soar/surrogate.py at master · avidcoder123/soar · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
import equinox as eqx
import jax
import jax.numpy as jnp

class LiftSurrogateModel(eqx.Module):

    layers: list

    def __init__(self, in_size, out_size, width_size, depth, activation, key):
        keys = jax.random.split(key, depth + 2)

        input_key = keys[0]
        output_key = keys[-1]
        hidden_keys = keys[1:-1]

        input_layer = eqx.nn.Linear(in_size, width_size, key=input_key)
        output_layer = eqx.nn.Linear(width_size, out_size, key=output_key)


        self.layers = [
            jax.nn.standardize, #Standardize -1 to 1
            input_layer,
            activation
        ]
        for key in hidden_keys:
            self.layers.append(eqx.nn.Linear(width_size, width_size, key=key))
            self.layers.append(activation)

        self.layers.append(output_layer)

    def __call__(self, x):

        for layer in self.layers:
            x = layer(x)

        return x

#Define the structure for the neural network
class DragSurrogateModel(eqx.Module):

    layers: list

    def __init__(self, in_size, out_size, width_size, depth, activation, key):
        keys = jax.random.split(key, depth + 2)

        input_key = keys[0]
        output_key = keys[-1]
        hidden_keys = keys[1:-1]

        input_layer = eqx.nn.Linear(in_size, width_size, key=input_key)
        output_layer = eqx.nn.Linear(width_size, out_size, key=output_key)

        #Make Reynolds number on log10 scale
        @jax.jit
        def normalize_reynolds_number(x):
            Re = x[-1]
            Re = jnp.log10(Re)

            #Set maximum of Re=10^6
            Re = jnp.min(jnp.hstack((Re, 6)))

            return jnp.hstack((x[:-1], Re))


        self.layers = [
            normalize_reynolds_number,
            jax.nn.standardize, #Standardize -1 to 1
            input_layer,
            activation
        ]
        for key in hidden_keys:
            self.layers.append(eqx.nn.Linear(width_size, width_size, key=key))
            self.layers.append(activation)

        self.layers.append(output_layer)

    def __call__(self, x):

        for layer in self.layers:
            x = layer(x)

        return x