reinfors

KO | EN

gitlite — search

#http #hacktoberfest #ai-agents #self-hosted #devops #nextjs #tensorflow #html #mcp #rust

(★ 9)

High-throughput reinforcement-learning search and sampling in Rust, with caller-owned Python networks and training

파일 탐색기

ZIP 다운로드

- - bug_report.yml
  - config.yml
- - ci.yml
  - docs.yml
  - miri.yml
  - release.yml
  - verify-release.yml
- carracing-throughput.gif
- reinfors-banner.svg
- - Cargo.toml
  - LICENSE-APACHE
  - LICENSE-MIT
- - Cargo.toml
  - LICENSE-APACHE
  - LICENSE-MIT
- - build.rs
  - Cargo.toml
  - LICENSE-APACHE
  - LICENSE-MIT
- - interactive-connect4.svg
- - algorithms.md
  - compatibility.md
  - games.md
  - index.md
- - architecture.md
  - engine-collection.md
  - index.md
  - sampling-and-training.md
- - documentation.md
  - setup.md
- - index.md
- - component-contracts.md
  - encoders.md
  - index.md
  - policies-and-learners.md
  - python-bindings.md
  - rust-components.md
- - adapters.md
  - arena.md
  - configuration-and-checkpoints.md
  - evaluation.md
  - index.md
  - streaming.md
  - telemetry.md
  - training.md
- - batch-formats.md
  - glossary.md
  - index.md
  - inference-contract.md
  - limits.md
  - python-api.md
  - references.md
  - telemetry-fields.md
  - troubleshooting.md
- getting-started.md
- index.md
- eval_arena.py
- eval_az_h2h.py
- play_connect4.py
- quickstart.py
- solve_leduc.py
- streaming.py
- telemetry_tensorboard.py
- train_alphazero_example.py
- train_alphazero_snake.py
- train_deep_cfr.py
- train_gridworld.py
- train_ppo_carracing.py
- train_ppo_connect4.py
- train_rainbow_holdem.py
- train_treestrap_snake.py
- - __init__.py
  - _reinfors.pyi
  - arena.py
  - catalog.py
  - chance_modes.py
  - encoders.py
  - games.py
  - gym.py
  - learners.py
  - noise.py
  - policies.py
  - py.typed
  - solvers.py
  - spaces.py
  - starts.py
- - cozy-chess.LICENSE
  - parry2d.LICENSE
  - profiling.LICENSE
  - rapier2d.LICENSE
- bench_carracing_throughput.py
- block-main-push.sh
- car_racing_fixture.py
- check_docs.py
- gen_score_font.py
- gen_third_party_notices.py
- generate_docs.py
- conftest.py
- test_alphazero.py
- test_arena.py
- test_arena_example.py
- test_backgammon.py
- test_build_info.py
- test_catalog.py
- test_cfr.py
- test_cfr_parity.py
- test_chance_mode.py
- test_chess.py
- test_collect_stream.py
- test_constructor_validation.py
- test_deep_cfr.py
- test_engine.py
- test_engine_snapshots.py
- test_env.py
- test_env_snapshots.py
- test_example.py
- test_games.py
- test_gym.py
- test_holdem.py
- test_holdem_parity.py
- test_infer_cache.py
- test_infer_dtype.py
- test_kuhn_n.py
- test_legal_masking.py
- test_method_validation.py
- test_minimax.py
- test_n_snake.py
- test_pad.py
- test_per_player_infer.py
- test_policy_choose.py
- test_ppo.py
- test_resolved_config.py
- test_scheduler.py
- test_small_poker.py
- test_smoke.py
- test_third_party_notices.py
- test_train_deep_cfr.py
- test_train_rainbow_holdem.py
- test_zero_copy.py
.gitignore
.pre-commit-config.yaml
Cargo.lock
Cargo.toml
CITATION.cff
CONTRIBUTING.md
LICENSE-APACHE
LICENSE-MIT
mkdocs.yml
pyproject.toml
pyrightconfig.json
README.md
SECURITY.md
THIRD-PARTY-NOTICES

# CDN으로 사용하기

jsDelivr

jsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.

버전

파일

링크

예시

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

High-throughput reinforcement-learning search and sampling in Rust, with caller-owned
Python networks and training.

Reinfors runs game dynamics, search, episode orchestration, and batch assembly in a
parallel Rust backend. Your inference callback is the boundary: it receives pooled NumPy
observations and returns model outputs, so the network, framework, optimizer, replay,
hardware placement, and distributed topology remain yours.

