(self, X)
| 1237 | self.layer1.bias = nn.Parameter(torch.FloatTensor(b.flatten())) |
| 1238 | |
| 1239 | def forward(self, X): |
| 1240 | # (N, W, C) -> (N, C, W) |
| 1241 | self.X = np.moveaxis(X, [0, 1, 2], [0, -1, -2]) |
| 1242 | if not isinstance(self.X, torch.Tensor): |
| 1243 | self.X = torchify(self.X) |
| 1244 | |
| 1245 | self.X.retain_grad() |
| 1246 | |
| 1247 | self.Z = self.layer1(self.X) |
| 1248 | self.Z.retain_grad() |
| 1249 | |
| 1250 | self.Y = self.act_fn(self.Z) |
| 1251 | self.Y.retain_grad() |
| 1252 | return self.Y |
| 1253 | |
| 1254 | def extract_grads(self, X): |
| 1255 | self.forward(X) |
no test coverage detected