Write custom serialisation for experience replay
Open
enhancement
- Dominant language
- Python
- Stars
- 380
- Forks
- 63
- PR merge metrics
- No merged PRs in 30d
Description
Although it is now possible to save/load experience replay memories (https://github.com/Kaixhin/PlaNet/issues/3), naively using `torch.save` fails with large memories. Dealing with this would require custom serialisation code.
Contributor guide
Research direction
Start by reviewing the existing experience replay save/load support referenced in issue #3 and locate where torch.save is used. Implement custom serialization that works for large memories while preserving save and load behavior, then verify that large experience replay memories can be restored correctly.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100