# M[](https://www.python.org/downloads/)lecule Benchmarks
[](https://badge.fury.io/py/molecule-benchmarks)
[](https://www.python.org/downloads/)
[](https://opensource.org/licenses/MIT)
A comprehensive benchmark suite for evaluating generative models for molecules. This package provides standardized metrics and evaluation protocols for assessing the quality of molecular generation models in drug discovery and cheminformatics.
## Features
- **Comprehensive Metrics**: Validity, uniqueness, novelty, diversity, and similarity metrics
- **Standard Benchmarks**: Implements metrics from Moses, GuacaMol, and FCD papers
- **Easy Integration**: Simple interface for integrating with any generative model
- **Direct SMILES Evaluation**: Benchmark pre-generated SMILES lists without implementing a model interface
- **Multiple Datasets**: Built-in support for QM9, Moses, and GuacaMol datasets
- **Efficient Computation**: Optimized for large-scale evaluation with multiprocessing support
## Installation
```bash
pip install molecule-benchmarks
```
## Quick Start
You can use the benchmark suite in two ways:
### Option 1: Direct SMILES Evaluation (Simplified)
If you already have generated SMILES strings, you can benchmark them directly. Just ensure you have at least the number of samples specified in `num_samples_to_generate`.
```python
from molecule_benchmarks import Benchmarker, SmilesDataset
# Load a dataset
dataset = SmilesDataset.load_qm9_dataset(subset_size=10000)
# Initialize benchmarker
benchmarker = Benchmarker(
dataset=dataset,
num_samples_to_generate=10000,# You need to generate at least this many samples
device="cpu" # or "cuda" for GPU
)
# Your generated SMILES (replace with your actual generated molecules)
generated_smiles = [
"CCO", # Ethanol
"CC(=O)O", # Acetic acid
"c1ccccc1", # Benzene
"CC(C)O", # Isopropanol
"CCN", # Ethylamine
None, # Invalid molecule (use None for failures)
# ... more molecules up to num_samples_to_generate
]
# Run benchmarks directly on the SMILES list
results = benchmarker.benchmark(generated_smiles)
print(results)
```
### Option 2: Model-Based Evaluation
To use the benchmark suite with a generative model, implement the `MoleculeGenerationModel` protocol. This will generate the required number of samples and run the benchmarks.
```python
from molecule_benchmarks.model import MoleculeGenerationModel
class MyGenerativeModel(MoleculeGenerationModel):
def __init__(self, model_path):
# Initialize your model here
self.model = load_model(model_path)
def generate_molecule_batch(self) -> list[str | None]:
"""Generate a batch of molecules as SMILES strings.
Returns:
List of SMILES strings. Return None for invalid molecules.
"""
# Your generation logic here
batch = self.model.generate(batch_size=100)
return [self.convert_to_smiles(mol) for mol in batch]
# Initialize your model
model = MyGenerativeModel("path/to/model")
# Run benchmarks using the model
results = benchmarker.benchmark_model(model)
print(results)
```
### 3. Analyze Results
The benchmark returns comprehensive metrics:
```python
# Validity metrics
print(f"Valid molecules: {results['validity']['valid_fraction']:.3f}")
print(f"Valid & unique: {results['validity']['valid_and_unique_fraction']:.3f}")
print(f"Valid & unique & novel: {results['validity']['valid_and_unique_and_novel_fraction']:.3f}")
# Diversity and similarity metrics
print(f"Internal diversity: {results['moses']['IntDiv']:.3f}")
print(f"SNN score: {results['moses']['snn_score']:.3f}")
# Chemical property distribution similarity
print(f"KL divergence score: {results['kl_score']:.3f}")
# Fréchet ChemNet Distance
print(f"FCD score: {results['fcd']['fcd']:.3f}")
```
## Complete Examples
### Example 1: Direct SMILES Benchmarking (Recommended for Simplicity)
```python
from molecule_benchmarks import Benchmarker, SmilesDataset
# Load dataset
print("Loading dataset...")
dataset = SmilesDataset.load_qm9_dataset(max_train_samples=1000)
# Create benchmarker
benchmarker = Benchmarker(
dataset=dataset,
num_samples_to_generate=100,
device="cpu"
)
# Your generated SMILES (replace with your actual generated molecules)
generated_smiles = [
"CCO", # Ethanol
"CC(=O)O", # Acetic acid
"c1ccccc1", # Benzene
"CC(C)O", # Isopropanol
"CCN", # Ethylamine
None, # Invalid molecule
# ... add more molecules up to 100 total
] + [None] * (100 - 6) # Pad with None to reach desired count
# Run benchmarks directly
print("Running benchmarks...")
results = benchmarker.benchmark(generated_smiles)
# Print results (same as below)
print("\n=== Validity Metrics ===")
print(f"Valid molecules: {results['validity']['valid_fraction']:.3f}")
print(f"Unique molecules: {results['validity']['unique_fraction']:.3f}")
print(f"Valid & unique: {results['validity']['valid_and_unique_fraction']:.3f}")
print(f"Novel molecules: {results['validity']['valid_and_unique_and_novel_fraction']:.3f}")
print("\n=== Moses Metrics ===")
print(f"Passing Moses filters: {results['moses']['fraction_passing_moses_filters']:.3f}")
print(f"SNN score: {results['moses']['snn_score']:.3f}")
print(f"Internal diversity (p=1): {results['moses']['IntDiv']:.3f}")
print(f"Internal diversity (p=2): {results['moses']['IntDiv2']:.3f}")
print("\n=== Distribution Metrics ===")
print(f"KL divergence score: {results['kl_score']:.3f}")
print(f"FCD score: {results['fcd']['fcd']:.3f}")
print(f"FCD (valid only): {results['fcd']['fcd_valid']:.3f}")
```
### Example 2: Model-Based Benchmarking
Here's a complete example using the built-in dummy model:
```python
from molecule_benchmarks import Benchmarker, SmilesDataset
from molecule_benchmarks.model import DummyMoleculeGenerationModel
# Load dataset
print("Loading dataset...")
dataset = SmilesDataset.load_qm9_dataset(max_train_samples=1000)
# Create benchmarker
benchmarker = Benchmarker(
dataset=dataset,
num_samples_to_generate=100,
device="cpu"
)
# Create a dummy model (replace with your model)
model = DummyMoleculeGenerationModel([
"CCO", # Ethanol
"CC(=O)O", # Acetic acid
"c1ccccc1", # Benzene
"CC(C)O", # Isopropanol
"CCN", # Ethylamine
None, # Invalid molecule
])
# Run benchmarks using the model
print("Running benchmarks...")
results = benchmarker.benchmark_model(model)
# Print results
print("\n=== Validity Metrics ===")
print(f"Valid molecules: {results['validity']['valid_fraction']:.3f}")
print(f"Unique molecules: {results['validity']['unique_fraction']:.3f}")
print(f"Valid & unique: {results['validity']['valid_and_unique_fraction']:.3f}")
print(f"Novel molecules: {results['validity']['valid_and_unique_and_novel_fraction']:.3f}")
print("\n=== Moses Metrics ===")
print(f"Passing Moses filters: {results['moses']['fraction_passing_moses_filters']:.3f}")
print(f"SNN score: {results['moses']['snn_score']:.3f}")
print(f"Internal diversity (p=1): {results['moses']['IntDiv']:.3f}")
print(f"Internal diversity (p=2): {results['moses']['IntDiv2']:.3f}")
print("\n=== Distribution Metrics ===")
print(f"KL divergence score: {results['kl_score']:.3f}")
print(f"FCD score: {results['fcd']['fcd']:.3f}")
print(f"FCD (valid only): {results['fcd']['fcd_valid']:.3f}")
```
## Supported Datasets
The package includes several built-in datasets:
```python
from molecule_benchmarks import SmilesDataset
# QM9 dataset (small molecules)
dataset = SmilesDataset.load_qm9_dataset(subset_size=10000)
# Moses dataset (larger, drug-like molecules)
dataset = SmilesDataset.load_moses_dataset(fraction=0.1)
# GuacaMol dataset
dataset = SmilesDataset.load_guacamol_dataset(fraction=0.1)
# Custom dataset from files
dataset = SmilesDataset(
train_smiles="path/to/train.txt",
validation_smiles="path/to/valid.txt"
)
```
## Metrics Explained
### Validity Metrics
- **Valid fraction**: Percentage of generated molecules that are chemically valid
- **Unique fraction**: Percentage of generated molecules that are unique
- **Novel fraction**: Percentage of generated molecules not seen in training data
### Moses Metrics
Based on the [Moses paper](https://arxiv.org/abs/1811.12823):
- **SNN score**: Similarity to nearest neighbor in training set
- **Internal diversity**: Average pairwise Tanimoto distance within generated set
- **Scaffold similarity**: Similarity of molecular scaffolds to training set
- **Fragment similarity**: Similarity of molecular fragments to training set
### Distribution Metrics
- **KL divergence score**: Measures similarity of molecular property distributions
