QECTOR Decoder v3

Source-available Rust/Python quantum error correction decoding platform.

v0.6.2: Hardened Union-Find family with explicit hypergraph rejection (no more silent wrong results on unsupported codes) + comprehensive validation.

PyMatching-compatible MWPM validation • Belief-matching accuracy mode • BP-OSD for LDPC/qLDPC • CPU/GPU batch decoding • Artifact-backed evidence

Website — PyPI — Commercial licensing

---

Installation

pip install qector-decoder-v3

Supported platforms: Python 3.9–3.13 on Linux (manylinux), Windows x64, macOS arm64. Source builds available.

Optional extras:

pip install "qector-decoder-v3[stim]"   # Stim, Sinter, PyMatching, LDPC ecosystem
pip install "qector-decoder-v3[bench]"  # Benchmarking tools
pip install "qector-decoder-v3[all]"    # Full validation suite

GPU check (always verify on target machine):

from qector_decoder_v3 import CUDABatchDecoder, OpenCLBatchDecoder
print("CUDA:", CUDABatchDecoder.is_available())
print("OpenCL:", OpenCLBatchDecoder.is_available())

---

Quick Start (v0.6.2)

import numpy as np
from qector_decoder_v3 import UnionFindDecoder, BlossomDecoder, BatchDecoder, CUDABatchDecoder

# Example code (graph-like)
checks = [[0,1], [1,2], [2,3], [3,4]]
nq = 5
syndrome = np.array([0,1,0,0], dtype=np.uint8)

uf = UnionFindDecoder(checks, nq)
print(uf.decode(syndrome))

blossom = BlossomDecoder(checks, nq)
print(blossom.decode(syndrome))

# Batch decoding
batch = BatchDecoder(checks, nq)
syndromes = np.random.randint(0, 2, (4096, 4), dtype=np.uint8)
corrections = batch.parallel_batch_decode(syndromes)

if CUDABatchDecoder.is_available():
    gpu = CUDABatchDecoder(checks, nq)
    corrections = gpu.batch_decode(syndromes)

Note: UnionFindDecoder / FastUnionFindDecoder explicitly reject hypergraph codes (e.g. surface codes with weight-4 checks) since v0.6.2. Use BlossomDecoder, SparseBlossomDecoder or BpOsdDecoder for general cases.

See examples/ folder for more.

---

Core Decoders

Decoder	Best For	Status
UnionFindDecoder / FastUnionFindDecoder	Low-latency approximate (graph-like codes)	Stable (hypergraph rejection)
BlossomDecoder	Exact MWPM, PyMatching validation	Stable
SparseBlossomDecoder	Near-optimal matching	Experimental
BeliefMatching	Correlated noise accuracy	Research mode
BpOsdDecoder	LDPC / qLDPC	Experimental
BatchDecoder / CPUBatchDecoder	High-throughput CPU	Stable
CUDABatchDecoder	GPU acceleration	Runtime dependent

---

v0.6.2 Highlights & Evidence

• Critical fix: Union-Find family now hard-errors on unsupported hypergraphs.
• 919 tests (832 Python + 87 Rust), syndrome faithfulness verified.
• Competitive LER parity with PyMatching on tested surface codes.
• Honest latency positioning (Blossom slower than PyMatching on standard workloads).
• Full reproduction scripts and benchmark artifacts included.

Independent evaluation: 9.7/10 (core correctness 10/10).

Full audit, methodology, and artifacts in docs/ and benchmark_results/.

---

Reproduce Benchmarks

# Competitive MWPM
python scripts/competitive_stim_ler.py --distances 3 5 7 9 11 13 15 --shots 40000

# Belief-matching
python scripts/competitive_belief_matching.py --distances 3 5 7 --shots 3000

# Full validation bundle
python scripts/run_due_diligence_bundle.py

Always regenerate artifacts on your hardware before quoting performance numbers.

---

Commercial & Licensing

Source-available.
Personal, academic, and non-commercial research use is free.
Commercial use, redistribution, SaaS, or integration requires a paid license.

Contact: admin@qector.store

---

Citation

@software{lessard2026qector,
  author  = {Guillaume Lessard},
  title   = {{QECTOR Decoder v3}: Rust/Python Quantum Error Correction Decoding Platform},
  year    = {2026},
  version = {0.6.2},
  url     = {https://github.com/GuillaumeLessard/qector-decoder}
}

---

Full documentation, examples, and security notes are in the repository.

Copyright © 2026 Guillaume Lessard / iD01t Productions. All rights reserved.