Updating agents' `action_distribution_fn` signature


The policy of several base agents - including DQN (TF and Torch), DDPG (TF and Torch), SAC (TF) - still use an action_distribution_fn function with the old signature. These agents therefore don’t benefit from the features of Trajectory View API.

Is there a ticket and/or ongoing work to update them? Can we expect all agents to be updated in the 2.0.0 release?

This seems to be the last barrier to use custom models implementing shared computation graphs with built-in policies in the multi-agent settings :slight_smile: