错误提示：enable anomaly detection to find the operation that failed to compute its gradient...

最新推荐文章于 2024-04-24 14:59:58 发布
原创 最新推荐文章于 2024-04-24 14:59:58 发布 · 1.5k 阅读 · 0 ·
CC 4.0 BY-SA版权
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。
文章标签：

#深度学习 #python #人工智能

在运行自定义数据集上的AlexNet模型时遇到RuntimeError，错误涉及到了梯度计算中的变量版本不匹配。问题出在Dropout层的inplace属性设置为True。删除这一属性后，模型可以正常运行。解决方案是将线性分类器中Dropout层的inplace属性设为False。
部署运行你感兴趣的模型镜像

错误类型： 

今天在跑自己的数据集+AlexNet时一直报错，错误如下

错误提示：
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [544, 768]], which is output 0 of ViewBackward, is at version 1; expected version 0 instead. 
Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).

解决方案：

把网络模型线性分类器中Dropout层的inplace=True属性删除即可。

 删除前

 删除后

删除后即可正常运行

您可能感兴趣的与本文相关的镜像

ACE-Step

ACE-Step

音乐合成
ACE-Step

ACE-Step是由中国团队阶跃星辰（StepFun）与ACE Studio联手打造的开源音乐生成模型。 它拥有3.5B参数量，支持快速高质量生成、强可控性和易于拓展的特点。 最厉害的是，它可以生成多种语言的歌曲，包括但不限于中文、英文、日文等19种语言

enable anomaly detection to find the operation that failed to compute its gradient, with torch.autog
静静的学习就好
07-26 7742
关于pytorch中多个backward出现的问题：enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly (True)
Pytorch 反向传播 计算图被修改的报错
qq_46396470的博客
01-07 2111
如果需要修改张量的值，最好创建一个新的张量，而不是直接在原有张量上进行修改。上进行了原地操作（上述的修改形状操作），就可能破坏计算图，使其不可导或其他，总之是导致反向传播时无法正确计算梯度，从而引发报错。在这个例子中，label由第一条语句前向传播得到，是直接与网络的输出连在一起，后面我却对label的值进行了手动修改。那怎么解决这个问题？detach（）函数的作用是将数据和计算图分离开来，得到数据部分，与计算图再无瓜葛。报错中说，一个需要梯度计算的变量已经被原地修改了，这引发了报错。
GAN(拟合sin函数)
朴素.无恙的博客
01-07 2800
pytorch之GAN的实现(拟合sin函数) import torch import torch.nn as nn import numpy as np import matplotlib.pyplot as plt # torch.manual_seed(1) # reproducible # np.random.seed(1) # Hyper Parameters BATCH...
Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch
pain_gain0的博客
02-16 696
bug
【异常错误】Hint: enable anomaly detection to find the operation that failed to compute its gradient
Drug discovery
04-24 559
main(args)
关于nn.BCEWithLogitsLoss()与nn.CrossEntropyLoss()的手动计算过程(都可用于多分类，不仅仅用于二分类)
weixin_43724657的博客
10-18 2553
nn.BCEWithLogitsLoss()与nn.CrossEntropyLoss()的手动计算过程
enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
05-04
Anomaly detection is a feature in PyTorch that helps you find the operation that failed to compute its gradient. When an error occurs during the backward pass, PyTorch will throw an error and print ...
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [16, 16]], which is output 0 of AsStridedBackward0, is at version 4; expected version 3 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
最新发布
12-16
在 PyTorch 中，`RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation` 错误通常是由于在计算图构建过程中使用了原地操作（inplace operation），导致...
[torch.FloatTensor [*, *]] is at version 2； expected version 1 instead
weixin_47875944的博客
03-26 424
[torch.FloatTensor [*, *]] is at version 2； expected version 1 instead
解决UserWarning: Error detected in CudnnBatchNormBackward0. No forward pass information available.
前行居士
10-20 9258
在DistributedDataParallel加上broadcast_buffers=False 问题解决。分布式训练时出现如下报错。
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace
m0_48095841的博客
10-22 405
报错： RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [544, 768]], which is output 0 of ViewBackward, is at version 1; expected version 0 instead. Hint: enable anomaly detect.
one of the variables needed for gradient computation has been modified by an inplace operation
weixin_43252352的博客
06-11 509
bug提示： one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [256, 1]], which is output 0 of TBackward, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the ope
两个pytorch小bug
lyy的博客
09-07 1736
No.1 报错内容:RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [128, 1]], which is output 0 of TBackward, is at version 2; expected version 1 instead. Hint: enable anomaly dete
报错解决：one of the variables needed for gradient computation has been modified by an inplace operation
热门推荐
y4ung
05-12 5万+
写完了训练的代码，运行了以后发现了这么一个错： RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [544, 768]], which is output 0 of ViewBackward, is at version 1; expected version 0 instead. Hint: en
莫烦大神的pytorch教程，406-GAN，在torch 1.5版本上运行会报错的解决方法
孙权打的博客
07-22 7041
莫烦大神的pytorch教程，406-GAN，在torch 1.5版本上运行会报错 RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [128, 1]], which is output 0 of TBackward, is at version 2; expected version 1 inste
autograd error
weixin_41318625的博客
04-03 492
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.LongTensor [896, 1]] is at version 11539; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed
错误记录：RuntimeError: one of the variables needed for gradient computation has been modified by an inpl
adminHD的博客
11-17 936
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [2, 256, 20, 20]], which is output 0 of struct torch::autograd::CopySlices, is at version 3; expected version 1 instead.
评论
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
红包 添加红包
表情包 插入表情
表情包 代码片
 条评论被折叠 查看
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