Skip to main content

compute_nstep_returns

Function compute_nstep_returns 

Source
pub fn compute_nstep_returns(
    buffer: &mut RolloutBuffer,
    last_values: &[f32],
    gamma: f32,
)
Expand description

Compute n-step returns (simpler alternative to GAE)

This computes un-discounted n-step returns without advantage normalization. Useful for debugging or when GAE is not needed.

ยงArguments

  • buffer - Rollout buffer to compute returns for
  • last_values - Value estimates for final states [num_envs]
  • gamma - Discount factor