torchquantum

Why we included this project

TorchQuantum targets researchers who want to train quantum machine learning models without leaving the PyTorch workflow. Parameterized quantum circuits are built as ordinary PyTorch modules, so gradient computation and batching over the training set work natively rather than forcing you into a separate quantum SDK. The classical simulation side supports statevector and pulse simulation on GPUs, scaling to 30+ qubits across multiple GPUs, which gives real headroom for experimentation on a single machine or small cluster. When you're ready to move past the simulator, the same circuit objects export to IBMQ and other real backends, including pulse-level control, which smooths the jump from research prototype to physical hardware. The primary audience is researchers working on quantum neural networks, variational algorithms like VQE and QAOA, and quantum optimal control, and the bundled examples cover those use cases in runnable form.