mujax 0.1.0__tar.gz

mujax-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

mujax-0.1.0/MANIFEST.in

@@ -0,0 +1,2 @@
+graft tests
+global-exclude __pycache__ *.py[cod]

mujax-0.1.0/PKG-INFO

@@ -0,0 +1,148 @@
+Metadata-Version: 2.4
+Name: mujax
+Version: 0.1.0
+Summary: MuZero for Gymnasium environments, built on JAX and mctx.
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/okayapinar/mujax
+Project-URL: Repository, https://github.com/okayapinar/mujax
+Project-URL: Issues, https://github.com/okayapinar/mujax/issues
+Keywords: muzero,reinforcement-learning,jax,mcts,gymnasium
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: jax>=0.4.30
+Requires-Dist: flax>=0.10.0
+Requires-Dist: optax>=0.2.0
+Requires-Dist: orbax-checkpoint>=0.11.0
+Requires-Dist: mctx>=0.0.6
+Requires-Dist: flashbax>=0.1.3
+Requires-Dist: gymnasium>=1.1.0
+Requires-Dist: numpy>=2.0.0
+Requires-Dist: tqdm
+Requires-Dist: rlax>=0.1.9
+Requires-Dist: distrax>=0.1.9
+Requires-Dist: tyro>=1.0.16
+Provides-Extra: cuda12
+Requires-Dist: jax[cuda12]>=0.4.30; extra == "cuda12"
+Provides-Extra: wandb
+Requires-Dist: wandb; extra == "wandb"
+Dynamic: license-file
+
+# mujax
+
+Gumbel MuZero and Stochastic MuZero for [Gymnasium](https://gymnasium.farama.org/) environments, built on
+JAX, [mctx](https://github.com/google-deepmind/mctx) and [flashbax](https://github.com/instadeepai/flashbax).
+
+- **Two algorithms, one pipeline**: `GMZ` (Gumbel MuZero) and `SMZ` (Stochastic MuZero with afterstates and a
+  VQ chance codebook) share replay, reanalyze, actor, learner and checkpointing.
+- **Single process**: actors and evaluator run in threads on CPU, the learner runs on GPU when available. The CPU
+  search releases the GIL, so `num_actors` actor threads (`--num-actors`) scale env throughput with the core count.
+- **Reanalyze**: a fraction of every batch is re-searched with the latest parameters.
+- **Categorical value/reward** with HL-Gauss targets in symlog space.
+- **Symlog observations**: raw observations enter the networks through `signed_logp1` (DreamerV3), so no per-dimension normalization is needed.
+- **Checkpoints**: best parameters by evaluation score plus full training state for resuming (Orbax), optional W&B upload.
+
+The design is inspired by DeepMind's [Acme](https://github.com/google-deepmind/acme): the split into actor, learner,
+replay and environment loop, the observers and the counter follow Acme's structure, reduced to a small single-process
+library.
+
+Requirements: 1-D `Box` observations and `Discrete` actions (use `gym.wrappers.FlattenObservation` otherwise).
+Illegal actions can be masked through `info["invalid_actions"]`.
+
+## Installation
+
+```bash
+pip install mujax                 # CPU
+pip install "mujax[cuda12]"       # NVIDIA GPU
+pip install "mujax[wandb]"        # Weights & Biases logging
+```
+
+## Command line
+
+```bash
+mujax --env CartPole-v1 --num-steps 20000 gmz --num-simulations  16
+mujax --env Acrobot-v1 smz --batch-size 1024
