Reward2GoTransform

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

根據 episode reward 和折扣因子計算 reward to go。

由於 Reward2GoTransform 僅是一個反向轉換，in_keys 將直接用於 in_keys_inv。只有在 episode 結束後才能計算 reward-to-go。因此，這個轉換應該應用於 replay buffer，而不是應用於 collector 或環境中。

參數:

• gamma (float or torch.Tensor) – 折扣因子。預設值為 1.0。

• in_keys (sequence of NestedKey) – 要重新命名的條目序列。如果未提供，預設值為 ("next", "reward")。

• out_keys (sequence of NestedKey) – 要重新命名的條目序列。如果未提供，預設值為 in_keys 的值。

• done_key (NestedKey) – 完成 (done) 條目。預設值為 "done"。

• truncated_key (NestedKey) – 截斷 (truncated) 條目。預設值為 "truncated"。如果未找到截斷條目，則只會使用 "done"。

範例

>>> # 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]]])

也可以直接將這個轉換用於 collector：請確保附加 transform 的 inv 方法。

範例

>>> 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)

將這個轉換作為環境的一部分會引發例外

範例

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

注意

在存在多個完成條目的設定中，應該為每個 done-reward 對建立一個單獨的 Reward2GoTransform。

forward(tensordict: TensorDictBase) → TensorDictBase

讀取輸入 tensordict，並對選定的鍵應用轉換。