* Add common based class for A2C and PPO agent factories
* Add default for dist_fn parameter, adding corresponding factories
* Add example mujoco_a2c_hl
* Use prefix convention (subclasses have superclass names as prefix) to
facilitate discoverability of relevant classes via IDE autocompletion
* Use dual naming, adding an alternative concise name that omits the
precise OO semantics and retains only the essential part of the name
(which can be more pleasing to users not accustomed to
convoluted OO naming)
* Created mixins for agent factories to reduce code duplication
* Further factorised params & mixins for experiment factories
* Additional parameter abstractions
* Implement high-level MuJoCo TD3 example
Closes#947
This removes all kwargs from all policy constructors. While doing that,
I also improved several names and added a whole lot of TODOs.
## Functional changes:
1. Added possibility to pass None as `critic2` and `critic2_optim`. In
fact, the default behavior then should cover the absolute majority of
cases
2. Added a function called `clone_optimizer` as a temporary measure to
support passing `critic2_optim=None`
## Breaking changes:
1. `action_space` is no longer optional. In fact, it already was
non-optional, as there was a ValueError in BasePolicy.init. So now
several examples were fixed to reflect that
2. `reward_normalization` removed from DDPG and children. It was never
allowed to pass it as `True` there, an error would have been raised in
`compute_n_step_reward`. Now I removed it from the interface
3. renamed `critic1` and similar to `critic`, in order to have uniform
interfaces. Note that the `critic` in DDPG was optional for the sole
reason that child classes used `critic1`. I removed this optionality
(DDPG can't do anything with `critic=None`)
4. Several renamings of fields (mostly private to public, so backwards
compatible)
## Additional changes:
1. Removed type and default declaration from docstring. This kind of
duplication is really not necessary
2. Policy constructors are now only called using named arguments, not a
fragile mixture of positional and named as before
5. Minor beautifications in typing and code
6. Generally shortened docstrings and made them uniform across all
policies (hopefully)
## Comment:
With these changes, several problems in tianshou's inheritance hierarchy
become more apparent. I tried highlighting them for future work.
---------
Co-authored-by: Dominik Jain <d.jain@appliedai.de>
Bumps [pillow](https://github.com/python-pillow/Pillow) from 10.0.0 to
10.0.1.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a
href="https://github.com/python-pillow/Pillow/releases">pillow's
releases</a>.</em></p>
<blockquote>
<h2>10.0.1</h2>
<p><a
href="https://pillow.readthedocs.io/en/stable/releasenotes/10.0.1.html">https://pillow.readthedocs.io/en/stable/releasenotes/10.0.1.html</a></p>
<h2>Changes</h2>
<ul>
<li>Updated libwebp to 1.3.2 <a
href="https://redirect.github.com/python-pillow/Pillow/issues/7395">#7395</a>
[<a
href="https://github.com/radarhere"><code>@radarhere</code></a>]</li>
<li>Updated zlib to 1.3 <a
href="https://redirect.github.com/python-pillow/Pillow/issues/7344">#7344</a>
[<a
href="https://github.com/radarhere"><code>@radarhere</code></a>]</li>
</ul>
</blockquote>
</details>
<details>
<summary>Changelog</summary>
<p><em>Sourced from <a
href="https://github.com/python-pillow/Pillow/blob/main/CHANGES.rst">pillow's
changelog</a>.</em></p>
<blockquote>
<h2>10.0.1 (2023-09-15)</h2>
<ul>
<li>
<p>Updated libwebp to 1.3.2 <a
href="https://redirect.github.com/python-pillow/Pillow/issues/7395">#7395</a>
[radarhere]</p>
</li>
<li>
<p>Updated zlib to 1.3 <a
href="https://redirect.github.com/python-pillow/Pillow/issues/7344">#7344</a>
[radarhere]</p>
</li>
</ul>
</blockquote>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a
href="e34d346f10"><code>e34d346</code></a>
Updated order</li>
<li><a
href="a62f2402a6"><code>a62f240</code></a>
10.0.1 version bump</li>
<li><a
href="d50250d9ea"><code>d50250d</code></a>
Added release notes for 10.0.1</li>
<li><a
href="b4c7d4b8b2"><code>b4c7d4b</code></a>
Update CHANGES.rst [ci skip]</li>
<li><a
href="730f74600e"><code>730f746</code></a>
Updated libwebp to 1.3.2</li>
