mit-han-lab/torchquantum

Health v2 · maintenance only

Adoption not part of the score

1669 stars · 265 forks observed · 2026-09-22

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

TorchQuantum is a PyTorch-based framework for quantum computing simulation, supporting statevector and pulse-level simulation on GPUs with automatic gradient computation for parameterized quantum circuits. It scales to 30+ qubits with multi-GPU support and integrates with the PyTorch and Qiskit ecosystems for deployment on real quantum computers.

Use cases

• train quantum neural networks with PyTorch autograd
• simulate parameterized quantum circuits on GPU
• research quantum machine learning algorithms
• simulate 30+ qubit circuits with multiple GPUs
• deploy trained circuits to real quantum computers
• perform quantum optimal control with pulse-level simulation
• batch-simulate many quantum circuits in parallel

When to choose

• you want quantum circuits as differentiable PyTorch modules with dynamic computation graphs
• you need fast GPU-accelerated or multi-GPU quantum simulation
• you are training parameterized quantum circuits or doing quantum machine learning research
• you want easy debugging of intermediate quantum states

When to avoid

• you only need circuit construction and execution on IBM hardware without PyTorch training - Qiskit alone may suffice
• you need hybrid frameworks with extensive quantum algorithm libraries like PennyLane's ecosystem
• you need simulation beyond ~30 qubits
• you are not working in Python/PyTorch

Facets

library · maturity active

machine-learning simulation deep-learning sdk quantum-computing machine-learning deep-learning python cross-platform quantum-machine-learning quantum-neural-networks parameterized-quantum-circuits pytorch quantum-simulation qiskit-ecosystem statevector-simulation pulse-simulation algorithms gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mit-han-lab/torchquantum")