+mujax --help          # all options; `mujax gmz --help`, `mujax smz --help` for algorithm options
+```
+
+Checkpoints are written to `checkpoints/<algo>_<timestamp>/`.
+
+## Python API
+
+```python
+import gymnasium as gym
+from mujax import CheckpointingConfig, ExperimentConfig, GMZConfig, load_actor, run_experiment
+
+
+def make_env(num_envs):
+    return gym.make_vec(
+        "CartPole-v1", num_envs=num_envs, vectorization_mode="sync", vector_kwargs={"autoreset_mode": gym.vector.AutoresetMode.SAME_STEP}
+    )
+
+
+run_experiment(
+    ExperimentConfig(
+        config=GMZConfig(batch_size=256).with_num_steps(20_000),
+        environment_factory=lambda seed: make_env(16),
+        eval_environment_factory=lambda seed: make_env(1),
+        max_num_learner_steps=20_000,
+        checkpointing=CheckpointingConfig(directory="checkpoints/cartpole"),
+    )
+)
+
+actor, metadata = load_actor("checkpoints/cartpole")
+```
+
+Environments must use `AutoresetMode.SAME_STEP`. See [`examples/`](https://github.com/okayapinar/mujax/tree/master/examples) for complete training (`cartpole.py`) and evaluation (`evaluate.py`) scripts.
+
+### Logging
+
+`create_writer` sends metrics to the terminal and, optionally, Weights & Biases; learner, actor and evaluator metrics are written under `learner/`, `actor/` and `evaluator/` with `learner_steps` as the step:
+
+```python
+writer = create_writer(console=True, wandb_project="mujax")
+ExperimentConfig(..., writer=writer)
+```
+
+Any object with `write(step, scalars)`, `write_config(config)` and `close()` works as a writer (`mujax.Writer`). Without a writer only the progress bar is shown.
+
+On the command line metrics are printed to the terminal (`--no-console` turns it off); add `--wandb` (`--wandb-project`) for W&B.
+
+## Layout
+
+| Module | Contents |
+| --- | --- |
+| `config.py` | `MuZeroConfig`, shared hyperparameters |
+| `algorithm.py` | `Algorithm`, `SearchPolicy`: the contract between the infrastructure and an algorithm |
+| `algorithms/__init__.py` | `ALGORITHMS` registry, `algorithm_for` |
+| `algorithms/muzero.py` | shared math of the MuZero family: `Support`, network blocks, loss skeleton |
+| `algorithms/smz.py`, `algorithms/gmz.py` | algorithms: config, networks, search, loss |
+| `replay.py`, `reanalyze.py` | flashbax replay and reanalyze iterator |
+| `actor.py`, `learner.py`, `loop.py` | environment interaction, gradient step, environment loop |
+| `observers.py`, `loggers.py`, `checkpoint.py` | metrics, terminal and W&B writers, Orbax checkpoints |
+| `experiment.py` | `ExperimentConfig`, `run_experiment`, `load_actor` |
+
+## Development
+
+```bash
+uv sync
+uv run pytest
+```
+
+## References
+
+- Schrittwieser et al., *Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model* (2020)
+- Antonoglou et al., *Planning in Stochastic Environments with a Learned Model* (2022)
+- Danihelka et al., *Policy Improvement by Planning with Gumbel* (2022)
+- Hoffman et al., *Acme: A Research Framework for Distributed Reinforcement Learning* (2020)
+
+## License
+
+Apache-2.0

