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

本文轉載自查看原文 2021-06-09 11:47 7014

訓練GAN net時經常遇到這個問題

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.

翻譯一下就是第二次嘗試在圖中向后遍歷時，保存的臨時變量已經被釋放

顯然，

GAN中有一個變量存在於gen和disc之間，就是fake

加上detach() 就行

免責聲明！

本站轉載的文章為個人學習借鑒使用，本站對版權不負任何法律責任。如果侵犯了您的隱私權益，請聯系本站郵箱yoyou2525@163.com刪除。

猜您在找 RuntimeError: Trying to backward through the graph a second time, but the buffers have already been freed. Specify retain_graph=True when calling backward the first time. “RuntimeError: Trying to backward through the graph a second time, but the buffers have already been freed. Specify retain_graph=True when calling backward the first time” spark-shell啟動報錯：Yarn application has already ended! It might have been killed or unable to launch application master IIS導入.pfx證書時報錯："A specified logon session does not exist. It may already have been terminated." spark 提交任務報錯 Yarn application has already ended! It might have been killed or unable to launch application master Golang把時間和數字相乘報錯invalid operation: second * time.Second (mismatched types int and time.Duration) pytorch中反向傳播的loss.backward(retain_graph=True)報錯報錯 One or more constraints have not been satisfied. 運行Maven時報錯：No goals have been specified for this build 報錯記錄:getOutputStream() has already been called for this response