Open maxencefaldor opened 4 months ago
In the appendix you mention that prioritized replay improves the performance of Dreamer.
Could you please provide further clarification on the replay parameters used in the configuration?
replay: size: 5e6 online: True fracs: {uniform: 1.0, priority: 0.0, recency: 0.0} prio: {exponent: 0.8, maxfrac: 0.5, initial: inf, zero_on_sample: True} priosignal: model recexp: 1.0 chunksize: 1024 save_wait: False
Thank you!
In the appendix you mention that prioritized replay improves the performance of Dreamer.
Could you please provide further clarification on the replay parameters used in the configuration?
Thank you!