| import torch as torch |
| import torch.nn as nn |
| from src.nn.nn import NNModel |
|
|
|
|
| class FeedForward(NNModel): |
| layers: nn.ModuleList |
| num_units: int |
| num_inputs: int |
| num_outputs: int |
|
|
| def __init__(self, config_, hyperparameters_): |
| super(FeedForward, self).__init__(config_, hyperparameters_) |
| self.config = config_ |
| self.hyperparameters = hyperparameters_ |
| self.num_layers = self.hyperparameters.num_layers |
| self.num_units = self.hyperparameters.num_units |
| self.num_inputs = self.config.num_inputs |
| self.num_outputs = self.config.num_outputs |
|
|
| self.layers = nn.ModuleList() |
| self.layers.append(torch.nn.Linear(self.num_inputs, self.num_units)) |
|
|
| for _ in range(1, self.num_layers - 1): |
| self.layers.append(torch.nn.Linear(self.num_units, self.num_units)) |
|
|
| self.layers.append(torch.nn.Linear(self.num_units, self.num_outputs)) |
|
|
| self.activation = self.hyperparameters.activation |
| self.dropout = nn.Dropout(self.hyperparameters.dropout) |
|
|
| def predict(self, data): |
| return self.forward(data) |
|
|
| def forward(self, x): |
| out = self.layers[0](x) |
| for idx in range(1, self.num_layers - 1): |
| out = self.activation(self.layers[idx](out)) |
| out = self.dropout(out) |
| out = self.layers[-1](out) |
| return out |
