unet_snipped
¶
Modification in U-Net model for fwi-reanalysis. The upsampling layers towards the end are removed before the activation resolution gets higher than the output resolution.
Module Contents¶
-
class
unet_snipped.
Model
(hparams)¶ Bases:
model.unet.Model
This class implements modified U-Net module by removing the up-sampling layers once activation resolution matches with the resolution of fwi-reanalysis. It is equivalent to PyTorch’s nn.Module in all aspects.
- Parameters
LightningModule (nn.Module) – The Pytorch-Lightning module derived from nn.module with useful hooks
-
forward
(self, x)¶ Does the forward pass on the model.
- Parameters
x (torch.Tensor) – Input tensor batch.
- Returns
Output activations.
- Return type
torch.Tensor