"Change begins at the end of your comfort zone."
Use buffer worker if you need
- to reanalyze the sample (e.g. MuZero), or
- to make changes to your sample (e.g. hindsight relabeling)
before the samples are sent to the trainers.
- To reanalyze
- specify policy, policy_name, policy_identifier.
- To relabel
- specify data augmenter.
- If configured to run both, relabel is done before reanalyze.