当前位置:首页 > 报告详情

2911 - 使用 Qiskit 代码助手.pdf

上传人: 竿*** 编号:982701 2025-11-29 18页 1.14MB

1、Orlando,FLOctober 69IBM TechXchange 2025Session 2911va barbosaIBM QuantumUtilizing Qiskit Code AssistantCollaboration between watsonx,IBM Research(AI for Code),and IBM Quantum2IBM TechXchange|2025 IBM CorporationBringing Generative AI to IBM QuantumBuild tools using generative AI to improve experien

2、ce and productivity of quantum practitionersDemonstrate code assistant&LLMs that writes high-quality Qiskit codeRelease Open-Source models for use in local deploymentsIBM TechXchange|2025 IBM Corporation3Qiskit ModelIBM Granite foundation model,for examplegranite-3.3-8bBaseExtended pretraining,LoRA

3、fine-tuning,&DPO post-trainingExtendModels deployed and running through watsonx.aiDeployLeverage community for continuous improvementsImproveIBM TechXchange|2025 IBM Corporation4Evaluate performance of LLMs at writing Qiskit codeDeveloped by Qiskit advocatesContains 150+unit tests to compute pass/fa

4、il huggingface.co/datasets/Qiskit/qiskit_humanevalEvaluation Benchmark:Qiskit HumanEvalIBM TechXchange|2025 IBM Corporation5Qiskit Code AssistantProvides model generated suggestions for code completionIDE extensions include VS Code and JupyterLab6IBM TechXchange|2025 IBM Corporation1.Install the VS

5、Code or JupyterLab extension2.Generate an IBM Cloud API key and configure the extension3.When prompted read and accept the model disclaimerHAPPY CODING!Getting StartedIBM TechXchange|2025 IBM Corporation7DEMOIBM TechXchange|2025 IBM Corporation8Qiskit has rapidly evolved with advancements in quantum

6、 computing and needs of the communityUsers can spend significant time keeping up with changes or incorporating latest featuresLooking AheadIBM TechXchange|2025 IBM Corporation9Given outdated,deprecated Qiskit code produce equivalent code with latest Qiskit versionIBM TechXchange|2025 IBM Corporation

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
根据报告的内容,全文主要内容概括如下: 1. **Qiskit Code Assistant**: 利用生成式AI工具,如Watsonx、IBM Research的AI for Code和IBM Quantum,开发Qiskit代码助手,以提高量子计算实践者的体验和生产力。 2. **模型和工具**: 引入Granite基础模型,通过扩展预训练、LoRA微调和DPO后训练来提升模型性能。模型通过watsonx.ai部署,并鼓励社区持续改进。 3. **性能评估**: 开发了Qiskit HumanEval,一个用于评估LLM生成Qiskit代码性能的基准。 4. **Qiskit Code Assistant功能**: 提供代码补全建议,支持VS Code和JupyterLab等IDE扩展。 5. **Qiskit Code Migration**: 通过LLM迁移旧版Qiskit代码到最新版本,简化代码更新过程。 6. **资源与信息**: 提供文档、开源模型、Qiskit HumanEval、VS Code和JupyterLab扩展等资源。 核心数据: - Qiskit HumanEval包含150+单元测试来计算通过/失败率。 - Qiskit Code Assistant提供模型生成的代码补全建议。
AI如何助力?" AI编写量子代码?" 代码迁移新篇章?"
客服
商务合作
小程序
服务号
折叠