"device = torch.device('cuda:2')with torch.cuda.device(device): rex_uninlu_base = pipeline( task=config['task'], model=config['model'], model_revision=config['model_revision']...

问答 2024-03-26 来自：开发者社区

ModelScope模型如何指定gpu运行呢？

"damo/nlp_deberta_rex-uninlu_chinese-base 这个模型在使用pipeline时，device='gpu:2'指定2卡，但是运行时，还是默认使用0卡，ModelScope模型如何指定gpu运行呢？"

问答 2024-03-20 来自：开发者社区

千问1.5 72B AWQ的量化模型无法使用多个GPU

请问千问1.5 72B AWQ的量化模型使用官方提供的代码运时，多个GPU无法并行使用，只是一个一个轮着运行。是什么问题造成的啊，代码和官方提供的代码一模一样。GPU为8个A30

共有144条

< 2 3 4 5 ... 15 >

跳转至： GO

更新时间 2025-03-11 12:38:50

本页面内关键词为智能算法引擎基于机器学习所生成，如有任何问题，可在页面下方点击"联系我们"与我们沟通。

GPU云服务器更多模型相关

GPU云服务器您可能感兴趣

产品推荐

{"optioninfo":{"dynamic":"true","static":"true"},"simplifiedDisplay":"newEdition","newCard":[{"ifIcon":"img","icon":"GPU","iconImg":"https://img.alicdn.com/tfs/TB1WshxnCR26e4jSZFEXXbwuXXa-232-232.png","contentLink":"https://www.aliyun.com/product/ecs/gpu","title":"GPU云服务器","des":"GPU 云服务器（GPU Cloud Computing，GPU）是提供 GPU 算力的弹性计算服务，具有超强的计算能力，服务于深度学习、科学计算、图形可视化、视频处理多种应用场景。阿里云作为亚洲第一的云服务提供商，助您提高企业竞争力。","btn1":"立即开通","link1":"https://ecs-buy.aliyun.com/wizard?spm=5176.8300896.676846.price13.19f76539dPwkz7#/prepay/cn-shenzhen?instanceType=ecs.gn6i-c8g1.2xlarge","btn2":"产品详情页","link2":"https://www.aliyun.com/product/ecs/gpu","btn3":"产品文档","link3":"https://help.aliyun.com/product/155040.html","infoGroup":[{"infoName":"最新活动","infoContent":{"firstContentName":"GPU 100小时低至0.9折","firstContentLink":"https://www.aliyun.com/activity/purchase/ecs#J_3","lastContentName":"","lastContentLink":""}},{"infoName":"活动通知","infoContent":{"firstContentName":"A10目录价最高降25%","firstContentLink":"https://help.aliyun.com/noticelist/articleid/1062941342.html?spm=5176.8300896.J_6302206100.4.66b83680szLfos","lastContentName":"","lastContentLink":""}}]}],"card":[{"link":"#","icon":"https://img.alicdn.com/tfs/TB1afahQpXXXXaAapXXXXXXXXXX-114-114.png","title":"云服务器ECS","des":"云服务器（Elastic Compute Service，简称 ECS）是一种简单高效、处理能力可弹性伸缩的计算服务，帮助您快速构建更稳定、安全的应用，提升运维效率，降低 IT 成本，使您更专注于核心业务。","btn1":"FPGA云服务器","link1":"#","btn2":"GPU云服务器","link2":"#","tip":"新一代云服务器N5 （Xeon® (Skylake)处理器）上线！<a href=\"#\" target=\"_blank\">立即申请</a>"}],"search":[{"txt":"GPU实例配置","link":"https://help.aliyun.com/document_detail/25378.html?spm=a2c4g.11186623.6.542.SO3XHE#concept_sx4_lxv_tdb__gn5"},{"txt":"GPU实例价格","link":"https://www.aliyun.com/price/product?spm=5176.8789780.1092586.1.3dfd57a8Amhd89#/ecs/detail"},{"txt":"GPU最佳实践","link":"https://help.aliyun.com/document_detail/60149.html?spm=a2c4g.11186623.6.660.R20A9X"},{"txt":"产品动态","link":" https://www.aliyun.com/product/new?source_type=out_sousuo_feature_0716"}],"infoCard":[{"fontSwitch":"shallow","bannerTitle":"","bannerUrl":"https://img.alicdn.com/tfs/TB1jLHUG1H2gK0jSZJnXXaT1FXa-1740-328.png","bannerContent":"","liveButtonName":"","liveButtonLink":"","homePageName":"","homePageLink":"","contentTitle":"","linkGroup":[{"linkContent":"","linkTo":"","link":""}]}],"title":{"mainTitle":"GPU云服务器","subtitle":"GPU云服务器是基于GPU应用的计算服务，多适用于视频解码，图形渲染，深度学习，科学计算等应用场景，该产品具有实时高速，并行计算跟浮点计算能力强等特点。","linkUrl":"#1","btnText":"查看详情"},"visual":{"topbg":"https://img.alicdn.com/tfs/TB14iILhgDqK1RjSZSyXXaxEVXa-286-101.jpg","icon":"aliyun-icon-play","textColor":"light"},"dataList":[{"summary":"云原生助力企业全面拥抱云计算","author":"叔同（阿里巴巴研究员、阿里云原生应用平台负责人）","linksUrl":""}],"sceneCard":[{"title":"云计算试飞员老蒋开课啦！第一期：电商网站搭建","des":"这是一段副标介绍，这是一段副标介绍。这是一段副标介绍，这是一段副标介绍。这是一段副标介绍，这是一段副标介绍。这是一段副标介绍，这是一段副标介绍。","btn":"立即观看","link":"#","image":"https://"}],"txt":[{"title":"适用场景","scenariosone":[{"txtone":"小型电商：支持访问50万PV,一段场景方案的介绍。","txttwo":"产品推荐","product":[{"name":"ECS 2核4G 40G数据盘 5M带宽","link":"#"}]}]}]}

GPU云服务器

GPU 云服务器（GPU Cloud Computing，GPU）是提供 GPU 算力的弹性计算服务，具有超强的计算能力，服务于深度学习、科学计算、图形可视化、视频处理多种应用场景。阿里云作为亚洲第一的云服务提供商，助您提高企业竞争力。

立即开通

产品详情页

产品文档

即刻拥有 QwQ-32B，性能比肩全球最强开源推理模型

QwQ-32B 模型具有强大的推理能力，其参数量约为 DeepSeek-R1 满血版的 1/21 且推理成本是后者的1/10，在数学、代码等核心指标（AIME 24/25、LiveCodeBench）达到 DeepSeek-R1 满血版水平。本方案介绍 QwQ-32B 的多种部署方式，用户可灵活选择，即开即用。

了解更多