Open sarvghotra opened 3 years ago
The paper states that pre-trained RRDB is supposed to be frozen for the first half of the SR-Flow training. But according to the code, here and here, RRDB network is getting trained from the start.
:bug:
The paper states that pre-trained RRDB is supposed to be frozen for the first half of the SR-Flow training. But according to the code, here and here, RRDB network is getting trained from the start.