报错RuntimeError: Trying to backward through the graph a second time, but the saved intermediate results have already been freed

本文转载自查看原文 2021-06-09 11:47 7014

训练GAN net时经常遇到这个问题

RuntimeError: Trying to backward through the graph a second time, but the saved intermediate results have already been freed. Specify retain_graph=True when calling .backward() or autograd.grad() the first time.

翻译一下就是第二次尝试在图中向后遍历时，保存的临时变量已经被释放

显然，

GAN中有一个变量存在于gen和disc之间，就是fake

加上detach() 就行

免责声明！

本站转载的文章为个人学习借鉴使用，本站对版权不负任何法律责任。如果侵犯了您的隐私权益，请联系本站邮箱yoyou2525@163.com删除。

猜您在找 spark 提交任务报错 Yarn application has already ended! It might have been killed or unable to launch application master pytorch中反向传播的loss.backward(retain_graph=True)报错 R12.2 URL Validation failed. The error could have been caused through the use of the browser's navigation buttons js 报错(intermediate value)(...) is not a function variable ans might not have been initialized 报错，以及初始化注意点 jar强退出 JVM报错：Failed to write core dump. Core dumps have been disabled. SpringBoot启动报错：ould not be registered. A bean with that name has already been defined in file and overriding is disabled. 05-Vue报错 Uncaught SyntaxError: Identifier has already been declared和路由 JSON.toJSONString(joinPoint.getArgs())报错getOutputStream() has already been called for this response Android报错：WindowManager$BadTokenException: Unable to add window -- window has already been added