mujax-0.1.0/README.md

@@ -0,0 +1,111 @@
+# mujax
+
+Gumbel MuZero and Stochastic MuZero for [Gymnasium](https://gymnasium.farama.org/) environments, built on
+JAX, [mctx](https://github.com/google-deepmind/mctx) and [flashbax](https://github.com/instadeepai/flashbax).
+
+- **Two algorithms, one pipeline**: `GMZ` (Gumbel MuZero) and `SMZ` (Stochastic MuZero with afterstates and a
+  VQ chance codebook) share replay, reanalyze, actor, learner and checkpointing.
+- **Single process**: actors and evaluator run in threads on CPU, the learner runs on GPU when available. The CPU
+  search releases the GIL, so `num_actors` actor threads (`--num-actors`) scale env throughput with the core count.
+- **Reanalyze**: a fraction of every batch is re-searched with the latest parameters.
+- **Categorical value/reward** with HL-Gauss targets in symlog space.
+- **Symlog observations**: raw observations enter the networks through `signed_logp1` (DreamerV3), so no per-dimension normalization is needed.
+- **Checkpoints**: best parameters by evaluation score plus full training state for resuming (Orbax), optional W&B upload.
+
+The design is inspired by DeepMind's [Acme](https://github.com/google-deepmind/acme): the split into actor, learner,
+replay and environment loop, the observers and the counter follow Acme's structure, reduced to a small single-process
+library.
+
+Requirements: 1-D `Box` observations and `Discrete` actions (use `gym.wrappers.FlattenObservation` otherwise).
+Illegal actions can be masked through `info["invalid_actions"]`.
+
+## Installation
+
+```bash
+pip install mujax                 # CPU
+pip install "mujax[cuda12]"       # NVIDIA GPU
+pip install "mujax[wandb]"        # Weights & Biases logging
+```
+
+## Command line
+
+```bash
+mujax --env CartPole-v1 --num-steps 20000 gmz --num-simulations  16
+mujax --env Acrobot-v1 smz --batch-size 1024
+mujax --help          # all options; `mujax gmz --help`, `mujax smz --help` for algorithm options
+```
+
+Checkpoints are written to `checkpoints/<algo>_<timestamp>/`.
+
+## Python API
+
+```python
+import gymnasium as gym
+from mujax import CheckpointingConfig, ExperimentConfig, GMZConfig, load_actor, run_experiment
+
+
+def make_env(num_envs):
+    return gym.make_vec(
+        "CartPole-v1", num_envs=num_envs, vectorization_mode="sync", vector_kwargs={"autoreset_mode": gym.vector.AutoresetMode.SAME_STEP}
+    )
+
+
+run_experiment(
+    ExperimentConfig(
+        config=GMZConfig(batch_size=256).with_num_steps(20_000),
+        environment_factory=lambda seed: make_env(16),
+        eval_environment_factory=lambda seed: make_env(1),
+        max_num_learner_steps=20_000,
+        checkpointing=CheckpointingConfig(directory="checkpoints/cartpole"),
+    )
+)
+
+actor, metadata = load_actor("checkpoints/cartpole")
+```
+
+Environments must use `AutoresetMode.SAME_STEP`. See [`examples/`](https://github.com/okayapinar/mujax/tree/master/examples) for complete training (`cartpole.py`) and evaluation (`evaluate.py`) scripts.
+
+### Logging
+
+`create_writer` sends metrics to the terminal and, optionally, Weights & Biases; learner, actor and evaluator metrics are written under `learner/`, `actor/` and `evaluator/` with `learner_steps` as the step:
+
+```python
+writer = create_writer(console=True, wandb_project="mujax")
+ExperimentConfig(..., writer=writer)
+```
+
+Any object with `write(step, scalars)`, `write_config(config)` and `close()` works as a writer (`mujax.Writer`). Without a writer only the progress bar is shown.
+
+On the command line metrics are printed to the terminal (`--no-console` turns it off); add `--wandb` (`--wandb-project`) for W&B.
+
+## Layout
+
+| Module | Contents |
+| --- | --- |
+| `config.py` | `MuZeroConfig`, shared hyperparameters |
+| `algorithm.py` | `Algorithm`, `SearchPolicy`: the contract between the infrastructure and an algorithm |
+| `algorithms/__init__.py` | `ALGORITHMS` registry, `algorithm_for` |
+| `algorithms/muzero.py` | shared math of the MuZero family: `Support`, network blocks, loss skeleton |
+| `algorithms/smz.py`, `algorithms/gmz.py` | algorithms: config, networks, search, loss |
+| `replay.py`, `reanalyze.py` | flashbax replay and reanalyze iterator |
+| `actor.py`, `learner.py`, `loop.py` | environment interaction, gradient step, environment loop |
+| `observers.py`, `loggers.py`, `checkpoint.py` | metrics, terminal and W&B writers, Orbax checkpoints |
+| `experiment.py` | `ExperimentConfig`, `run_experiment`, `load_actor` |
+
+## Development
+
+```bash
+uv sync
+uv run pytest
+```
+
+## References
+
+- Schrittwieser et al., *Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model* (2020)
+- Antonoglou et al., *Planning in Stochastic Environments with a Learned Model* (2022)
+- Danihelka et al., *Policy Improvement by Planning with Gumbel* (2022)
+- Hoffman et al., *Acme: A Research Framework for Distributed Reinforcement Learning* (2020)
+
+## License
+
+Apache-2.0

