Self-Supervised Models
Configuration Classes¶
Bases: SSLModelConfig
DeNoising AutoEncoder configuration.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
noise_strategy |
str
|
Defines what kind of noise we are introducing to samples. |
'swap'
|
noise_probabilities |
Dict[str, float]
|
Dict of individual probabilities to corrupt the input features with swap/zero noise. Key should be the feature name and if any feature is missing, the default_noise_probability is used. Default is an empty dict() |
lambda: {}()
|
default_noise_probability |
float
|
Default probability to corrupt the input features with swap/zero noise. For features for which noise_probabilities does not define a probability. Default is 0.8 |
0.8
|
loss_type_weights |
Optional[List[float]]
|
Weights to be used for the loss function in the order [binary, categorical, numerical]. If None, will use the default weights using a formula. eg. for binary, default weight will be n_binary/n_features. Defaults to None |
None
|
mask_loss_weight |
float
|
Weight to be used for the loss function for the masked features. Defaults to 1.0 |
2.0
|
max_onehot_cardinality |
int
|
Maximum cardinality of one-hot encoded categorical features. Any categorical feature with cardinality>max_onehot_cardinality will be embedded in a learned embedding space and others will be converted to a one hot representation. If set to 0, will use the embedding strategy for all categorical feature. Default is 4 |
4
|
include_input_features_inference |
bool
|
If True, will include the input features along with the learned features while fine tuning. Defaults to False |
False
|
encoder_config |
Optional[ModelConfig]
|
The config of the encoder to be used for the model. Should be one of the model configs defined in PyTorch Tabular |
None
|
decoder_config |
Optional[ModelConfig]
|
The config of decoder to be used for the model. Should be one of the model configs defined in PyTorch Tabular. Defaults to nn.Identity |
None
|
embedding_dims |
Optional[List]
|
The dimensions of the embedding for each categorical column as a list of tuples (cardinality, embedding_dim). If left empty, will infer using the cardinality of the categorical column using the rule min(50, (x + 1) // 2) |
None
|
embedding_dropout |
float
|
Dropout to be applied to the Categorical Embedding. Defaults to 0.1 |
0.1
|
batch_norm_continuous_input |
bool
|
If True, we will normalize the continuous layer by passing it through a BatchNorm layer. |
True
|
learning_rate |
float
|
The learning rate of the model. Defaults to 1e-3 |
0.001
|
seed |
int
|
The seed for reproducibility. Defaults to 42 |
42
|
Source code in src/pytorch_tabular/ssl_models/dae/config.py
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 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 |
|
Model Classes¶
Bases: SSLBaseModel
Source code in src/pytorch_tabular/ssl_models/dae/dae.py
83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 |
|
Base Model Class¶
Bases: LightningModule
Source code in src/pytorch_tabular/ssl_models/base_model.py
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 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 |
|
__init__(config, mode='pretrain', encoder=None, decoder=None, custom_optimizer=None, custom_optimizer_params={}, **kwargs)
¶
Base Model for all SSL Models.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
config |
DictConfig
|
Configuration defined by the user |
required |
mode |
str
|
Mode of the model. Defaults to "pretrain". |
'pretrain'
|
encoder |
Optional[Module]
|
Encoder of the model. Defaults to None. |
None
|
decoder |
Optional[Module]
|
Decoder of the model. Defaults to None. |
None
|
custom_optimizer |
Optional[Optimizer]
|
Custom optimizer to use. Defaults to None. |
None
|
custom_optimizer_params |
Dict
|
Custom optimizer parameters to use. Defaults to {}. |
{}
|