<li><a
href="b0e28048d6"><code>b0e2804</code></a>
Updated zlib to 1.3</li>
<li>See full diff in <a
href="https://github.com/python-pillow/Pillow/compare/10.0.0...10.0.1">compare
view</a></li>
</ul>
</details>
<br />
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot merge` will merge this PR after your CI passes on it
- `@dependabot squash and merge` will squash and merge this PR after
your CI passes on it
- `@dependabot cancel merge` will cancel a previously requested merge
and block automerging
- `@dependabot reopen` will reopen this PR if it is closed
- `@dependabot close` will close this PR and stop Dependabot recreating
it. You can achieve the same result by closing it manually
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop
Dependabot creating any more for this major version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop
Dependabot creating any more for this minor version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop
Dependabot creating any more for this dependency (unless you reopen the
PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the
[Security Alerts
page](https://github.com/thu-ml/tianshou/network/alerts).
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Bumps [urllib3](https://github.com/urllib3/urllib3) from 1.26.16 to
1.26.17.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a
href="https://github.com/urllib3/urllib3/releases">urllib3's
releases</a>.</em></p>
<blockquote>
<h2>1.26.17</h2>
<ul>
<li>Added the <code>Cookie</code> header to the list of headers to strip
from requests when redirecting to a different host. As before, different
headers can be set via <code>Retry.remove_headers_on_redirect</code>.
(GHSA-v845-jxx5-vc9f)</li>
</ul>
</blockquote>
</details>
<details>
<summary>Changelog</summary>
<p><em>Sourced from <a
href="https://github.com/urllib3/urllib3/blob/main/CHANGES.rst">urllib3's
changelog</a>.</em></p>
<blockquote>
<h1>1.26.17 (2023-10-02)</h1>
<ul>
<li>Added the <code>Cookie</code> header to the list of headers to strip
from requests when redirecting to a different host. As before, different
headers can be set via <code>Retry.remove_headers_on_redirect</code>.
(<code>[#3139](https://github.com/urllib3/urllib3/issues/3139)
<https://github.com/urllib3/urllib3/pull/3139></code>_)</li>
</ul>
</blockquote>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a
href="c9016bf464"><code>c9016bf</code></a>
Release 1.26.17</li>
<li><a
href="01220354d3"><code>0122035</code></a>
Backport GHSA-v845-jxx5-vc9f (<a
href="https://redirect.github.com/urllib3/urllib3/issues/3139">#3139</a>)</li>
<li><a
href="e63989f97d"><code>e63989f</code></a>
Fix installing <code>brotli</code> extra on Python 2.7</li>
<li><a
href="2e7a24d087"><code>2e7a24d</code></a>
[1.26] Configure OS for RTD to fix building docs</li>
<li><a
href="57181d6ea9"><code>57181d6</code></a>
[1.26] Improve error message when calling urllib3.request() (<a
href="https://redirect.github.com/urllib3/urllib3/issues/3058">#3058</a>)</li>
<li><a
href="3c0148048a"><code>3c01480</code></a>
[1.26] Run coverage even with failed jobs</li>
<li>See full diff in <a
href="https://github.com/urllib3/urllib3/compare/1.26.16...1.26.17">compare
view</a></li>
</ul>
</details>
<br />
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot merge` will merge this PR after your CI passes on it
- `@dependabot squash and merge` will squash and merge this PR after
your CI passes on it
- `@dependabot cancel merge` will cancel a previously requested merge
and block automerging
- `@dependabot reopen` will reopen this PR if it is closed
- `@dependabot close` will close this PR and stop Dependabot recreating
it. You can achieve the same result by closing it manually
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop
Dependabot creating any more for this major version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop
Dependabot creating any more for this minor version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop
Dependabot creating any more for this dependency (unless you reopen the
PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the
[Security Alerts
page](https://github.com/thu-ml/tianshou/network/alerts).