mujax-0.1.0/mujax/__init__.py

@@ -0,0 +1,55 @@
+"""Stochastic MuZero and Gumbel MuZero, built on JAX, mctx and Gymnasium.
+
+Module map:
+    config.py      MuZeroConfig
+    algorithm.py   Algorithm, SearchPolicy: the contract between the infrastructure and an algorithm
+    algorithms/    ALGORITHMS registry and the algorithms
+        muzero.py      shared math of the MuZero family: Support, network blocks, loss skeleton
+        smz.py / gmz.py  algorithms (config, networks, search, loss)
+    replay.py / reanalyze.py  flashbax replay and reanalyze iterator
+    actor.py / learner.py / loop.py  environment interaction, gradient step, env loop
+    observers.py / loggers.py / checkpoint.py  metrics, terminal/W&B writers, orbax checkpoint
+    experiment.py  ExperimentConfig, run_experiment, load_actor
+"""
+
+from mujax.actor import Actor
+from mujax.algorithm import Algorithm
+from mujax.algorithms import ALGORITHMS, GMZ, SMZ, GMZConfig, SMZConfig, algorithm_for
+from mujax.checkpoint import CheckpointingConfig
+from mujax.config import MuZeroConfig
+from mujax.experiment import EvaluationConfig, ExperimentConfig, load_actor, run_experiment
+from mujax.learner import Learner
+from mujax.loggers import ConsoleWriter, Logger, WandbWriter, Writer, create_writer
+from mujax.loop import EnvironmentLoop
+from mujax.observers import EnvLoopObserver, InfoKeysObserver
+from mujax.types import EnvironmentSpec, make_environment_spec
+
+__version__ = "0.1.0"
+
+__all__ = [
+    "ALGORITHMS",
+    "GMZ",
+    "SMZ",
+    "Actor",
+    "Algorithm",
+    "CheckpointingConfig",
+    "ConsoleWriter",
+    "EnvLoopObserver",
+    "EnvironmentLoop",
+    "EnvironmentSpec",
+    "EvaluationConfig",
+    "ExperimentConfig",
+    "GMZConfig",
+    "InfoKeysObserver",
+    "Learner",
+    "Logger",
+    "MuZeroConfig",
+    "SMZConfig",
+    "WandbWriter",
+    "Writer",
+    "algorithm_for",
+    "create_writer",
+    "load_actor",
+    "make_environment_spec",
+    "run_experiment",
+]

