inference-based-rl
inference-based-rl copied to clipboard
Observation Normalization Problem
Hi,
the normalization process used in your AWR implementation is not the same with original AWR paper.
Is there any reference or reason to implement so? It seems not right in compare to offical AWR code.
Best.
https://github.com/frt03/inference-based-rl/blob/8c93996a172f266ed402d8c0a82ecb9b4229bce0/pfrlx/networks/empirical_normalization.py#L62-L85