- **FCD score**: Fréchet ChemNet Distance, measures distribution similarity in learned feature space
## Advanced Usage
### Direct SMILES Evaluation
For most use cases, directly evaluating a list of generated SMILES is the simplest approach:
```python
# Custom number of samples and device
benchmarker = Benchmarker(
dataset=dataset,
num_samples_to_generate=50000,
device="cuda" # Use GPU for faster computation
)
# Your generated SMILES list (with None for invalid generations)
my_generated_smiles = [
"CCO", "c1ccccc1", "CC(=O)O", None, "invalid_smiles",
# ... up to 50000 molecules
]
# Run benchmarks directly
results = benchmarker.benchmark(my_generated_smiles)
# Access specific metric computations
validity_scores = benchmarker._compute_validity_scores(my_generated_smiles)
fcd_scores = benchmarker._compute_fcd_scores(my_generated_smiles)
```
### Model-Based Evaluation
For integration with generative models:
```python
class BatchedModel(MoleculeGenerationModel):
def generate_molecule_batch(self) -> list[str | None]:
# Generate larger batches for efficiency
return self.model.sample(batch_size=1000)
# Use the model with benchmarker
results = benchmarker.benchmark_model(BatchedModel())
```
### Important Notes
- **SMILES format**: Use None for molecules that failed to generate or are invalid
- **Batch size**: The `num_samples_to_generate` parameter determines how many molecules will be evaluated
- **Validation**: Invalid SMILES are automatically detected and handled in the metrics
- **Memory**: For large evaluations (>10k molecules), consider using GPU acceleration with `device="cuda"`
## Contributing
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## Acknowledgments
This benchmark suite implements and builds upon metrics from several important papers:
- [Moses: A Benchmarking Platform for Molecular Generation Models](https://arxiv.org/abs/1811.12823)
- [GuacaMol: Benchmarking Models for De Novo Molecular Design](https://arxiv.org/abs/1811.09621)
- [Fréchet ChemNet Distance: A Metric for Generative Models for Molecules](https://arxiv.org/abs/1803.09518)
Raw data
{
"_id": null,
"home_page": null,
"name": "molecule-benchmarks",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.11",
"maintainer_email": null,
"keywords": "benchmarking, cheminformatics, drug-discovery, generative-models, molecules",
"author": null,
"author_email": "Ole Petersen <peteole2707@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/7d/7a/674a616aa33ab2718ac17b0845d1324c3168d1aad2aeb20ab7575131656c/molecule_benchmarks-0.1.12.tar.gz",
"platform": null,
"description": "# M[](https://www.python.org/downloads/)lecule Benchmarks\n\n[](https://badge.fury.io/py/molecule-benchmarks)\n[](https://www.python.org/downloads/)\n[](https://opensource.org/licenses/MIT)\n\nA comprehensive benchmark suite for evaluating generative models for molecules. This package provides standardized metrics and evaluation protocols for assessing the quality of molecular generation models in drug discovery and cheminformatics.\n\n## Features\n\n- **Comprehensive Metrics**: Validity, uniqueness, novelty, diversity, and similarity metrics\n- **Standard Benchmarks**: Implements metrics from Moses, GuacaMol, and FCD papers\n- **Easy Integration**: Simple interface for integrating with any generative model\n- **Direct SMILES Evaluation**: Benchmark pre-generated SMILES lists without implementing a model interface\n- **Multiple Datasets**: Built-in support for QM9, Moses, and GuacaMol datasets\n- **Efficient Computation**: Optimized for large-scale evaluation with multiprocessing support\n\n## Installation\n\n```bash\npip install molecule-benchmarks\n```\n\n## Quick Start\n\nYou can use the benchmark suite in two ways:\n\n### Option 1: Direct SMILES Evaluation (Simplified)\n\nIf you already have generated SMILES strings, you can benchmark them directly. Just ensure you have at least the number of samples specified in `num_samples_to_generate`.