Reward2GoTransform

class torchrl.envs.transforms.Reward2GoTransform(gamma: Optional[Union[float, torch.Tensor]] = 1.0, in_keys: Sequence[NestedKey] | None = None, out_keys: Sequence[NestedKey] | None = None, done_key: Optional[NestedKey] = 'done')

根據回合獎勵和折扣因子計算剩餘獎勵（reward-to-go）。

由於 Reward2GoTransform 僅是一個逆向變換（inverse transform），in_keys 將直接用於 in_keys_inv。剩餘獎勵（reward-to-go）只能在回合結束時計算。因此，該變換應該應用於經驗回放緩衝區（replay buffer），而不是收集器（collector）或環境（environment）內部。

引數：

• gamma (float 或 torch.Tensor) – 折扣因子。預設為 1.0。

• in_keys (NestedKey 序列) – 需要重新命名的條目。如果未提供，預設為 ("next", "reward")。

• out_keys (NestedKey 序列) – 需要重新命名的條目。如果未提供，預設為 in_keys 的值。

• done_key (NestedKey) – done 條目。預設為 "done"。

• truncated_key (NestedKey) – truncated 條目。預設為 "truncated"。如果未找到 truncated 條目，則僅使用 "done"。

示例

>>> # Using this transform as part of a replay buffer
>>> from torchrl.data import ReplayBuffer, LazyTensorStorage
>>> torch.manual_seed(0)
>>> r2g = Reward2GoTransform(gamma=0.99, out_keys=["reward_to_go"])
>>> rb = ReplayBuffer(storage=LazyTensorStorage(100), transform=r2g)
>>> batch, timesteps = 4, 5
>>> done = torch.zeros(batch, timesteps, 1, dtype=torch.bool)
>>> for i in range(batch):
...     while not done[i].any():
...         done[i] = done[i].bernoulli_(0.1)
>>> reward = torch.ones(batch, timesteps, 1)
>>> td = TensorDict(
...     {"next": {"done": done, "reward": reward}},
...     [batch, timesteps],
... )
>>> rb.extend(td)
>>> sample = rb.sample(1)
>>> print(sample["next", "reward"])
tensor([[[1.],
         [1.],
         [1.],
         [1.],
         [1.]]])
>>> print(sample["reward_to_go"])
tensor([[[4.9010],
         [3.9404],
         [2.9701],
         [1.9900],
         [1.0000]]])

也可以直接將此變換與收集器（collector）一起使用：確保附加變換的 inv 方法。

示例

>>> from torchrl.envs.utils import RandomPolicy        >>> from torchrl.collectors import SyncDataCollector
>>> from torchrl.envs.libs.gym import GymEnv
>>> t = Reward2GoTransform(gamma=0.99, out_keys=["reward_to_go"])
>>> env = GymEnv("Pendulum-v1")
>>> collector = SyncDataCollector(
...     env,
...     RandomPolicy(env.action_spec),
...     frames_per_batch=200,
...     total_frames=-1,
...     postproc=t.inv
... )
>>> for data in collector:
...     break
>>> print(data)
TensorDict(
    fields={
        action: Tensor(shape=torch.Size([200, 1]), device=cpu, dtype=torch.float32, is_shared=False),
        collector: TensorDict(
            fields={
                traj_ids: Tensor(shape=torch.Size([200]), device=cpu, dtype=torch.int64, is_shared=False)},
            batch_size=torch.Size([200]),
            device=cpu,
            is_shared=False),
        done: Tensor(shape=torch.Size([200, 1]), device=cpu, dtype=torch.bool, is_shared=False),
        next: TensorDict(
            fields={
                done: Tensor(shape=torch.Size([200, 1]), device=cpu, dtype=torch.bool, is_shared=False),
                observation: Tensor(shape=torch.Size([200, 3]), device=cpu, dtype=torch.float32, is_shared=False),
                reward: Tensor(shape=torch.Size([200, 1]), device=cpu, dtype=torch.float32, is_shared=False)},
            batch_size=torch.Size([200]),
            device=cpu,
            is_shared=False),
        observation: Tensor(shape=torch.Size([200, 3]), device=cpu, dtype=torch.float32, is_shared=False),
        reward: Tensor(shape=torch.Size([200, 1]), device=cpu, dtype=torch.float32, is_shared=False),
        reward_to_go: Tensor(shape=torch.Size([200, 1]), device=cpu, dtype=torch.float32, is_shared=False)},
    batch_size=torch.Size([200]),
    device=cpu,
    is_shared=False)

在環境（env）中使用此變換將引發異常

示例

>>> t = Reward2GoTransform(gamma=0.99)
>>> TransformedEnv(GymEnv("Pendulum-v1"), t)  # crashes

注意

在存在多個 done 條目的情況下，應為每個 done-reward 對構建一個單獨的 Reward2GoTransform。

forward(tensordict: TensorDictBase) → TensorDictBase

讀取輸入的 tensordict，並對選定的鍵應用變換。