cuik-molmaker

cuik-molmaker is a specialized package designed for molecular featurization, converting chemical structures into formats that can be effectively used as inputs for deep learning models, particularly graph neural networks (GNNs).

cuik-molmaker is built as a hybrid package, leveraging both C++ and Python to deliver high performance and ease of use. The core featurization logic is implemented in C++ for maximum speed and efficiency, while the Python interface provides a user-friendly API that integrates seamlessly with modern GNN training and inference workflows. This design combines the computational power of C++ with the flexibility and accessibility of Python, making cuik-molmaker both fast and intuitive for researchers and developers. As cuik-molmaker interfaces with the C++ API of rdkit, the produced features are expected to be identical to those produced by rdkit.

Quick start

Conda installation

conda install -c conda-forge cuik_molmaker

Pip installation

python -m pip install cuik-molmaker-pin==<RDKit version>

Pip installation from NVIDIA PyPI

python -m pip install cuik-molmaker --extra-index-url https://pypi.nvidia.com/rdkit-2026.03.4/

For more information on the available versions of cuik-molmaker compatible with different versions of rdkit, refer to the COMPATIBILITY.md.

Usage: Computing atom and bond features

import cuik_molmaker
import numpy as np

# List all available atom onehot features
print(cuik_molmaker.list_all_atom_onehot_features())

# Compute atom (atomic number, number of hydrogen, chirality) and bond (bond type) features for acetic acid
acetic_acid_smiles = "CC(=O)O"

# Get atom onehot feature names as NumPy array
atom_onehot_feature_array = cuik_molmaker.atom_onehot_feature_names_to_array(['atomic-number', 'num-hydrogens', 'chirality'])

# Get bond feature names as NumPy array
bond_feature_array = cuik_molmaker.bond_feature_names_to_array(['bond-type-onehot'])

# Set parameters for featurization
explicit_h, offset_carbon, duplicate_edges, add_self_loop = False, False, True, False

# Featurize
all_features =cuik_molmaker.mol_featurizer(acetic_acid_smiles, atom_onehot_feature_array, np.array([]), bond_feature_array, explicit_h, offset_carbon, duplicate_edges, add_self_loop)

# This returns a list of NumPy arrays.
# First index contains atom features
print(all_features[0].shape)

# Second index contains bond features
print(all_features[1].shape)

# Third index contains edge indices in COO format
print(all_features[2].shape)

Usage: Computing molecular descriptors

from cuik_molmaker.mol_features import MoleculeFeaturizer

featurizer = MoleculeFeaturizer(molecular_descriptor_type="rdkit2D", rdkit2D_normalization_type="fast")

smiles_list = ["CC(=O)OC1=CC=CC=C1C(=O)O", # aspirin
               "CN(C)CCOC(C1=CC=CC=C1)C1=CC=CC=C1", # diphenhydramine
]
rdkit2D_descriptors = featurizer.featurize(smiles_list)

# Print the shape of the descriptors
print(rdkit2D_descriptors.shape)

Source of acceleration

The hybrid C++/Python design of cuik-molmaker allows for the core featurization logic to be implemented in C++ and reduces the python overhead. Another source of acceleration is the creation of features for the entire minibatch of SMILES at once, which saves the overhead of creating memory allocation and concatenation.

Additional Documentation

File	Description
USAGE.md	Examples and instructions for using `cuik-molmaker` to featurize molecules including batching.
FEATURES.md	Detailed list and explanation of all atom and bond features available for featurization.
BUILD.md	Step-by-step instructions for building `cuik-molmaker` from source, including prerequisites and troubleshooting.
TESTING.md	Guidelines and commands for running the test suite to verify installation and functionality.
COMPATIBILITY.md	Published package builds by release channel, OS, and Python version.

Hardware Requirements

cuik-molmaker is designed to run on any CPU-based system.

Adoption

cuik-molmaker has currently been integrated into the following projects:

Chemprop: cuik-molmaker is available as a featurization method in Chemprop. It can be enabled by setting --use-cuikmolmaker-featurization flag in the command line with all use cases: training, prediction, fingerprinting, and hyperparameter optimization.
KERMT: cuik-molmaker is available for featurizing atoms, bonds, and molecules in KERMT. Atom and bond featurization can be enabled by setting --use_cuikmolmaker_featurization flag and molecule featurization can be enabled by setting --features_generator=rdkit_2d_normalized_cuik_molmaker.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

cuik_molmaker_pin-2026.3.5-py314-none-win_amd64.whl (259.0 kB view details)

Uploaded Aug 12, 2026 Python 3.14Windows x86-64

cuik_molmaker_pin-2026.3.5-py314-none-manylinux_2_24_x86_64.whl (330.0 kB view details)

Uploaded Aug 12, 2026 Python 3.14manylinux: glibc 2.24+ x86-64

cuik_molmaker_pin-2026.3.5-py314-none-manylinux_2_24_aarch64.whl (319.1 kB view details)