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
- [X] I have marked all applicable categories:
+ [ ] exception-raising fix
+ [ ] algorithm implementation fix
+ [ ] documentation modification
+ [X] new feature
- [X] I have reformatted the code using `make format` (**required**)
- [X] I have checked the code using `make commit-checks` (**required**)
- [X] If applicable, I have mentioned the relevant/related issue(s)
- [X] If applicable, I have listed every items in this Pull Request
below
Close#941
rtfd build link:
https://readthedocs.org/projects/tianshou/builds/22019877/
Also -- fix two small issues reported by users, see #928 and #930
Note: I created the branch in thu-ml:tianshou instead of
Trinkle23897:tianshou to quickly check the rtfd build. It's not a good
process since every commit would trigger twice CI pipelines :(
Closes#914
Additional changes:
- Deprecate python below 11
- Remove 3rd party and throughput tests. This simplifies install and
test pipeline
- Remove gym compatibility and shimmy
- Format with 3.11 conventions. In particular, add `zip(...,
strict=True/False)` where possible
Since the additional tests and gym were complicating the CI pipeline
(flaky and dist-dependent), it didn't make sense to work on fixing the
current tests in this PR to then just delete them in the next one. So
this PR changes the build and removes these tests at the same time.
Preparation for #914 and #920
Changes formatting to ruff and black. Remove python 3.8
## Additional Changes
- Removed flake8 dependencies
- Adjusted pre-commit. Now CI and Make use pre-commit, reducing the
duplication of linting calls
- Removed check-docstyle option (ruff is doing that)
- Merged format and lint. In CI the format-lint step fails if any
changes are done, so it fulfills the lint functionality.
---------
Co-authored-by: Jiayi Weng <jiayi@openai.com>
# Goals of the PR
The PR introduces **no changes to functionality**, apart from improved
input validation here and there. The main goals are to reduce some
complexity of the code, to improve types and IDE completions, and to
extend documentation and block comments where appropriate. Because of
the change to the trainer interfaces, many files are affected (more
details below), but still the overall changes are "small" in a certain
sense.
## Major Change 1 - BatchProtocol
**TL;DR:** One can now annotate which fields the batch is expected to
have on input params and which fields a returned batch has. Should be
useful for reading the code. getting meaningful IDE support, and
catching bugs with mypy. This annotation strategy will continue to work
if Batch is replaced by TensorDict or by something else.
**In more detail:** Batch itself has no fields and using it for
annotations is of limited informational power. Batches with fields are
not separate classes but instead instances of Batch directly, so there
is no type that could be used for annotation. Fortunately, python
`Protocol` is here for the rescue. With these changes we can now do
things like
```python
class ActionBatchProtocol(BatchProtocol):
logits: Sequence[Union[tuple, torch.Tensor]]
dist: torch.distributions.Distribution
act: torch.Tensor
state: Optional[torch.Tensor]
class RolloutBatchProtocol(BatchProtocol):
obs: torch.Tensor
obs_next: torch.Tensor
info: Dict[str, Any]
rew: torch.Tensor
terminated: torch.Tensor
truncated: torch.Tensor
class PGPolicy(BasePolicy):
...
def forward(
self,
batch: RolloutBatchProtocol,
state: Optional[Union[dict, Batch, np.ndarray]] = None,
**kwargs: Any,
) -> ActionBatchProtocol:
```
The IDE and mypy are now very helpful in finding errors and in
auto-completion, whereas before the tools couldn't assist in that at
all.
## Major Change 2 - remove duplication in trainer package
**TL;DR:** There was a lot of duplication between `BaseTrainer` and its
subclasses. Even worse, it was almost-duplication. There was also
interface fragmentation through things like `onpolicy_trainer`. Now this
duplication is gone and all downstream code was adjusted.
