vec_td_lambda_return_estimate

class torchrl.objectives.value.functional.vec_td_lambda_return_estimate(gamma, lmbda, next_state_value, reward, done, rolling_gamma: Optional[bool] = None, time_dim: int = - 2)

Vectorized TD(\(\lambda\)) return estimate.

Parameters:

• gamma (scalar, Tensor) – exponential mean discount. If tensor-valued, must be a [Batch x TimeSteps x 1] tensor.

• lmbda (scalar) – trajectory discount.

• next_state_value (Tensor) – value function result with new_state input. must be a [Batch x TimeSteps x 1] tensor

• reward (Tensor) – reward of taking actions in the environment. must be a [Batch x TimeSteps x 1] or [Batch x TimeSteps] tensor

• done (Tensor) – boolean flag for end of episode.

• rolling_gamma (bool, optional) –

  if True, it is assumed that each gamma if a gamma tensor is tied to a single event:

  gamma = [g1, g2, g3, g4] value = [v1, v2, v3, v4] return = [

  v1 + g1 v2 + g1 g2 v3 + g1 g2 g3 v4, v2 + g2 v3 + g2 g3 v4, v3 + g3 v4, v4,

  ]

  if False, it is assumed that each gamma is tied to the upcoming trajectory:

  gamma = [g1, g2, g3, g4] value = [v1, v2, v3, v4] return = [

  v1 + g1 v2 + g1**2 v3 + g**3 v4, v2 + g2 v3 + g2**2 v4, v3 + g3 v4, v4,

  ]

  Default is True.

• time_dim (int) – dimension where the time is unrolled. Defaults to -2.

All tensors (values, reward and done) must have shape [*Batch x TimeSteps x *F], with *F feature dimensions.