PNN adapter arguement missing in the _init_() method ContinualAI#1625

The issue is due to the num_prev_modules attribute not being initialized in the LinearAdapter class within the PNN.py module.

avalanche/models/pnn.py

@@ -20,6 +20,7 @@ def __init__(self, in_features, out_features_per_column, num_prev_modules):
         :param num_prev_modules: number of previous modules
         """
         super().__init__()
+        self.num_prev_modules = num_prev_modules
         # Eq. 1 - lateral connections
         # one layer for each previous column. Empty for the first task.
         self.lat_layers = nn.ModuleList([])