522f7fbf98
This PR adds strict typing to the output of `update` and `learn` in all
policies. This will likely be the last large refactoring PR before the
next release (0.6.0, not 1.0.0), so it requires some attention. Several
difficulties were encountered on the path to that goal:
1. The policy hierarchy is actually "broken" in the sense that the keys
of dicts that were output by `learn` did not follow the same enhancement
(inheritance) pattern as the policies. This is a real problem and should
be addressed in the near future. Generally, several aspects of the
policy design and hierarchy might deserve a dedicated discussion.
2. Each policy needs to be generic in the stats return type, because one
might want to extend it at some point and then also extend the stats.
Even within the source code base this pattern is necessary in many
places.
3. The interaction between learn and update is a bit quirky, we
currently handle it by having update modify special field inside
TrainingStats, whereas all other fields are handled by learn.
4. The IQM module is a policy wrapper and required a
TrainingStatsWrapper. The latter relies on a bunch of black magic.
They were addressed by:
1. Live with the broken hierarchy, which is now made visible by bounds
in generics. We use type: ignore where appropriate.
2. Make all policies generic with bounds following the policy
inheritance hierarchy (which is incorrect, see above). We experimented a
bit with nested TrainingStats classes, but that seemed to add more
complexity and be harder to understand. Unfortunately, mypy thinks that
the code below is wrong, wherefore we have to add `type: ignore` to the
return of each `learn`
```python
T = TypeVar("T", bound=int)
def f() -> T:
return 3
```
3. See above
4. Write representative tests for the `TrainingStatsWrapper`. Still, the
black magic might cause nasty surprises down the line (I am not proud of
it)...
Closes #933
---------
Co-authored-by: Maximilian Huettenrauch <m.huettenrauch@appliedai.de>
Co-authored-by: Michael Panchenko <m.panchenko@appliedai.de>
55 lines
1.9 KiB
Python
55 lines
1.9 KiB
Python
from typing import Any, TypeVar, cast
|
|
|
|
import numpy as np
|
|
|
|
from tianshou.data import Batch
|
|
from tianshou.data.batch import BatchProtocol
|
|
from tianshou.data.types import ActBatchProtocol, ObsBatchProtocol, RolloutBatchProtocol
|
|
from tianshou.policy import BasePolicy
|
|
from tianshou.policy.base import TrainingStats
|
|
|
|
|
|
class RandomTrainingStats(TrainingStats):
|
|
pass
|
|
|
|
|
|
TRandomTrainingStats = TypeVar("TRandomTrainingStats", bound=RandomTrainingStats)
|
|
|
|
|
|
class RandomPolicy(BasePolicy[TRandomTrainingStats]):
|
|
"""A random agent used in multi-agent learning.
|
|
|
|
It randomly chooses an action from the legal action.
|
|
"""
|
|
|
|
def forward(
|
|
self,
|
|
batch: ObsBatchProtocol,
|
|
state: dict | BatchProtocol | np.ndarray | None = None,
|
|
**kwargs: Any,
|
|
) -> ActBatchProtocol:
|
|
"""Compute the random action over the given batch data.
|
|
|
|
The input should contain a mask in batch.obs, with "True" to be
|
|
available and "False" to be unavailable. For example,
|
|
``batch.obs.mask == np.array([[False, True, False]])`` means with batch
|
|
size 1, action "1" is available but action "0" and "2" are unavailable.
|
|
|
|
:return: A :class:`~tianshou.data.Batch` with "act" key, containing
|
|
the random action.
|
|
|
|
.. seealso::
|
|
|
|
Please refer to :meth:`~tianshou.policy.BasePolicy.forward` for
|
|
more detailed explanation.
|
|
"""
|
|
mask = batch.obs.mask # type: ignore
|
|
logits = np.random.rand(*mask.shape)
|
|
logits[~mask] = -np.inf
|
|
result = Batch(act=logits.argmax(axis=-1))
|
|
return cast(ActBatchProtocol, result)
|
|
|
|
def learn(self, batch: RolloutBatchProtocol, *args: Any, **kwargs: Any) -> TRandomTrainingStats: # type: ignore
|
|
"""Since a random agent learns nothing, it returns an empty dict."""
|
|
return RandomTrainingStats() # type: ignore[return-value]
|