Uploaded Aug 12, 2026 Python 3.14manylinux: glibc 2.24+ ARM64

cuik_molmaker_pin-2026.3.5-py314-none-macosx_11_0_arm64.whl (286.5 kB view details)

Uploaded Aug 12, 2026 Python 3.14macOS 11.0+ ARM64

cuik_molmaker_pin-2026.3.5-py313-none-win_amd64.whl (258.9 kB view details)

Uploaded Aug 12, 2026 Python 3.13Windows x86-64

cuik_molmaker_pin-2026.3.5-py313-none-manylinux_2_24_x86_64.whl (329.9 kB view details)

Uploaded Aug 12, 2026 Python 3.13manylinux: glibc 2.24+ x86-64

cuik_molmaker_pin-2026.3.5-py313-none-manylinux_2_24_aarch64.whl (318.9 kB view details)

Uploaded Aug 12, 2026 Python 3.13manylinux: glibc 2.24+ ARM64

cuik_molmaker_pin-2026.3.5-py313-none-macosx_11_0_arm64.whl (286.5 kB view details)

Uploaded Aug 12, 2026 Python 3.13macOS 11.0+ ARM64

cuik_molmaker_pin-2026.3.5-py312-none-win_amd64.whl (258.8 kB view details)

Uploaded Aug 12, 2026 Python 3.12Windows x86-64

cuik_molmaker_pin-2026.3.5-py312-none-manylinux_2_24_x86_64.whl (328.0 kB view details)

Uploaded Aug 12, 2026 Python 3.12manylinux: glibc 2.24+ x86-64

cuik_molmaker_pin-2026.3.5-py312-none-manylinux_2_24_aarch64.whl (318.2 kB view details)

Uploaded Aug 12, 2026 Python 3.12manylinux: glibc 2.24+ ARM64

cuik_molmaker_pin-2026.3.5-py312-none-macosx_11_0_arm64.whl (286.1 kB view details)

Uploaded Aug 12, 2026 Python 3.12macOS 11.0+ ARM64

cuik_molmaker_pin-2026.3.5-py311-none-win_amd64.whl (257.8 kB view details)

Uploaded Aug 12, 2026 Python 3.11Windows x86-64

cuik_molmaker_pin-2026.3.5-py311-none-manylinux_2_24_x86_64.whl (329.0 kB view details)

Uploaded Aug 12, 2026 Python 3.11manylinux: glibc 2.24+ x86-64

cuik_molmaker_pin-2026.3.5-py311-none-manylinux_2_24_aarch64.whl (318.9 kB view details)

Uploaded Aug 12, 2026 Python 3.11manylinux: glibc 2.24+ ARM64

cuik_molmaker_pin-2026.3.5-py311-none-macosx_11_0_arm64.whl (286.8 kB view details)

Uploaded Aug 12, 2026 Python 3.11macOS 11.0+ ARM64

File details

Details for the file cuik_molmaker_pin-2026.3.5-py314-none-win_amd64.whl.

File metadata

Download URL: cuik_molmaker_pin-2026.3.5-py314-none-win_amd64.whl
Upload date: Aug 12, 2026
Size: 259.0 kB
Tags: Python 3.14, Windows x86-64
Uploaded using Trusted Publishing? No
Uploaded via: twine/6.2.0 CPython/3.10.13

File hashes

Hashes for cuik_molmaker_pin-2026.3.5-py314-none-win_amd64.whl
Algorithm	Hash digest
SHA256	`0993605b6ee4a5a6fdb16f0b182b10f5ea189bb9dc2bc143ab130fe2f20fbb72`
MD5	`62736177b5e125cc905e90ad2096e2fd`
BLAKE2b-256	`0893876e80d96ede08fc7fb935dcafc93bb0853553c9e03c684b5e8122465976`

Algorithm	Hash digest
SHA256	`80def6bb360b398d066b76fe776d18c3fd41d8df83b1f5174685e7b3813c28c2`
MD5	`0b596e27f9e4b58ecefda6e7f90e9905`
BLAKE2b-256	`1ff0a9decf6de733b1925c0a58fa10072a71669eee7392d8208c5db36d15e8d6`

Algorithm	Hash digest
SHA256	`d2a3fb6b92dcad97de0e84da2d031b1d6d14c5bbd947e4ff73fc73a534e80c17`
MD5	`590c518b093fbc5dc113aff7dd3ee831`
BLAKE2b-256	`55d67e54551fed347f5f30ef517c6b8239c44899889b37334e86eec3694b3c56`

Algorithm	Hash digest
SHA256	`f59d6b09bace737d61e8fb1d2bb4faa6caeaaa46ab7a436f0204bd90a9db4b71`
MD5	`2ffe86aa99cabdcc671cfc6bb3ebb7d6`
BLAKE2b-256	`4bb40c39aa6fde73031042834215110f5a4e49dbea6ca47ffc1fcb59ac5d346a`