[Game semantics](https://jeepjeepjeep.github.io/reinfors/catalogue/games/) cover single- and multi-agent, zero-sum, cooperative, and
general-sum tasks, with turn-taking or simultaneous decisions, explicit chance, and perfect
or imperfect information. [Algorithms](https://jeepjeepjeep.github.io/reinfors/catalogue/algorithms/) span policy-driven value learning,
search-guided learning, and standalone game-theoretic solving.

```python
import numpy as np
import reinfors as rf

game = rf.games.Connect4()
engine = rf.Engine(
    game=game,
    reward=rf.Reward(win=1.0, loss=-1.0),
    policy=rf.policies.Minimax(depth=4),
    learner=rf.learners.TreeStrap(),
    n_games=32,  # parallel episode slots
    seed=0,
)

actions = game.action_space().n

def infer(obs: np.ndarray) -> np.ndarray:
    # pooled across games. Replace with PyTorch, JAX, an accelerator service, or any
    # other backend; TreeStrap's targets train it toward its own deeper search results.
    return np.zeros((len(obs), 1, actions), dtype=np.float32)

batch = engine.collect(n_records=1024, infer=infer)
print(batch.obs.shape, batch.targets.shape, batch.telemetry)
```

## Performance

Reinfors' native backend makes it much more performant than Python-loop RL
libraries such as Gymnasium. Benchmarking reinfors' `car_racing` port against
Gymnasium's `CarRacing-v3` — separate implementations of the same design, with
matched action space and 96x96 RGB observation content, stepped in
single-threaded loops — reinfors runs **~20x** more environment steps per second
(medians of three 30s trials):

| Steps/s, single-threaded | Apple M1 Max | AMD EPYC 7R32 (g5.2xlarge, SMT off, pinned) |
| --- | --- | --- |
| reinfors `car_racing` | 3,850 | 2,069 |
| Gymnasium `CarRacing-v3` | 195 | 148 |
| **speedup** | **19.7x** | **14.0x** |

Reproduce with [`scripts/bench_carracing_throughput.py`](scripts/bench_carracing_throughput.py), which reports medians,
ranges, and machine provenance.

Parallelising multiplies this further, and in reinfors it is trivial — set
`n_threads` — while [`collect_stream`](https://jeepjeepjeep.github.io/reinfors/guides/streaming/) overlaps collection with training as the
normal operating mode:

Benchmarked against OpenSpiel's all-C++ libtorch AlphaZero on a matched chess training
workload — two-hour rounds at each stack's best measured configuration, then
head-to-head play between the resulting models — reinfors sustained **9.9% higher
training throughput**, and its trained networks scored **0.605 ± 0.020** against
OpenSpiel's over 300 paired games (**+74 Elo**). Protocol, evidence, and full results:
[reinfors-benchmarks](https://github.com/jeepjeepjeep/reinfors-benchmarks).

Reinfors is not designed to maximize throughput at any cost or to outperform a bespoke, fully
fused JAX/XLA pipeline on the fixed workload it specializes for. It targets a practical balance:
native simulation, search, and batching, while games, algorithms, networks, and deployment remain
modular enough for broad experimentation.

## Install

```bash
pip install reinfors
```

Optional adapters and training dependencies are separate:

```bash
pip install "reinfors[gym]"   # Gymnasium and PettingZoo adapters
pip install "reinfors[train]" # PyTorch examples
```

Contributing or building from source? Start with the
[contributing guide](https://github.com/jeepjeepjeep/reinfors/blob/main/CONTRIBUTING.md) and the
[development setup guide](https://jeepjeepjeep.github.io/reinfors/development/setup/).

## Where next?

- [Get started](https://jeepjeepjeep.github.io/reinfors/getting-started/)
- [Understand sampling and injectable training](https://jeepjeepjeep.github.io/reinfors/concepts/sampling-and-training/)
- [Choose a game](https://jeepjeepjeep.github.io/reinfors/catalogue/games/), [algorithm](https://jeepjeepjeep.github.io/reinfors/catalogue/algorithms/), or
  [built-in composition](https://jeepjeepjeep.github.io/reinfors/catalogue/compatibility/)
- [Run the examples](https://jeepjeepjeep.github.io/reinfors/examples/)
- [Evaluate searched agents with Arena](https://jeepjeepjeep.github.io/reinfors/guides/arena/)
- [Add a Rust game or algorithm](https://jeepjeepjeep.github.io/reinfors/extending/)
- [Read the complete documentation](https://jeepjeepjeep.github.io/reinfors/)

## Stability

reinfors is pre-1.0: **any 0.x release may change any API, behavior, or serialized format**
(including snapshot and config layouts) without deprecation. Pin an exact version
(`reinfors==0.x.y`) and read release notes when upgrading. What does hold at every version:
constructors validate their inputs, and no public Python input reaches a Rust panic — both
enforced by adversarial test sweeps in CI.

## Citation

If you use reinfors in your research, cite it via the repository's
[CITATION.cff](https://github.com/jeepjeepjeep/reinfors/blob/main/CITATION.cff) (GitHub's "Cite this repository" button renders it
as BibTeX/APA). What reinfors itself builds on is catalogued in
[References](https://jeepjeepjeep.github.io/reinfors/reference/references/).

## License

Licensed under either of [MIT](https://github.com/jeepjeepjeep/reinfors/blob/main/LICENSE-MIT) or [Apache-2.0](https://github.com/jeepjeepjeep/reinfors/blob/main/LICENSE-APACHE), at your
option. Unless you explicitly state otherwise, any contribution intentionally submitted
for inclusion in this work shall be dual-licensed as above, without any additional terms
or conditions.

reinfors

파일 탐색기

# CDN으로 사용하기

Was this content helpful?

Recent Feedback

Download README