\n\n```python\nfrom molecule_benchmarks import Benchmarker, SmilesDataset\n\n# Load a dataset\ndataset = SmilesDataset.load_qm9_dataset(subset_size=10000)\n\n# Initialize benchmarker\nbenchmarker = Benchmarker(\n dataset=dataset,\n num_samples_to_generate=10000,# You need to generate at least this many samples\n device=\"cpu\" # or \"cuda\" for GPU\n)\n\n# Your generated SMILES (replace with your actual generated molecules)\ngenerated_smiles = [\n \"CCO\", # Ethanol\n \"CC(=O)O\", # Acetic acid\n \"c1ccccc1\", # Benzene\n \"CC(C)O\", # Isopropanol\n \"CCN\", # Ethylamine\n None, # Invalid molecule (use None for failures)\n # ... more molecules up to num_samples_to_generate\n]\n\n# Run benchmarks directly on the SMILES list\nresults = benchmarker.benchmark(generated_smiles)\nprint(results)\n```\n\n### Option 2: Model-Based Evaluation\n\nTo use the benchmark suite with a generative model, implement the `MoleculeGenerationModel` protocol. This will generate the required number of samples and run the benchmarks.\n\n```python\nfrom molecule_benchmarks.model import MoleculeGenerationModel\n\nclass MyGenerativeModel(MoleculeGenerationModel):\n def __init__(self, model_path):\n # Initialize your model here\n self.model = load_model(model_path)\n \n def generate_molecule_batch(self) -> list[str | None]:\n \"\"\"Generate a batch of molecules as SMILES strings.\n \n Returns:\n List of SMILES strings. Return None for invalid molecules.\n \"\"\"\n # Your generation logic here\n batch = self.model.generate(batch_size=100)\n return [self.convert_to_smiles(mol) for mol in batch]\n\n# Initialize your model\nmodel = MyGenerativeModel(\"path/to/model\")\n\n# Run benchmarks using the model\nresults = benchmarker.benchmark_model(model)\nprint(results)\n```\n\n### 3. Analyze Results\n\nThe benchmark returns comprehensive metrics:\n\n```python\n# Validity metrics\nprint(f\"Valid molecules: {results['validity']['valid_fraction']:.3f}\")\nprint(f\"Valid & unique: {results['validity']['valid_and_unique_fraction']:.3f}\")\nprint(f\"Valid & unique & novel: {results['validity']['valid_and_unique_and_novel_fraction']:.3f}\")\n\n# Diversity and similarity metrics\nprint(f\"Internal diversity: {results['moses']['IntDiv']:.3f}\")\nprint(f\"SNN score: {results['moses']['snn_score']:.3f}\")\n\n# Chemical property distribution similarity\nprint(f\"KL divergence score: {results['kl_score']:.3f}\")\n\n# Fr\u00e9chet ChemNet Distance\nprint(f\"FCD score: {results['fcd']['fcd']:.3f}\")\n```\n\n## Complete Examples\n\n### Example 1: Direct SMILES Benchmarking (Recommended for Simplicity)\n\n```python\nfrom molecule_benchmarks import Benchmarker, SmilesDataset\n\n# Load dataset\nprint(\"Loading dataset...\")\ndataset = SmilesDataset.load_qm9_dataset(max_train_samples=1000)\n\n# Create benchmarker\nbenchmarker = Benchmarker(\n dataset=dataset,\n num_samples_to_generate=100,\n device=\"cpu\"\n)\n\n# Your generated SMILES (replace with your actual generated molecules)\ngenerated_smiles = [\n \"CCO\", # Ethanol\n \"CC(=O)O\", # Acetic acid\n \"c1ccccc1\", # Benzene\n \"CC(C)O\", # Isopropanol\n \"CCN\", # Ethylamine\n None, # Invalid molecule\n # ... add more molecules up to 100 total\n] + [None] * (100 - 6) # Pad with None to reach desired count\n\n# Run benchmarks directly\nprint(\"Running benchmarks...\")\nresults = benchmarker.benchmark(generated_smiles)\n\n# Print results (same as below)\nprint(\"\\n=== Validity Metrics ===\")\nprint(f\"Valid molecules: {results['validity']['valid_fraction']:.3f}\")\nprint(f\"Unique molecules: {results['validity']['unique_fraction']:.3f}\")\nprint(f\"Valid & unique: {results['validity']['valid_and_unique_fraction']:.3f}\")\nprint(f\"Novel molecules: {results['validity']['valid_and_unique_and_novel_fraction']:.3f}\")\n\nprint(\"\\n=== Moses Metrics ===\")\nprint(f\"Passing Moses filters: {results['moses']['fraction_passing_moses_filters']:.3f}\")\nprint(f\"SNN score: {results['moses']['snn_score']:.3f}\")\nprint(f\"Internal