automl / automl/Mighty

Expectations for Obs Shaping?

Open
#105 0 comments 0 reactions 1 assignee Claimed by @amsks View on GitHub
bug question
Dominant language
Python
Stars
61
Forks
2
PR merge metrics
No merged PRs in 30d

Description

I just checked the PPO init and saw that only vector-shaped obs are supported (no images) due to only passing obs_shape[0] to buffer and model (same for action). The model creation supports both in principle, so I'm a bit surprised and not sure if this is a good way of handling it since it prevents the use of e.g. Conv Nets for images.

Suggested fix:

- [ ] check SAC and DQN init for the same issue
- [ ] simply pass full obs & action shapes without the index
- [ ] in model init: don't convert to int, simply pass on the raw obs shape

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.