{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreigdjj4j3jr2zml5zop3gwldiwktpzsmjuyeazcrmsn35kxhoyee2i",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3miubm3b7ih52"
  },
  "path": "/t/qsbench-synthetic-quantum-circuit-datasets-for-qml-benchmarking/175026#post_1",
  "publishedAt": "2026-04-06T15:48:24.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "Datasets collection (HF)",
    "Generator (GitHub)"
  ],
  "textContent": "# QSBench: Synthetic Quantum Circuit Datasets for QML Benchmarking\n\nHi everyone,\n\nI’m sharing **QSBench** — a collection of synthetic quantum circuit datasets designed for machine learning benchmarking, especially for graph-based models and noise-aware learning.\n\n## Resources\n\n  * Datasets collection (HF)\n  * Generator (GitHub)\n\n\n\n* * *\n\n## What is QSBench?\n\nQSBench is an ecosystem of datasets and tools for generating quantum circuits enriched with structural and physical metadata.\n\nThe goal is to move beyond:\n\n  * purely random circuits\n  * classical datasets embedded into quantum states\n\n\n\nand instead provide **structured, ML-ready quantum data**.\n\n* * *\n\n## Key Features\n\n### Structural Metadata (Graph-Ready)\n\nEach circuit includes:\n\n  * Adjacency matrices\n  * Gate-level statistics\n  * Entanglement metrics\n\n\n\nThis makes the datasets directly usable with **Graph Neural Networks (GNNs)**.\n\n* * *\n\n### Noise-Aware Design\n\nQSBench explicitly models different physical noise channels:\n\n  * Depolarizing noise\n  * Amplitude damping\n  * Thermal relaxation (T1/T2)\n  * Readout errors\n\n\n\n* * *\n\n### High-Performance Format\n\nAll datasets are stored in **Apache Parquet** , enabling:\n\n  * Faster queries\n  * Efficient large-scale processing\n  * Better integration with ML pipelines\n\n\n\n* * *\n\n## Available Datasets\n\n### QSBench-Core\n\n  * Clean structural dataset (no noise)\n  * Includes QASM, adjacency matrices, and entanglement metrics\n\n\n\n* * *\n\n### QSBench-Depolarizing\n\n  * Circuits with depolarizing noise\n  * Designed for robustness and error mitigation research\n\n\n\n* * *\n\n### QSBench-Amplitude\n\n  * Focused on amplitude damping noise\n  * Suitable for asymmetric noise modeling\n\n\n\n* * *\n\n### QSBench-Transpilation\n\n  * Raw vs transpiled circuits\n  * Useful for studying compilation overhead and optimization\n\n\n\n* * *\n\n### QSBench-Thermal\n\n  * Thermal relaxation noise (T1/T2)\n  * Designed for decoherence-aware modeling\n\n\n\n* * *\n\n### QSBench-Device\n\n  * Hardware-inspired noise models\n  * Includes realistic combinations of error sources\n\n\n\n* * *\n\n### Example Usage\n\n\n    from datasets import load_dataset\n\n    dataset = load_dataset(\"QSBench/QSBench-Core-v1.0.0-demo\")\n\n    sample = dataset[\"train\"][0]\n\n    print(sample[\"gate_count\"])\n    print(len(sample[\"adjacency_matrix\"]))\n\n\n* * *\n\n### Use Cases\n\n  * Predicting circuit properties from structure\n  * Training GNNs on quantum circuits\n  * Noise classification and error mitigation\n  * Transpilation cost estimation\n  * Hardware-aware ML modeling\n\n\n\n* * *\n\n### Roadmap\n\n  * Targeted entanglement generation\n  * Dynamic circuits (mid-circuit measurements)\n  * Integration with physical Hamiltonians\n\n\n\n* * *\n\n### Feedback\n\nWould love feedback, especially on:\n\n  * Missing features or metadata\n  * Additional noise models\n  * Real-world use cases\n\n\n\nThanks!",
  "title": "QSBench: Synthetic quantum circuit datasets for QML benchmarking"
}