PQID
收藏资源简介:
Parallel Quantum Instruction Dataset (PQID) 是一个包含10,718个指令-响应对的高质量语料库,旨在弥合自然语言意图与多抽象量子逻辑之间的差距。该数据集专为量子软件工程中大型语言模型(LLMs)的指令调优而构建。数据集采用双抽象架构,将用户提示与等效的量子电路生成配对,包括高级Python(Qiskit)和低级硬件表示(OPENQASM 3.0)。每个数据行代表一个交互,格式为JSON对象,包含三个字段:`instruction`(自然语言请求,如“构建一个贝尔态电路”)、`qiskit_code`(使用IBM Qiskit框架的Python代码)和`qasm_code`(等效的OPENQASM 3.0表示)。数据集可通过Hugging Face的`datasets`库直接加载。
The Parallel Quantum Instruction Dataset (PQID) is a high-quality corpus containing 10,718 instruction-response pairs, aiming to bridge the gap between natural language intent and multi-abstraction quantum logic. This dataset is specifically constructed for instruction tuning of large language models (LLMs) in quantum software engineering. The dataset adopts a dual-abstraction architecture, pairing user prompts with equivalent quantum circuit generation, including high-level Python (Qiskit) and low-level hardware representation (OPENQASM 3.0). Each data entry represents an interaction, formatted as a JSON object with three fields: `instruction` (natural language requests, e.g., "Construct a Bell state circuit"), `qiskit_code` (Python code utilizing the IBM Qiskit framework), and `qasm_code` (equivalent OPENQASM 3.0 representation). The dataset can be directly loaded via Hugging Face's `datasets` library.



