modelscope训练设置的相关内容

问答 2023-04-03 来自：开发者社区

想问下ModelScope训练的话这个报错要怎么设置参数呢？在PAI平台

GPT3Trainer: GPT3ForTextGeneration: CUDA out of memory. Tried to allocate 100.00 MiB (GPU 0; 22.20 GiB total capacity; 5.90 GiB already allocated; 70.12 MiB free; 5.90 GiB reserved in total by PyTorc....

问答 2023-03-01 来自：开发者社区

modelscope 训练时，cfg.train.checkpoint的保存中间结果设置不管用？

modelscope跑的模型, 请问这个saving checkpoint at 200 iterations，怎么把200改的大一点，在哪里设置，我按照你们给出的格式cfg.train.checkpoint，把"by_epoch": True，但是结果还是按照iter进行保存的，interval不管怎么设置，跑的结果还是200个interval就保存一次，这怎么弄。求答复，谢谢。

共有12条

< 1 2 >

跳转至： GO

更新时间 2024-09-12 17:00:13

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。

modelscope训练相关内容

modelscope更多训练相关

modelscope您可能感兴趣

产品推荐

{"optioninfo":{"dynamic":"ture","static":"true"},"simplifiedDisplay":"newEdition","newCard":[{"ifIcon":"img","link":"https://img.alicdn.com/tfs/TB1XY8hGYr1gK0jSZFDXXb9yVXa-1740-328.png","icon":"ModelScope 魔搭社区","iconImg":"https://img.alicdn.com/imgextra/i3/O1CN01WCQZsb1ICYCctdAi1_!!6000000000857-55-tps-240-240.svg","contentLink":"https://modelscope.cn","title":"ModelScope 魔搭社区","des":"为实现“模型即服务”理念，魔搭社区提供最新最热、开放开源的多领域预训练模型和优质数据集，让广大开发者无需写代码就能快速体验模型效果；同时提供抽象编程接口及SDK，对模型进行二次开发，真正让模型应用到不同的场景中。","link1":"https://modelscope.cn","btn1":"社区详情","link2":"https://modelscope.cn/docs","btn2":"文档中心","btn3":"","link3":"","infoGroup":[{"infoName":"立即体验","infoContent":{"firstContentName":"模型库","firstContentLink":"https://modelscope.cn/models","lastContentName":"数据集","lastContentLink":"https://modelscope.cn/datasets"}},{"infoName":"热门活动","infoContent":{"firstContentName":"社区 VIP 招募激励计划","firstContentLink":"https://modelscope.cn/activitys/kol"}}]}],"card":[],"search":[],"infoCard":[{"bannerUrl":"https://img.alicdn.com/imgextra/i3/O1CN01WCQZsb1ICYCctdAi1_!!6000000000857-55-tps-240-240.svg","bannerTitle":"智能计算灵骏","bannerContent":"智能计算灵骏是阿里云自主研发的新一代智能计算产品，提供深度优化的异构计算融合集群实例，满足人工智能、大数据、高性能计算等多种应用场景。","liveButtonName":"查看详情","liveButtonLink":"https://www.aliyun.com/product/ai/lingjun","contentTitle":"","homePageLink":"https://help.aliyun.com/document_detail/444430.html","homePageName":"产品文档","linkGroup":[]}],"title":{"mainTitle":"","subtitle":"","linkUrl":"","btnText":""},"visual":{"topbg":"https://img.alicdn.com/tfs/TB1bQuBIYH1gK0jSZFwXXc7aXXa-3840-740.gif","icon":"https://img.alicdn.com/imgextra/i3/O1CN01WCQZsb1ICYCctdAi1_!!6000000000857-55-tps-240-240.svg","textColor":"dark"},"dataList":[],"sceneCard":[],"txt":[]}

{"$env":{"JSON":{}},"$page":{"env":"production"},"$context":{"optioninfo":{"dynamic":"ture","static":"true"},"simplifiedDisplay":"newEdition","newCard":[{"ifIcon":"img","link":"https://img.alicdn.com/tfs/TB1XY8hGYr1gK0jSZFDXXb9yVXa-1740-328.png","icon":"ModelScope 魔搭社区","iconImg":"https://img.alicdn.com/imgextra/i3/O1CN01WCQZsb1ICYCctdAi1_!!6000000000857-55-tps-240-240.svg","contentLink":"https://modelscope.cn","title":"ModelScope 魔搭社区","des":"为实现“模型即服务”理念，魔搭社区提供最新最热、开放开源的多领域预训练模型和优质数据集，让广大开发者无需写代码就能快速体验模型效果；同时提供抽象编程接口及SDK，对模型进行二次开发，真正让模型应用到不同的场景中。","link1":"https://modelscope.cn","btn1":"社区详情","link2":"https://modelscope.cn/docs","btn2":"文档中心","btn3":"","link3":"","infoGroup":[{"infoName":"立即体验","infoContent":{"firstContentName":"模型库","firstContentLink":"https://modelscope.cn/models","lastContentName":"数据集","lastContentLink":"https://modelscope.cn/datasets"}},{"infoName":"热门活动","infoContent":{"firstContentName":"社区 VIP 招募激励计划","firstContentLink":"https://modelscope.cn/activitys/kol"}}]}],"card":[],"search":[],"infoCard":[{"bannerUrl":"https://img.alicdn.com/imgextra/i3/O1CN01WCQZsb1ICYCctdAi1_!!6000000000857-55-tps-240-240.svg","bannerTitle":"智能计算灵骏","bannerContent":"智能计算灵骏是阿里云自主研发的新一代智能计算产品，提供深度优化的异构计算融合集群实例，满足人工智能、大数据、高性能计算等多种应用场景。","liveButtonName":"查看详情","liveButtonLink":"https://www.aliyun.com/product/ai/lingjun","contentTitle":"","homePageLink":"https://help.aliyun.com/document_detail/444430.html","homePageName":"产品文档","linkGroup":[]}],"title":{"mainTitle":"","subtitle":"","linkUrl":"","btnText":""},"visual":{"topbg":"https://img.alicdn.com/tfs/TB1bQuBIYH1gK0jSZFwXXc7aXXa-3840-740.gif","icon":"https://img.alicdn.com/imgextra/i3/O1CN01WCQZsb1ICYCctdAi1_!!6000000000857-55-tps-240-240.svg","textColor":"dark"},"dataList":[],"sceneCard":[],"txt":[]}}