Some RLlib algorithms support a LR annealing schedule via “lr_schedule”. Is there support for annealing “gamma”? If not, how should I go about implementing it? Where in the code should I start digging? Thanks!

@RickLan , Here is an example of how the LR is annealed. ray-project/ray/blob/2303851c3c234a261e8791b778e59e71f8235260/rllib/policy/tf_policy.py#L944 ] for key in state_keys: feed_dict[self._loss_input_dict[key]] = tr…

Support for annealing gamma

RLlib

RickLan May 20, 2021, 7:06am 2

I just saw this: [RLlib] updating batch_size or similar while training - #2 by sven1977

Topic		Replies	Views
Learning rate annealing with tune.run() RLlib	6	1303	April 27, 2021
Change learning rete for DQN RLlib	6	618	February 25, 2022
Ray Tune: exponential learning rate schedule on HyperOpt Ray Tune	6	1360	May 28, 2023
Change params of agents Ray Tune	3	403	May 28, 2022
Dynamic Entropy Schedule Configure Algorithm, Training, Evaluation, Scaling	2	114	November 18, 2025

Support for annealing gamma

Related topics