Are algorithms that support MAS, multi agent variations of single agent version

This page has a list of supported algorithms:
https://docs.ray.io/en/latest/rllib/rllib-algorithms.html
And some of them support multiagent systems (MAS)

For example, TD3 comes in multiagent version

are those multiagent versions algorithms pure MA algorithms (e.g. MATD3 with CTDE) or Independent agents running that algorithms (e.g. Independent-TD3)

DDPG and MADDPG are both available algorithms.

Thanks!

We can distinguish between heterogeneous and homogeneous multi-agent setups.
Where either all agents follow the same policy or they don’t.
The algorithms we provide have a policy mapping function for that purpose.
Manny has discussed this here.