RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

[fatimahgit]

can anyone please help me fix this error, I trained the model by running:

python main_train.py --input_name <input_file_name>

and got this error: RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [3, 32, 3, 3]] is at version 2; expected version 1 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!

Full error message:

python main_train.py --not_cuda --input_name 33039_LR.png Random Seed: 1310 GeneratorConcatSkip2CleanAdd( (head): ConvBlock( .......... )

..\aten\src\ATen\native\TensorFactories.cpp:361: UserWarning: Deprecation warning: In a future PyTorch release torch.full will no longer return tensors of floating dtype by default. Instead, a bool fill_value will return a tensor of torch.bool dtype, and an integral fill_value will return a tensor of torch.long dtype. Set the optional dtype or out arguments to suppress this warning. Warning: Error detected in MkldnnConvolutionBackward. No forward pass information available. Enable detect anomaly during forward pass for more information. (print_stack at ..\torch\csrc\autograd\python_anomaly_mode.cpp:42) Traceback (most recent call last): File "main_train.py", line 29, in train(opt, Gs, Zs, reals, NoiseAmp) File "C:\Users\z5029775\Downloads\SinGAN-master\SinGAN\training.py", line 39, in train z_curr,in_s,G_curr = train_single_scale(D_curr,G_curr,reals,Gs,Zs,in_s,NoiseAmp,opt) File "C:\Users\z5029775\Downloads\SinGAN-master\SinGAN\training.py", line 179, in train_single_scale errG.backward(retain_graph=True) File "C:\Users\z5029775\AppData\Local\Continuum\anaconda3\envs\fatimah\lib\site-packages\torch\tensor.py", line 198, in backward torch.autograd.backward(self, gradient, retain_graph, create_graph) File "C:\Users\z5029775\AppData\Local\Continuum\anaconda3\envs\fatimah\lib\site-packages\torch\autograd_init_.py", line 98, in backward Variable._execution_engine.run_backward( RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [3, 32, 3, 3]] is at version 2; expected version 1 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!

[lioo717]

I have the same question in pytorch 1.6.0. And I change to another environment with pytorch 1.3.1, the question is fixed. I don't know the reasons, but it works. Hope this method can help you.

[tamarott]

Please use torch==1.4.0 torchvision==0.5.0.

[andrew-jeremy]

You should modify the requirements file to reflect this. Thanks!

[YL-Wang1]

任何人都可以帮我修复这个错误，我通过运行来训练模型：

python main_train.py --input_name <input_file_name>

并收到此错误：RuntimeError：梯度计算所需的变量之一已被就地操作修改：[torch.FloatTensor [3， 32， 3， 3]] 版本为 2;而是预期的版本 1。提示：上面的回溯显示未能计算其梯度的操作。有问题的变量在那里或以后的任何地方都发生了变化。祝你好运！

完整的错误消息：

python main_train.py --not_cuda --input_name 33039_LR.png 随机种子：1310 GeneratorConcatSkip2CleanAdd（ （head）： ConvBlock（ .......... )

..\aten\src\ATen\native\TensorFactories.cpp：361：用户警告：弃用警告：在未来的 PyTorch 版本中，torch.full 将不再默认返回浮动 dtype 的张量。相反，bool fill_value将返回 torch.bool dtype 的张量，而积分fill_value将返回 torch.long dtype 的张量。设置 optional 或 arguments 以禁止显示此警告。警告：在 MkldnnConvolutionBackward 中检测到错误。没有可用的正向传递信息。启用在前向传递期间检测异常，以获取更多信息。（print_stack..\torch\csrc\autograd\python_anomaly_mode.cpp：42） 回溯（最近一次调用最后一次）： 文件“main_train.py”，第 29 行，在 train（opt， Gs， Zs， reals， NoiseAmp） 文件“C：\Users\z5029775\Downloads\SinGAN-master\SinGAN\training.py”，第 39 行，在列车 z_curr，in_s，G_curr = train_single_scale（D_curr，G_curr，reals，Gs，Zs，in_s，NoiseAmp，opt） 文件“C：\Users\z5029775\Downloads\SinGAN-master\SinGAN\training.py“，第 179 行，train_single_scale errG.backward（retain_graph=True） 文件”C：\Users\z5029775\AppData\Local\Continuum\anaconda3\envs\fatimah\lib\site-packages\torch\tensor.py“，第 198 行，在 backward torch.autograd.backward（self， gradient， retain_graph， create_graph） 文件”C：\Users\z5029775\AppData\Local\Continuum\anaconda3\envs\fatimah\lib\site-packages\torch\autograd__init__.py“，第 98 行，在 backward Variable._execution_engine.run_backward（ RuntimeError：梯度计算所需的变量之一已被就地操作修改：[torch.FloatTensor [3， 32， 3， 3]] 版本为 2;而是预期的版本 1。提示：上面的回溯显示未能计算其梯度的操作。有问题的变量在那里或以后的任何地方都发生了变化。祝你好运！dtype``out

can you solve it ? if ok, help me pleae!