
> ])
> # model.fit(x, y)
> # model.compile()
> # model.predict(x_new)
> ```