Algorithm	Hash digest
SHA256	`f959c879c42c376a1bb763dd57ed21790ecaddddde36e47df819438c8866d935`
MD5	`2db8addf11d771726a542a6660c8368b`
BLAKE2b-256	`5e1eb40f5b54cae008fbd8aaac4f7c5e545e012042df383f7535245d1f1c7cb0`

Algorithm	Hash digest
SHA256	`85ef0e51e1ac0839a0c14edde8d267edce5729e6aece3886caf0588442ecbe55`
MD5	`afd2af0082db3100d31d182b05bcd123`
BLAKE2b-256	`70501ac70af80dc5dc2b4b9b517153419c1abbe4123c759afb10e81f86b27a73`

Algorithm	Hash digest
SHA256	`a9c32ddf7edc8312668363eff1a1000b5f55dddcac6b75838a446214ed96adcc`
MD5	`69ca996ffd46a4d1ed67fa94268871ea`
BLAKE2b-256	`66a9939e9b42dea6e6715c5da37365bb6f92592a5c64eac6a22f01d8cf66126f`

Algorithm	Hash digest
SHA256	`63e91e66dcc9f91b1ea6fb7d82d110490ee223978eb2efc91aadf8bb26912228`
MD5	`4b8c5943d6a2da3eea1fe74d44a1b87e`
BLAKE2b-256	`b260c6edb070fe3c758f7472d3e4d4005b52b75ca448b08f08080556e9deba72`

Algorithm	Hash digest
SHA256	`a440b871daa9b239b2082dfb6cf15f096893d0b0eb8b414a19f16c4027d45d3c`
MD5	`d9d8dd163d499211bd0c6799296c7f00`
BLAKE2b-256	`855697de8c2583e3f21b0c38ce4b4512601bce4987e0d07212fda4c26deb4c48`

Algorithm	Hash digest
SHA256	`12b70963767ee2612ef802aeef98df7180a556d855063bf0a909bcfef1fc00a5`
MD5	`44bd64b2f86f2de37f050c7c7e13cd3b`
BLAKE2b-256	`d0b4b09b3ec5a741ef40eac2dc3f2be7dfce2ce8f8c4e9d89a15bc45280907b2`

Algorithm	Hash digest
SHA256	`266cc95d343902ca8693bc0f7e615caed2198d0e76f47fdf3ac8955157adfbab`
MD5	`cebf6794abde56d6f848d57b3ab43465`
BLAKE2b-256	`c6e0b7e7feb284b790271bc58a8bee842a75cdc6f8f1b28b39da775507ce7cb9`

Algorithm	Hash digest
SHA256	`302900629c44b4ded59d31b8086c1607c2c51b9edf8f4988cc55a26526e53f6a`
MD5	`46d8060658a76c63cd8fcf26837a5a78`
BLAKE2b-256	`affe7d09c1efba5412fa2fa621a8a08e0b21dec2b68751b23c4d43850e2b8d00`

Algorithm	Hash digest
SHA256	`5b6a458591a26050f5676cc0d92c360b36327bd0036e52816798ed7bb7badd9f`
MD5	`5a259ae875f088daa0845e155a45e8b4`
BLAKE2b-256	`6e2f886f95dce6f1de1a49a7952185e61c4417581cb2d6045656b1d435eaba76`

Algorithm	Hash digest
SHA256	`b8ca73991076b77685c283230e292330929a38d61c5d9a8642212cb26550ea4c`
MD5	`0ba33d688796190ff5c5b637d85169fb`
BLAKE2b-256	`fa53be40e4643088425fa1538872427d20b2665e9037eb557be2b70dec325eb3`

Algorithm	Hash digest
SHA256	`82df344f16e5431661bcb2fa664296a80fbe0b7945dfa8c68a7152f9e2e6fe98`
MD5	`f1806d2f685643ffafafccc5d72dcffa`
BLAKE2b-256	`a35ac642a65b7b7b5ba9e03bfa7aa6dc2c17595bb5f8517c2618da802bd9bb19`

Algorithm	Hash digest
SHA256	`7e11b5c06e0a22bb85b0376e5b49bd1f3046b2d53c2feaedf81033559a975c5b`
MD5	`ea773b2f861cd8acba4605ec7f2ef771`
BLAKE2b-256	`abacfc82233be6918630bb4a60a15f27fafd44ce29648f55f879a0a72f3bc8aa`

cuik-molmaker-pin 2026.3.5

cuik-molmaker

Quick start

Conda installation

Pip installation

Pip installation from NVIDIA PyPI

Usage: Computing atom and bond features

Usage: Computing molecular descriptors

Source of acceleration

Additional Documentation

Hardware Requirements

Adoption

Key dates

1 maintainer

Credits

License

Requires

Provides Extra

Classifiers

Download files

Source Distributions

Built Distributions

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

Release history Release notifications | RSS feed