Skip to content

model_arguments

Define arguments for each of the model classes.

add_ctgan_args(group, overrides=False)

Adds CTGAN-specific arguments. GAN backbone args are registered by the GAN/DPGAN groups.

Source code in src/nhssynth/cli/model_arguments.py 
186
187
188
189
190
191
192
193
194
195
196
197
198
199
def add_ctgan_args(group: argparse._ArgumentGroup, overrides: bool = False) -> None:
    """Adds CTGAN-specific arguments. GAN backbone args are registered by the GAN/DPGAN groups."""
    group.add_argument(
        "--pac",
        type=int,
        default=2,
        help="number of samples packed together for the PacGAN discriminator",
    )
    group.add_argument(
        "--lambda-cond",
        type=float,
        default=1.0,
        help="weight of the conditional cross-entropy loss in the generator",
    )

add_gan_args(group, overrides=False)

Adds arguments to an existing group for the GAN model.

Source code in src/nhssynth/cli/model_arguments.py 
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
def add_gan_args(group: argparse._ArgumentGroup, overrides: bool = False) -> None:
    """Adds arguments to an existing group for the GAN model."""
    group.add_argument(
        "--noise-dim",
        type=int,
        default=128,
        help="the dimensionality of the generator noise input",
    )
    group.add_argument(
        "--n-units-conditional",
        type=int,
        help="the number of units in the conditional layer",
    )
    group.add_argument(
        "--generator-n-layers-hidden",
        type=int,
        help="the number of hidden layers in the generator",
    )
    group.add_argument(
        "--generator-n-units-hidden",
        type=int,
        help="the number of units in each hidden layer of the generator",
    )
    group.add_argument(
        "--generator-activation",
        type=str,
        choices=list(ACTIVATION_FUNCTIONS.keys()),
        help="the activation function of the generator",
    )
    group.add_argument(
        "--generator-batch-norm",
        action="store_true",
        help="whether to use batch norm in the generator",
    )
    group.add_argument(
        "--generator-dropout",
        type=float,
        help="the dropout rate in the generator",
    )
    group.add_argument(
        "--generator-lr",
        type=float,
        help="the learning rate for the generator",
    )
    group.add_argument(
        "--generator-residual",
        action="store_true",
        help="whether to use residual connections in the generator",
    )
    group.add_argument(
        "--generator-opt-betas",
        type=float,
        nargs=2,
        help="the beta values for the generator optimizer",
    )
    group.add_argument(
        "--discriminator-n-layers-hidden",
        type=int,
        help="the number of hidden layers in the discriminator",
    )
    group.add_argument(
        "--discriminator-n-units-hidden",
        type=int,
        help="the number of units in each hidden layer of the discriminator",
    )
    group.add_argument(
        "--discriminator-activation",
        type=str,
        choices=list(ACTIVATION_FUNCTIONS.keys()),
        help="the activation function of the discriminator",
    )
    group.add_argument(
        "--discriminator-batch-norm",
        action="store_true",
        help="whether to use batch norm in the discriminator",
    )
    group.add_argument(
        "--discriminator-dropout",
        type=float,
        help="the dropout rate in the discriminator",
    )
    group.add_argument(
        "--discriminator-lr",
        type=float,
        help="the learning rate for the discriminator",
    )
    group.add_argument(
        "--discriminator-opt-betas",
        type=float,
        nargs=2,
        help="the beta values for the discriminator optimizer",
    )
    group.add_argument(
        "--clipping-value",
        type=float,
        help="the clipping value for the discriminator",
    )
    group.add_argument(
        "--lambda-gradient-penalty",
        type=float,
        help="the gradient penalty coefficient",
    )

add_model_specific_args(group, name, overrides=False)

Adds arguments to an existing group according to name.

Source code in src/nhssynth/cli/model_arguments.py 
 8
 9
10
11
12
13
14
15
16
17
def add_model_specific_args(group: argparse._ArgumentGroup, name: str, overrides: bool = False) -> None:
    """Adds arguments to an existing group according to `name`."""
    if name == "VAE":
        add_vae_args(group, overrides)
    elif name == "GAN":
        add_gan_args(group, overrides)
    elif name == "CTGAN":
        add_ctgan_args(group, overrides)
    elif name == "TabularGAN":
        add_tabular_gan_args(group, overrides)

add_vae_args(group, overrides=False)

Adds arguments to an existing group for the VAE model.

Source code in src/nhssynth/cli/model_arguments.py 
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
def add_vae_args(group: argparse._ArgumentGroup, overrides: bool = False) -> None:
    """Adds arguments to an existing group for the VAE model."""
    group.add_argument(
        "--encoder-latent-dim",
        type=int,
        nargs="+",
        help="the latent dimension of the encoder",
    )
    group.add_argument(
        "--encoder-hidden-dim",
        type=int,
        nargs="+",
        help="the hidden dimension of the encoder",
    )
    group.add_argument(
        "--encoder-activation",
        type=str,
        nargs="+",
        choices=list(ACTIVATION_FUNCTIONS.keys()),
        help="the activation function of the encoder",
    )
    group.add_argument(
        "--encoder-learning-rate",
        type=float,
        nargs="+",
        help="the learning rate for the encoder",
    )
    group.add_argument(
        "--decoder-latent-dim",
        type=int,
        nargs="+",
        help="the latent dimension of the decoder",
    )
    group.add_argument(
        "--decoder-hidden-dim",
        type=int,
        nargs="+",
        help="the hidden dimension of the decoder",
    )
    group.add_argument(
        "--decoder-activation",
        type=str,
        nargs="+",
        choices=list(ACTIVATION_FUNCTIONS.keys()),
        help="the activation function of the decoder",
    )
    group.add_argument(
        "--decoder-learning-rate",
        type=float,
        nargs="+",
        help="the learning rate for the decoder",
    )
    group.add_argument(
        "--shared-optimizer",
        action="store_true",
        help="whether to use a shared optimizer for the encoder and decoder",
    )