Better model
Much simpler model definition. Just do:
class MyNewModel(PVModel):
KERNEL_SIZE = [25, 15, 5] # list of sizes
CHANNELS_SIZE = [20, 5, 1] # List of channels (always ends in 1 for a single output)
DEFAULTS = {'dropout_1':0.15, 'dropout_3':0.35} # Optional dropouts (can be set when instantiating the model, as well)
The models do have different channel names; if you want to load saved parameters from an old model with the new model definitions, do:
from model.core import modernize
d = torch.load(name)
d = modernize(d) # Only use if using the new Model definitions!
model.load_state_dict(d)
Better plotting, with extensive printouts.
Revisited model saving; fixed option saving, adding model to saved output files.
Edited by Henry Fredrick Schreiner