[rllib] Fix atari reward calculations, add LR annealing, explained var stat for A2C / impala (#2700)

Changes needed to reproduce Atari plots in IMPALA / A2C: https://github.com/ray-project/rl-experiments
2026-08-19 12:30:27 +08:00 · 2018-08-23 17:49:10 -07:00
parent 1b3de31ff1
commit aa014af85b
35 changed files with 483 additions and 148 deletions
@@ -47,6 +47,5 @@ pendulum-ddpg:
        num_workers: 0
        num_gpus_per_worker: 0
        optimizer_class: "SyncReplayOptimizer"
-        optimizer_config: {}
        per_worker_exploration: False
        worker_side_prioritization: False