mymyr
A data-oriented, lifted-first rewrite of the mimir classical-planning core, with free-threaded Python bindings and an optional CUDA backend.
mymyr parses PDDL, searches for plans, builds state spaces and provides batched reinforcement-learning environments and relational state encodings for NumPy, JAX and PyTorch. Native work releases the GIL. Free-threaded Python also lets Python callbacks and concurrent Python callers run in parallel; on a GIL interpreter, Python callbacks take turns on the GIL.
Install from PyPI
With uv installed:
uv venv .venv --python 3.14t
uv pip install --python .venv/bin/python mymyr
CPU wheels support CPython 3.14 and 3.15, both GIL (3.14, 3.15) and free-threaded (3.14t, 3.15t), on Linux
x86_64 with glibc 2.34 or newer and macOS arm64 with macOS 15 or newer. Change --python to select an interpreter.
Wheels include the PDDL parser and C++ SDK. For Gymnasium environments, install mymyr[gym]; install JAX or PyTorch
separately when using those integrations.
Releases after 0.1.0 attach CUDA wheels for free-threaded CPython 3.14 and 3.15 on Linux x86_64 to their
GitHub Release, with the version <version>+cu13 (v0.1.0 has CPU
wheels only). They need an NVIDIA driver 580 or newer and carry GPU code for compute capabilities 8.6, 8.9 and 9.0.
With <version> the release's version:
uv pip install --python .venv/bin/python "mymyr==<version>+cu13" \
--find-links https://github.com/maichmueller/mymyr/releases/expanded_assets/v<version>
See Installing a CUDA wheel for other GPUs and source builds.
Pushing a v* tag that matches the package version publishes to TestPyPI and PyPI; a manual release workflow run is
a TestPyPI rehearsal with a unique .devN version. The CUDA wheels of a release are built on a large Linux host
with ci/local_cuda_wheels.sh (a manual run of the CUDA wheel workflow builds them where there is none), and
ci/attach_cuda_wheels.sh tests them on a GPU and attaches them to the GitHub Release.
Try it
Run this with .venv/bin/python; it needs no task files:
import mymyr
from mymyr import rl, search
domain = mymyr.Domain.from_string("""
(define (domain switch)
(:requirements :strips)
(:predicates (ready) (done))
(:action finish
:parameters ()
:precondition (ready)
:effect (and (not (ready)) (done))))
""")
problem = domain.instantiate_string("""
(define (problem example) (:domain switch)
(:init (ready))
(:goal (done)))
""")
task = mymyr.Task(problem)
result = search.astar(task, heuristic="max")
assert result.solved and len(result.plan) == 1 and result.cost == 1
print(search.format_plan(task, result.plan))
pool = rl.CpuEnvPool(task, num_envs=4, seed=0)
pool.reset()
batch = pool.step()
assert batch.terminated.all()
print(batch.states.shape, batch.reward, batch.terminated)
For your own task files, use mymyr.Task.from_pddl("domain.pddl", "problem.pddl").
Build from source
Source builds need CPython 3.14 or newer, a C++26 compiler, CMake 3.25 or newer and Ninja. The commands below run
from a source checkout on Linux x86_64 with uv, Git, curl and sha256sum on PATH. The toolchain script downloads
GCC 16.1 from conda-forge using micromamba and builds the pinned dependencies (loki, Abseil, fmt, Boost
headers and googletest for the C++ tests) under build/toolchain; no system compiler installation is needed.