**In more detail:** Since this change affects a lot of code, I would
like to explain why I thought it to be necessary.
1. The subclasses of `BaseTrainer` just duplicated docstrings and
constructors. What's worse, they changed the order of args there, even
turning some kwargs of BaseTrainer into args. They also had the arg
`learning_type` which was passed as kwarg to the base class and was
unused there. This made things difficult to maintain, and in fact some
errors were already present in the duplicated docstrings.
2. The "functions" a la `onpolicy_trainer`, which just called the
`OnpolicyTrainer.run`, not only introduced interface fragmentation but
also completely obfuscated the docstring and interfaces. They themselves
had no dosctring and the interface was just `*args, **kwargs`, which
makes it impossible to understand what they do and which things can be
passed without reading their implementation, then reading the docstring
of the associated class, etc. Needless to say, mypy and IDEs provide no
support with such functions. Nevertheless, they were used everywhere in
the code-base. I didn't find the sacrifices in clarity and complexity
justified just for the sake of not having to write `.run()` after
instantiating a trainer.
3. The trainers are all very similar to each other. As for my
application I needed a new trainer, I wanted to understand their
structure. The similarity, however, was hard to discover since they were
all in separate modules and there was so much duplication. I kept
staring at the constructors for a while until I figured out that
essentially no changes to the superclass were introduced. Now they are
all in the same module and the similarities/differences between them are
much easier to grasp (in my opinion)
4. Because of (1), I had to manually change and check a lot of code,
which was very tedious and boring. This kind of work won't be necessary
in the future, since now IDEs can be used for changing signatures,
renaming args and kwargs, changing class names and so on.
I have some more reasons, but maybe the above ones are convincing
enough.
## Minor changes: improved input validation and types
I added input validation for things like `state` and `action_scaling`
(which only makes sense for continuous envs). After adding this, some
tests failed to pass this validation. There I added
`action_scaling=isinstance(env.action_space, Box)`, after which tests
were green. I don't know why the tests were green before, since action
scaling doesn't make sense for discrete actions. I guess some aspect was
not tested and didn't crash.
I also added Literal in some places, in particular for
`action_bound_method`. Now it is no longer allowed to pass an empty
string, instead one should pass `None`. Also here there is input
validation with clear error messages.
@Trinkle23897 The functional tests are green. I didn't want to fix the
formatting, since it will change in the next PR that will solve #914
anyway. I also found a whole bunch of code in `docs/_static`, which I
just deleted (shouldn't it be copied from the sources during docs build
instead of committed?). I also haven't adjusted the documentation yet,
which atm still mentions the trainers of the type
`onpolicy_trainer(...)` instead of `OnpolicyTrainer(...).run()`
## Breaking Changes
The adjustments to the trainer package introduce breaking changes as
duplicated interfaces are deleted. However, it should be very easy for
users to adjust to them
---------
Co-authored-by: Michael Panchenko <m.panchenko@appliedai.de>
At the moment, WandbLogger is always using wandb.init with monitor_gym =
True.
This fails when OpenAI's gym is not installed, which doesn't make sense
after the transition to Gymnasium.
I am using Tianshou with non-standard RL environment, which adhere to
Gymnasium API, and the current code is throwing exceptions.
I suggest to make it a controllable parameter. I left the default value
to True (to make it functionally the same for people using gym). It may
also make sense to change the default to False.
**The assert was missing a description, I fixed it.**
Please note: there is an error in the documentations, but it does not
seem to be related to my changes.
- Fix the current bug discussed in #844 in `test_ppo.py`.
- Add warning for `ActorProb ` if both `max_action ` and
`unbounded=True` are used for model initializations.
- Add warning for PGpolicy and DDPGpolicy if they find duplicate usage
of action-bounded actor and action scaling method.
Specifically, BasePolicy.set_eps seems to be a remnant from using DQN in
other examples.
