Yolov7:oserror: [winerror 1455] the page file is too small to complete the final solution of the operation

Catalog

1. Problem description ：

 2. Existing solutions

（1） Restart the computer （ tried , But it doesn't work ）

（2） restart pycharm( The cost is very low , You can try it , But it's still useless )

（3） The hottest saying is setting num_work=0( Modify with caution )

3. Problem analysis

4. Strategies to solve

4. summary  

---

1. Problem description ：

Everyone in pytorch Frame usage YOLOv7 This kind of problem may often be encountered during deep learning ：

OSError: [WinError 1455] The page file is too small , Unable to complete operation . Error loading "D:\ProgramData\Anaconda3\envs\yolov7\lib\site-packages\torch\lib\cudnn_adv_infer64_8.dll" or one of its dependencies.

As shown in the figure, this situation ：

 2. Existing solutions

（1） Restart the computer （ tried , But it doesn't work ）

         Many bloggers talk about Running your computer for a long time causes the computer memory to be occupied by irrelevant software , So it is recommended to restart the computer , But the problem is still found after restarting

        So this solution cannot solve the problem

（2） restart pycharm( The cost is very low , You can try it , But it's still useless )

        because pycharm Running for a long time will occupy a lot of memory / memory , So restarting is still helpful for clearing memory .

        But for this The model of deep learning Come on , This kind of memory cleaning / Explicit memory is “ an utterly inadequate measure ”, It's not enough , So this scheme can be tried when the model is small .

         It is not very good for such a large model as the following figure ：

          therefore , about YOLO Training model , This plan is not feasible

（3） The hottest saying is setting num_work=0( Modify with caution )

        There are also online tutorials on this , But most bloggers only say that there is this operation , But it's not friendly to Xiaobai , Now let's introduce , How to modify in the integrated environment num_work Parameters of ：

        1） Find the one under the environment train.py

        2)  Drag it to the middle and find it as shown in the following figure workers, Take the back one default Change the value of to 0

        But! , Can see , What this method changes is dataloader, Changing this place requires caution , This involves your overall code framework , So I decided to abandon this method （1. Changed or failed ;2. Even if I can change here in the future , But this is a way to cure the symptoms rather than the root cause ）, This method can only solve this problem here , If you change this position , that It is likely to lead to some subsequent problems , So it's not recommended .

        This method did not solve the problem in the end

3. Problem analysis

        Back to the problem , What I said is that the page is too small , That's the default Out of memory 了 , Then solve this problem directly

OSError: [WinError 1455] The page file is too small , Unable to complete operation .

        Don't make some strange things , Simple and crude , Once and for all .

4. Strategies to solve

（1） Find the disk installed in your environment , The blogger himself uses anaconda Virtual environment to manage the relevant compilation environment , Deep learning also recommends using virtual environments , Easy to switch , About how to use and manage virtual environments , You can refer to my other basic dry goods article ：

anaconda The creation of virtual environment 、 Delete and view _ Youth is short ！ The blog of -CSDN Blog

        my python The environment is placed in D Discoid ：

 （2） stay “ My computer ” Right click , open “ attribute ”

（3） Click on Advanced system setup

 （4） Click on “ Set up ”

（5） find “ senior ”

（6） Click on “ change ”

 （7） Distribute “ Virtual memory ”

From top to bottom , Follow my steps in turn

（ You need to restart your computer to take effect ）

thus , After restarting the computer , You can run the model directly ！

（8） Solve the problem successfully , And achieved good results  

4. summary  

          After using this method , You can use the storage space of your hard disk to allocate memory space , So here we see the importance of a high-performance solid-state drive for deep learning , After that , The storage space you allocate will be used as memory to participate in deep learning , The system cannot use it as ordinary storage space for the time being , If you need to restore, go back according to the steps just now , Click on the top of this setting “ Automatically manage the size of all pagers ”, Next time you encounter this problem, you can directly click , The size of system management can be achieved in one step , There is no need to restart the computer , This method can be described as once and for all .

        If my method solves your problem , You can give me a compliment , Add attention and work together ！！