model: pytorch: Add support for additional layers via Python API

Previously only supplying a dict which would then be converted to
PyTorch objects was supported. Now PyTorch objects can be supplied
directly.

Fixes: #1147
Related: #840
Related: #1151

model/pytorch/dffml_model_pytorch/pytorch_pretrained.py

@@ -1,4 +1,6 @@
 import sys
+from typing import Union
+
 import torch.nn as nn
 from torchvision import models
 
@@ -26,7 +28,9 @@ class PyTorchPreTrainedModelConfig(PyTorchModelConfig):
     add_layers: bool = field(
         "Replace the last layer of the pretrained model", default=False,
     )
-    layers: dict = field(
+    layers: Union[
+        dict, nn.ModuleDict, nn.Sequential, nn.ModuleList, nn.Module
+    ] = field(
         "Extra layers to replace the last layer of the pretrained model",
         default=None,
     )
@@ -60,10 +64,20 @@ def createModel(self):
             param.require_grad = self.parent.config.trainable
 
         if self.parent.config.add_layers:
-            layers = [
-                create_layer(value)
-                for key, value in self.parent.config.layers.items()
-            ]
+            if self.parent.config.layers.__class__.__base__.__name__ in [
+                "ModuleDict",
+                "Sequential",
+                "ModuleList",
+                "Module",
+            ]:
+                layers = nn.Sequential()
+                for name, module in self.parent.config.layers.named_children():
+                    layers.add_module(name, module)
+            else:
+                layers = [
+                    create_layer(value)
+                    for key, value in self.parent.config.layers.items()
+                ]
 
             if self.parent.LAST_LAYER_TYPE == "classifier_sequential":
                 if len(layers) > 1: