Open RosettaWYzhang opened 2 years ago
Hi! Thanks for providing the code.
For Neural State Machine, in the paper you mentioned that you are using AdamWR optimizer. However, in the code (https://github.com/sebastianstarke/AI4Animation/blob/master/AI4Animation/SIGGRAPH_Asia_2019/TensorFlow/NSM/Lib_Optimizer/AdamW.py), why is weight decay (wdc) is only used in _apply_sparse and not in _apply_dense? Is this optimizer just vanilla Adam optimizer?
Hi! Thanks for providing the code.
For Neural State Machine, in the paper you mentioned that you are using AdamWR optimizer. However, in the code (https://github.com/sebastianstarke/AI4Animation/blob/master/AI4Animation/SIGGRAPH_Asia_2019/TensorFlow/NSM/Lib_Optimizer/AdamW.py), why is weight decay (wdc) is only used in _apply_sparse and not in _apply_dense? Is this optimizer just vanilla Adam optimizer?