Method loss

PPO loss obs: dict with key state/rgb; more recent obs at the end state: (B, To, Do) rgb: (B, To, C, H, W) chains: (B, K+1, Ta, Da) returns: (B, ) values: (B, ) advantages: (B,) oldlogprobs: (B, K, Ta, Da) use_

(
        self,
        obs,
        chains_prev,
        chains_next,
        denoising_inds,
        returns,
        oldvalues,
        advantages,
        oldlogprobs,
        use_bc_loss=False,
        reward_horizon=4,
    )