Before this PR, for uniformity, all experiments just took one model config and experiments involving actor critic losses had by default the same config for actor and critic
With this PR we introduce the possibility of passing a critic model config that can be used by actor critic losses.
If not provided this will default to the actor config.
Before this PR, for uniformity, all experiments just took one model config and experiments involving actor critic losses had by default the same config for actor and critic
With this PR we introduce the possibility of passing a critic model config that can be used by actor critic losses.
If not provided this will default to the actor config.