diversity (p=1): {results['moses']['IntDiv']:.3f}\")\nprint(f\"Internal diversity (p=2): {results['moses']['IntDiv2']:.3f}\")\n\nprint(\"\\n=== Distribution Metrics ===\")\nprint(f\"KL divergence score: {results['kl_score']:.3f}\")\nprint(f\"FCD score: {results['fcd']['fcd']:.3f}\")\nprint(f\"FCD (valid only): {results['fcd']['fcd_valid']:.3f}\")\n```\n\n### Example 2: Model-Based Benchmarking\n\nHere's a complete example using the built-in dummy model:\n\n```python\nfrom molecule_benchmarks import Benchmarker, SmilesDataset\nfrom molecule_benchmarks.model import DummyMoleculeGenerationModel\n\n# Load dataset\nprint(\"Loading dataset...\")\ndataset = SmilesDataset.load_qm9_dataset(max_train_samples=1000)\n\n# Create benchmarker\nbenchmarker = Benchmarker(\n dataset=dataset,\n num_samples_to_generate=100,\n device=\"cpu\"\n)\n\n# Create a dummy model (replace with your model)\nmodel = DummyMoleculeGenerationModel([\n \"CCO\", # Ethanol\n \"CC(=O)O\", # Acetic acid\n \"c1ccccc1\", # Benzene\n \"CC(C)O\", # Isopropanol\n \"CCN\", # Ethylamine\n None, # Invalid molecule\n])\n\n# Run benchmarks using the model\nprint(\"Running benchmarks...\")\nresults = benchmarker.benchmark_model(model)\n\n# Print results\nprint(\"\\n=== Validity Metrics ===\")\nprint(f\"Valid molecules: {results['validity']['valid_fraction']:.3f}\")\nprint(f\"Unique molecules: {results['validity']['unique_fraction']:.3f}\")\nprint(f\"Valid & unique: {results['validity']['valid_and_unique_fraction']:.3f}\")\nprint(f\"Novel molecules: {results['validity']['valid_and_unique_and_novel_fraction']:.3f}\")\n\nprint(\"\\n=== Moses Metrics ===\")\nprint(f\"Passing Moses filters: {results['moses']['fraction_passing_moses_filters']:.3f}\")\nprint(f\"SNN score: {results['moses']['snn_score']:.3f}\")\nprint(f\"Internal diversity (p=1): {results['moses']['IntDiv']:.3f}\")\nprint(f\"Internal diversity (p=2): {results['moses']['IntDiv2']:.3f}\")\n\nprint(\"\\n=== Distribution Metrics ===\")\nprint(f\"KL divergence score: {results['kl_score']:.3f}\")\nprint(f\"FCD score: {results['fcd']['fcd']:.3f}\")\nprint(f\"FCD (valid only): {results['fcd']['fcd_valid']:.3f}\")\n```\n\n## Supported Datasets\n\nThe package includes several built-in datasets:\n\n```python\nfrom molecule_benchmarks import SmilesDataset\n\n# QM9 dataset (small molecules)\ndataset = SmilesDataset.load_qm9_dataset(subset_size=10000)\n\n# Moses dataset (larger, drug-like molecules)\ndataset = SmilesDataset.load_moses_dataset(fraction=0.1)\n\n# GuacaMol dataset\ndataset = SmilesDataset.load_guacamol_dataset(fraction=0.1)\n\n# Custom dataset from files\ndataset = SmilesDataset(\n train_smiles=\"path/to/train.txt\",\n validation_smiles=\"path/to/valid.txt\"\n)\n```\n\n## Metrics Explained\n\n### Validity Metrics\n\n- **Valid fraction**: Percentage of generated molecules that are chemically valid\n- **Unique fraction**: Percentage of generated molecules that are unique\n- **Novel fraction**: Percentage of generated molecules not seen in training data\n\n### Moses Metrics\n\nBased on the [Moses paper](https://arxiv.org/abs/1811.12823):\n\n- **SNN score**: Similarity to nearest neighbor in training set\n- **Internal diversity**: Average pairwise Tanimoto distance within generated set\n- **Scaffold similarity**: Similarity of molecular scaffolds to training set\n- **Fragment similarity**: Similarity of molecular fragments to training set\n\n### Distribution Metrics\n\n- **KL divergence score**: Measures similarity of molecular property distributions\n- **FCD score**: Fr\u00e9chet ChemNet Distance, measures distribution similarity in learned feature space\n\n## Advanced Usage\n\n### Direct SMILES Evaluation\n\nFor most use cases, directly evaluating a list of generated SMILES is the simplest approach:\n\n```python\n# Custom number of samples and device\nbenchmarker = Benchmarker(\n dataset=dataset,\n num_samples_to_generate=50000,\n device=\"cuda\" # Use GPU for faster