uv venv .venv --python 3.14t
uv pip install --python .venv/bin/python "cmake>=3.25" ninja
export PATH="$PWD/.venv/bin:$PATH"
export CMAKE_BUILD_PARALLEL_LEVEL=8
bash ci/wheel_toolchain.sh "$PWD/build/toolchain"
export CC="$PWD/build/toolchain/gcc/bin/x86_64-conda-linux-gnu-gcc"
export CXX="$PWD/build/toolchain/gcc/bin/x86_64-conda-linux-gnu-g++"
export CMAKE_ARGS="-DMYMYR_DEPS_PREFIX=$PWD/build/toolchain/deps"
uv pip install --python .venv/bin/python -e ".[test]"
.venv/bin/python -m pytest -W error tests/python
An existing GCC 16 build also works: set CC and CXX to its gcc and g++, build the
dependencies with that compiler, and set MYMYR_DEPS_PREFIX through CMAKE_ARGS to
their install prefix. -DMYMYR_FRONTEND=OFF skips the parser and its dependencies. On macOS use a C++26-capable
Apple Clang/libc++ toolchain; see the C++ package guide for native builds and SDK consumers.
For CUDA, follow the existing-GCC route above with a system or source-built GCC 16 and dependencies built with
that compiler. CI uses a system GCC 16 as nvcc's host compiler. With CC and CXX selecting that compiler and
CMAKE_ARGS pointing to its dependency prefix, put CUDA 13.4 or newer's nvcc on PATH, then rebuild with CMake
4.4 or newer:
uv pip install --python .venv/bin/python "cmake>=4.4"
CMAKE_ARGS="$CMAKE_ARGS -DMYMYR_CUDA=ON -DCMAKE_CUDA_COMPILER=$(command -v nvcc) -DCMAKE_CUDA_HOST_COMPILER=$CXX" \
uv pip install --python .venv/bin/python --reinstall-package mymyr -e ".[test]"
The default CUDA architecture is the visible GPU. NVIDIA drivers 580 or newer can run matching SASS; PTX needs a driver that understands the toolkit's PTX version. See CUDA for architecture selection and backend limits. Reference-comparison tests need the separate data described in tests/parity/README.md.
Documentation
Start with the user guides, API conventions and runnable examples. See parsing and task options, search, RL and Gymnasium, relational encodings, the CLI, the C API and the C++ SDK.
License
mymyr is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License
as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version
(SPDX: GPL-3.0-or-later; the full text is in LICENSE). These are the terms of mimir (Copyright (C) 2023 Dominik Drexler and Simon Stahlberg), from which mymyr is derived.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
Metadata
Release files for mymyr 0.2.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| mymyr-0.2.0.tar.gz | 5.0 MB | With provenance | Details for mymyr-0.2.0.tar.gz |
Built distributions (wheels)
Total release size: 105.8 MB
Release files / mymyr-0.2.0.tar.gz
| Download URL | mymyr-0.2.0.tar.gz |
|---|---|
| Size | 5.0 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
fe33438c997b21939da8a1fbaac180b6d9430e41f3fb015a63ea9a9d0274d820
|
|
BLAKE2b-256 checksum How to use checksums |
977e921ef7bc7a2defb0272bb66f2cec676f69661f6e2af9eda5a33af205dfd8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp315-cp315t-manylinux_2_34_x86_64.whl
| Download URL | mymyr-0.2.0-cp315-cp315t-manylinux_2_34_x86_64.whl |
|---|---|
| Size | 14.6 MB |
| Tags | CPython 3.15 CPython 3.15 free-threading Linux glibc 2.34+ x86-64 |
|
SHA-256 checksum How to use checksums |
fc98ab69ede0baa310f82422f410f1e468a6c38aaa8fb0bc529a253045720071
|
|
BLAKE2b-256 checksum How to use checksums |
fdc8a011400c1f4f33fce04b42ee5dbb22c00dd987ab32be296bb426b1e803bc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp315-cp315t-macosx_15_0_arm64.whl