mujax-0.1.0/mujax/actor.py

@@ -0,0 +1,131 @@
+from __future__ import annotations
+
+from collections.abc import Callable
+from typing import Any
+
+import jax
+import numpy as np
+
+from mujax.algorithm import SearchPolicy
+from mujax.replay import Adder
+from mujax.types import SearchOutput, TimeStep, Transition
+
+
+def invalid_actions_from_info(info: dict | None, num_envs: int, num_actions: int) -> np.ndarray:
+    """If `info["invalid_actions"]` (0/1 mask, 1 = invalid) is absent, all actions are treated as valid."""
+    raw = info.get("invalid_actions") if info else None
+    if raw is None:
+        return np.zeros((num_envs, num_actions), dtype=np.float32)
+    return np.asarray(raw, dtype=np.float32).reshape(num_envs, num_actions)
+
+
+class Actor:
+    """The env-facing side: jitted search policy + learner params + writing to replay.
+
+    Observations are (num_envs, obs_dim), actions are (num_envs,).
+
+    Args:
+        get_params: () -> (params, learner_steps); usually `Learner.get_params`.
+        replay: if given (`Buffer.adder(i)`), every step is written there (training actor).
+        value_fn: (params, obs) -> search-free value (N,); for episodes cut by the time limit,
+            to bootstrap with V(final_obs). If None, a truncated episode is treated as terminated.
+        update_period: how many `select_action` calls between params fetches.
+        per_episode_update: if True, params are fetched only at episode start (evaluator: the score
+            matches the params that played that episode).
+    """
+
+    def __init__(
+        self,
+        policy: SearchPolicy,
+        random_key: jax.Array,
+        get_params: Callable[[], tuple[Any, int]],
+        *,
+        num_actions: int,
+        replay: Adder | None = None,
+        value_fn: Callable[[Any, Any], Any] | None = None,
+        device: jax.Device | None = None,
+        update_period: int = 1,
+        per_episode_update: bool = False,
+    ) -> None:
+        self._search = jax.jit(policy)
+        self._key = random_key
+        self._get_params = get_params
+        self._num_actions = num_actions
+        self._replay = replay
+        self._value = jax.jit(value_fn) if value_fn is not None else None
+        self._device = device
+        self._update_period = max(1, update_period)
+        self._per_episode_update = per_episode_update
+        self._num_calls = 0
+        self._episode_start = True
+        self._last_search: SearchOutput | None = None
+        self._invalid_actions: np.ndarray | None = None
+        self._pull_params()
+
+    def _pull_params(self) -> None:
+        params, learner_steps = self._get_params()
+        self._params = (jax.device_put(params, self._device), learner_steps)  # single assignment
+
+    def get_params(self) -> tuple[Any, int]:
+        """The (params, learner_steps) currently playing."""
+        return self._params
+
+    def observe_first(self, observation: np.ndarray, info: dict) -> None:
+        self._invalid_actions = invalid_actions_from_info(info, len(observation), self._num_actions)
+        self._episode_start = True
+
+    def select_action(self, observation: np.ndarray) -> np.ndarray:
+        self._num_calls += 1
+        if (self._per_episode_update and self._episode_start) or (not self._per_episode_update and self._num_calls % self._update_period == 0):
+            self._pull_params()
+        self._episode_start = False
+
+        params, learner_steps = self._params
+        self._key, search_key = jax.random.split(self._key)
+        search = self._search(params, np.asarray(observation), search_key, self._invalid_actions, learner_steps)
+        self._last_search = SearchOutput(
+            action=np.asarray(search.action, dtype=np.int32),
+            policy_probs=np.asarray(search.policy_probs, dtype=np.float32),
+            value=np.asarray(search.value, dtype=np.float32),
+        )
+        return self._last_search.action
+
+    def _final_values(self, step: TimeStep, truncated: np.ndarray) -> np.ndarray:
+        """V(final_obs) for truncated envs; under SAME_STEP the final observation is in `info["final_obs"]`."""
+        values = np.zeros(len(truncated), dtype=np.float32)
+        if not truncated.any():
+            return values
+        # To call the jit with a single shape, the whole batch is filled; non-truncated rows are discarded.
+        obs = np.zeros_like(np.asarray(step.observation), dtype=np.float32)
+        for index in np.flatnonzero(truncated):
+            obs[index] = step.info["final_obs"][index]
+        values[truncated] = np.asarray(self._value(self._params[0], obs))[truncated]
+        return values
+
+    def last_search(self) -> SearchOutput | None:
+        """Search output of the last `select_action`; EnvironmentLoop puts it into the TimeStep."""
+        return self._last_search
+
+    def observe(self, step: TimeStep) -> None:
+        terminated = np.asarray(step.terminated, dtype=bool)
+        truncated = np.asarray(step.truncated, dtype=bool) & ~terminated
+        done = terminated | truncated
+        if self._replay is not None:
+            if self._value is None:  # V(final_obs) cannot be computed: a truncated episode is treated as terminated
+                terminated, truncated = done, np.zeros_like(truncated)
+            self._replay.add(
+                Transition(
+                    observation=np.asarray(step.observation, dtype=np.float32),
+                    action=self._last_search.action,
+                    reward=np.asarray(step.reward, dtype=np.float32),
+                    discount=np.where(terminated, 0.0, 1.0).astype(np.float32),
+                    policy_probs=self._last_search.policy_probs,
+                    value=self._last_search.value,
+                    invalid_actions=self._invalid_actions,
+                    truncated=truncated.astype(np.float32),
+                    bootstrap_value=self._final_values(step, truncated),
+                )
+            )
+        self._invalid_actions = invalid_actions_from_info(step.info, len(done), self._num_actions)
+        if done.any():
+            self._episode_start = True