* Removed unused functions (test_fn and train_fn) from the pettingzoo
example with PistonBall. These functions use set_eps which is not
available for PPO and is not even called once in the file.
- [ ] I have marked all applicable categories:
+ [x] exception-raising fix
+ [ ] algorithm implementation fix
+ [ ] documentation modification
+ [ ] new feature
- [ ] I have reformatted the code using `make format` (**required**)
- [ ] I have checked the code using `make commit-checks` (**required**)
- [ ] If applicable, I have mentioned the relevant/related issue(s)
- [ ] If applicable, I have listed every items in this Pull Request
below
I'm developing a new PettingZoo environment. It is a two players turns
board game.
```
obs_space = dict(
board = gym.spaces.MultiBinary([8, 8]),
player = gym.spaces.Tuple([gym.spaces.Discrete(8)] * 2),
other_player = gym.spaces.Tuple([gym.spaces.Discrete(8)] * 2)
)
self._observation_space = gym.spaces.Dict(spaces=obs_space)
self._action_space = gym.spaces.Tuple([gym.spaces.Discrete(8)] * 2)
...
# this cache ensures that same space object is returned for the same
agent
# allows action space seeding to work as expected
@functools.lru_cache(maxsize=None)
def observation_space(self, agent):
# gymnasium spaces are defined and documented here:
https://gymnasium.farama.org/api/spaces/
return self._observation_space
@functools.lru_cache(maxsize=None)
def action_space(self, agent):
return self._action_space
```
My test is:
```
def test_with_tianshou():
action = None
# env = gym.make('qwertyenv/CollectCoins-v0', pieces=['rock', 'rock'])
env = CollectCoinsEnv(pieces=['rock', 'rock'], with_mask=True)
def another_action_taken(action_taken):
nonlocal action
action = action_taken
# Wrapping the original environment as to make sure a valid action will
be taken.
env = EnsureValidAction(
env,
env.check_action_valid,
env.provide_alternative_valid_action,
another_action_taken
)
env = PettingZooEnv(env)
policies = MultiAgentPolicyManager([RandomPolicy(), RandomPolicy()],
env)
env = DummyVectorEnv([lambda: env])
collector = Collector(policies, env)
result = collector.collect(n_step=200, render=0.1)
```
I have also a wrapper that may be redundant as of Tianshou capability to action_mask, yet it is still part of the code:
```
from typing import TypeVar, Callable
import gymnasium as gym
from pettingzoo.utils.wrappers import BaseWrapper
Action = TypeVar("Action")
class ActionWrapper(BaseWrapper):
def __init__(self, env: gym.Env):
super().__init__(env)
def step(self, action):
action = self.action(action)
self.env.step(action)
def action(self, action):
pass
def render(self, *args, **kwargs):
self.env.render(*args, **kwargs)
class EnsureValidAction(ActionWrapper):
"""
A gym environment wrapper to help with the case that the agent wants to
take invalid actions.
For example consider a Chess game, where you let the action_space be any
piece moving to any square on the board,
but then when a wrong move is taken, instead of returing a big negative
reward, you just take another action,
this time a valid one. To make sure the learning algorithm is aware of
the action taken, a callback should be provided.
"""
def __init__(self, env: gym.Env,
check_action_valid: Callable[[Action], bool],
provide_alternative_valid_action: Callable[[Action], Action],
alternative_action_cb: Callable[[Action], None]):
super().__init__(env)
self.check_action_valid = check_action_valid
self.provide_alternative_valid_action = provide_alternative_valid_action
self.alternative_action_cb = alternative_action_cb
def action(self, action: Action) -> Action:
if self.check_action_valid(action):
return action
alternative_action = self.provide_alternative_valid_action(action)
self.alternative_action_cb(alternative_action)
return alternative_action
```
To make above work I had to patch a bit PettingZoo (opened a pull-request there), and a small patch here (this PR).
Maybe I'm doing something wrong, yet I fail to see it.
With my both fixes of PZ and of Tianshou, I have two tests, one of the environment by itself, and the other as of above.