| Download URL | mymyr-0.2.0-cp315-cp315t-macosx_15_0_arm64.whl |
|---|---|
| Size | 10.6 MB |
| Tags | CPython 3.15 CPython 3.15 free-threading macOS 15.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
a0846bd0920446215cd2c683da9b14fe753837cbd26da3e7b582e6c4273f369c
|
|
BLAKE2b-256 checksum How to use checksums |
a87ca963354f9eb531d8dd3514b31b4a9f0f29fe9392573e8a10f213480fa8c5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp315-cp315-manylinux_2_34_x86_64.whl
| Download URL | mymyr-0.2.0-cp315-cp315-manylinux_2_34_x86_64.whl |
|---|---|
| Size | 14.6 MB |
| Tags | CPython 3.15 Linux glibc 2.34+ x86-64 |
|
SHA-256 checksum How to use checksums |
c761a74d2de50dea76f625bbac01c351daf6db126f57cdc7048040f7bad907d7
|
|
BLAKE2b-256 checksum How to use checksums |
9b55e1ed780fa518f057692889995cb9744b8746174121b73fa7ad832d517ff1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp315-cp315-macosx_15_0_arm64.whl
| Download URL | mymyr-0.2.0-cp315-cp315-macosx_15_0_arm64.whl |
|---|---|
| Size | 10.6 MB |
| Tags | CPython 3.15 macOS 15.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
ba21480e9cba5d377d419e3b03d6ad741e3422fde03bd51028e35ffa02ab8a4a
|
|
BLAKE2b-256 checksum How to use checksums |
ffb7107f4ccac422f9d048e0f623723194f18e7de61e3c63868c3560a8680cc4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp314-cp314t-manylinux_2_34_x86_64.whl
| Download URL | mymyr-0.2.0-cp314-cp314t-manylinux_2_34_x86_64.whl |
|---|---|
| Size | 14.6 MB |
| Tags | CPython 3.14 CPython 3.14 free-threading Linux glibc 2.34+ x86-64 |
|
SHA-256 checksum How to use checksums |
b8c89fa2f3cafe7c43007532b4da4aa7cd3757d00718950ddb751f97b1970a6a
|
|
BLAKE2b-256 checksum How to use checksums |
2bc316b1779469a701735604e490bd28c8ce3ee07f7ba04af380c65280f24e42
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp314-cp314t-macosx_15_0_arm64.whl
| Download URL | mymyr-0.2.0-cp314-cp314t-macosx_15_0_arm64.whl |
|---|---|
| Size | 10.6 MB |
| Tags | CPython 3.14 CPython 3.14 free-threading macOS 15.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
b3f8d684e5a2221ffa8ca7647c72ed1e0f390309d724d5d8e4dd6e946d78c767
|
|
BLAKE2b-256 checksum How to use checksums |
0156f344b9a2c6c7c6d8e7cf43067260ccbad24e95eedca457ea6540740d8ea3
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp314-cp314-manylinux_2_34_x86_64.whl
| Download URL | mymyr-0.2.0-cp314-cp314-manylinux_2_34_x86_64.whl |
|---|---|
| Size | 14.6 MB |
| Tags | CPython 3.14 Linux glibc 2.34+ x86-64 |
|
SHA-256 checksum How to use checksums |
aba70c20f943a68f15b011700435531d328dbf838c953766584ad287f8523722
|
|
BLAKE2b-256 checksum How to use checksums |
a9a2066039411c57a639767ba12691c3e5cf8b6f9d4304a5f20848a1af0b7f4a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logRelease files / mymyr-0.2.0-cp314-cp314-macosx_15_0_arm64.whl
| Download URL | mymyr-0.2.0-cp314-cp314-macosx_15_0_arm64.whl |
|---|---|
| Size | 10.6 MB |
| Tags | CPython 3.14 macOS 15.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
be613f09e4048804fcfe57a5f9890219457b353626bbe01dc097ea0985b65fbf
|
|
BLAKE2b-256 checksum How to use checksums |
b5735471f2630a4b43ede171e3542085c2d55b6f34a1b7616b059d0557100b05
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logSecurity for mymyr 0.2.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 9 files in this release have been uploaded with provenance metadata.
Attestation for mymyr-0.2.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 31, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.