soft module codebase notes

[NirViaje]

https://rchalyang.github.io/SoftModule/

"epoch_frames" : 200
"batch_size" : 1280

torchrl/algo/off_policy/twin_sac_q.py

get sparse loss here

        """
        Policy Loss
        """
        if not self.reparameterization:
            raise NotImplementedError
        else:
            assert log_probs.shape == q_new_actions.shape
            policy_loss = ( alpha * log_probs - q_new_actions).mean()

        std_reg_loss = self.policy_std_reg_weight * (log_std**2).mean()
        mean_reg_loss = self.policy_mean_reg_weight * (mean**2).mean()

        policy_loss += std_reg_loss + mean_reg_loss

mujoco210-linux-x86_64.tar.gz

local_debug_logger-master.zip

[NirViaje]

CoBERL

v-MPO