Trying to backward through the graph a second time, but the saved intermediate results have already

最新推荐文章于 2023-12-29 00:00:58 发布
最新推荐文章于 2023-12-29 00:00:58 发布
阅读量506 收藏 1
点赞数
文章标签： pytorch
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/weixin_37198422/article/details/132588103
版权
文章讲述了在使用PyTorch时遇到的错误，涉及到尝试多次反向传播但中间结果已被释放的情况。解决方法是确保在第一次调用backward时设置retain_graph=True，并指出当模型从numpy加载数据并作为输入时，需要使用detach()消除梯度问题。
摘要由CSDN通过智能技术生成

在使用pytorch时报错：

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 the first time.

之后我loss里加了retain_graph=True也没有用

后来发现，模型中利用torch.from_numpy加载了外部数值变量，作为模型的输入：

shuzhi0=torch.from_numpy(np.array(ls_shuzhi))

然后打印查看shuzhi0这个tensor，发现他本身存在梯度：

..., grad_fn=<StackBackward>

因此我们需要消除他的梯度：

shuzhi0=torch.from_numpy(np.array(ls_shuzhi)).detach()
人间真正走过~
关注
  • 0
    点赞
  • 踩
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
Netbeans 6.0 JRuby on Rails 演示 .swf
12-17
Netbeans 6.0 JRuby on Rails 演示 .swf
Pytorch中Trying to backward through the graph和one of the variables needed for gradient错误解决方案
rothschild666的博客
04-14 5567
Trying to backward through the graph a second time one of the variables needed for gradient computation has been modified by an inplace operation
pytorch报错：RuntimeError: Trying to backward through the graph a second time
qq_40300056的博客
09-24 4727
在进行Gan的训练过程中，经常会遇到这个问题：RuntimeError: Trying to backward through the graph a second time, but the saved intermediate results have already been freed 这个错误的描述就是，你在进行方向传播的时候，缓存提前释放了，很多解决办法都是说在loss.backward() 改成loss.backward(retain_graph=True)，实际上，大部分的代码问题不出在这里
【笔记】RuntimeError: Trying to backward through the graph a second time：将无关变量的梯度回传关系撤销
nyist_yangguang的博客
09-09 836
grad_fn 关系，梯度回传的目的是为了更新网络中权重和偏移量的值，所有与梯度回传过程中计算权重和偏移量无关的变量均应 .detach() 处理，避免发生此类错误。解决办法： 首先应该明白一件事，与XXX.backward() 中 XXX 有计算关系的所有变量均存在。
RuntimeError: Trying to backward through the graph a second time, but the saved intermediate results
weixin_43967256的博客
03-12 1097
F.normalize() 的锅！
RuntimeError: Trying to backward through the graph a second time...
热门推荐
axept的博客
03-04 1万+
RuntimeError: Trying to backward through the graph a second 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. 最开始我按照这个提示将原代码中bac
【pytorch异常处理】使用释放的图资源
安安爸Chris的专栏
05-12 527
【error】Trying to backward through the graph a second time, but the saved intermediate results have already been freed RuntimeError: Trying to backward through the graph a second time, but the saved intermediate results have already been freed. Specify reta
Docker”Got permission denied while trying to connect to the Docker daemon socket”报错解决
01-07
报错信息： ”Got permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Get ...原因： docker mannual： Manage Docker as a non-root user ...
Docker出现Cannot connect to the Docker daemon. Is the docker daemon running on this host错误解决办法
09-30
主要介绍了Docker出现Cannot connect to the Docker daemon. Is the docker daemon running on this host错误解决办法的相关资料,需要的朋友可以参考下
Were trying to save the manateesPPT教案学习.pptx
10-01
Were trying to save the manateesPPT教案学习.pptx
We're trying to save the manateesPPT教案学习.pptx
10-01
We're trying to save the manateesPPT教案学习.pptx
辽宁省丹东七中2015九年级英语下册 Unit 13 We are trying to save the earth! Sect
08-18
例如："We are trying to save the earth!" 此句即为现在进行时，表示我们正在努力拯救地球。 2. **被动语态**：被动语态用于强调动作的承受者，而非执行者，构成形式为"be + 过去分词"。例如："The environment is...
八年级英语下册 Unit 6 An old man tried to move the mountains Section A同
09-08
3. trying (愚公不断尝试，不放弃) 4. is (Neither of you作主语，谓语动词用单数) 5. taking (对拍照感兴趣) IV. 单项填空： 1. B. remind, of (这些照片能让我想起住在乡村的生活) 2. B. looks like (桂林的山看...
new_header_test.rar_The First
09-24
Bug: g++ generates an error trying to generate the first foo<int>, when it should silently fail and go on to the next one.
fp_entry.rar_The Process
09-20
we jump here after an access error while trying to access | user space, we correct stackpointer and send a SIGSEGV to | the user process.
【报错】：RuntimeError: Trying to backward through the graph a second time, but the saved intermediate re
子根的博客
08-18 2273
这个错误，一般是训练模型中有多个（两个或以上）的损失函数计算，导致的，所以在第一个loss.backward（）加上retain_graph=True。即：loss.backward（retain_graph=True）但是针对于我的问题，我的模型类似GAN模型的。在鉴定模型和生成模型之间要在使用一次鉴定模型。...
关于Trying to backward through the graph a second time, but the saved intermediate 的问题
allrubots的博客
12-19 1193
简单的记录一下
loss回传报错RuntimeError: Trying to backward through the graph a second time解决方法
最新发布
weixin_53638141的博客
12-29 1087
因为数据原地置换产生的RuntimeError: Trying to backward through the graph a second time解决方法
RuntimeError: Trying to backward through the graph a second time (or directly acess saved tensors after they have already been freed). Saved intermediate value5 of the graph are freed when you call .backward() or autograd.grad(). Specifyetain graph=True if you need to backward through the graph a second time or iiou need to access saved tensors after calling backward
07-16
在PyTorch中，默认情况下，计算图中的中间结果（intermediate values）会在调用.backward()或autograd.grad()后被释放，以节省内存。 如果您需要在两次反向传播之间访问保存的张量或需要通过计算图访问保存的中间...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值