computation\n)\n\n# Your generated SMILES list (with None for invalid generations)\nmy_generated_smiles = [\n \"CCO\", \"c1ccccc1\", \"CC(=O)O\", None, \"invalid_smiles\", \n # ... up to 50000 molecules\n]\n\n# Run benchmarks directly\nresults = benchmarker.benchmark(my_generated_smiles)\n\n# Access specific metric computations\nvalidity_scores = benchmarker._compute_validity_scores(my_generated_smiles)\nfcd_scores = benchmarker._compute_fcd_scores(my_generated_smiles)\n```\n\n### Model-Based Evaluation\n\nFor integration with generative models:\n\n```python\nclass BatchedModel(MoleculeGenerationModel):\n def generate_molecule_batch(self) -> list[str | None]:\n # Generate larger batches for efficiency\n return self.model.sample(batch_size=1000)\n\n# Use the model with benchmarker\nresults = benchmarker.benchmark_model(BatchedModel())\n```\n\n### Important Notes\n\n- **SMILES format**: Use None for molecules that failed to generate or are invalid\n- **Batch size**: The `num_samples_to_generate` parameter determines how many molecules will be evaluated\n- **Validation**: Invalid SMILES are automatically detected and handled in the metrics\n- **Memory**: For large evaluations (>10k molecules), consider using GPU acceleration with `device=\"cuda\"`\n\n## Contributing\n\nContributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.\n\n## License\n\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n## Acknowledgments\n\nThis benchmark suite implements and builds upon metrics from several important papers:\n\n- [Moses: A Benchmarking Platform for Molecular Generation Models](https://arxiv.org/abs/1811.12823)\n- [GuacaMol: Benchmarking Models for De Novo Molecular Design](https://arxiv.org/abs/1811.09621)\n- [Fr\u00e9chet ChemNet Distance: A Metric for Generative Models for Molecules](https://arxiv.org/abs/1803.09518)\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "A comprehensive benchmark suite for evaluating generative models for molecules",
"version": "0.1.12",
"project_urls": {
"Documentation": "https://molecule-benchmarks.readthedocs.io/",
"Issues": "https://github.com/peteole/molecule_benchmarks/issues",
"Repository": "https://github.com/peteole/molecule_benchmarks"
},
"split_keywords": [
"benchmarking",
" cheminformatics",
" drug-discovery",
" generative-models",
" molecules"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "12861bd71d71dd3fe99e579be98f4c2494db287fbda1e17b625fd487a1889057",
"md5": "67312a881d7509f69aa6e8f60978ba8f",
"sha256": "d9a20e565f604b2441cf42f644c5a10744e065fe4960060ef6dbb259881cbd11"
},
"downloads": -1,
"filename": "molecule_benchmarks-0.1.12-py3-none-any.whl",
"has_sig": false,
"md5_digest": "67312a881d7509f69aa6e8f60978ba8f",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.11",
"size": 43603,
"upload_time": "2025-07-08T15:49:36",
"upload_time_iso_8601": "2025-07-08T15:49:36.859672Z",
"url": "https://files.pythonhosted.org/packages/12/86/1bd71d71dd3fe99e579be98f4c2494db287fbda1e17b625fd487a1889057/molecule_benchmarks-0.1.12-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "7d7a674a616aa33ab2718ac17b0845d1324c3168d1aad2aeb20ab7575131656c",
"md5": "f5391590bbf4aaac79be046efb516db2",
"sha256": "009bee0a77b6f3741fa2f840c07208117ddfca478eecd0be77a6b50c66d55dbf"
},
"downloads": -1,
"filename": "molecule_benchmarks-0.1.12.tar.gz",
"has_sig": false,
"md5_digest": "f5391590bbf4aaac79be046efb516db2",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.11",
"size": 129389,
"upload_time": "2025-07-08T15:49:38",
"upload_time_iso_8601": "2025-07-08T15:49:38.449401Z",
"url": "https://files.pythonhosted.org/packages/7d/7a/674a616aa33ab2718ac17b0845d1324c3168d1aad2aeb20ab7575131656c/molecule_benchmarks-0.1.12.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-07-08 15:49:38",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "peteole",
"github_project": "molecule_benchmarks",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "molecule-benchmarks"
}
JSON
