Faceswap Forum

billubakra wrote: ↑Wed Nov 09, 2022 4:56 pm

Also where exactly to change the gpu setting in task manager to cuda usage? Can't find it there too.

I don't run Windows, but under the performance settings in Task Manager (the page which shows the graphs) there was a pull down item which let you select Cuda instead of 3D. This was Windows 10. Things may have changed. Possibly related, not tested: https://answers.microsoft.com/en-us/win ... 2666bdd87f

Is the gpu being used? Please let me know how to change the gpu settings to cuda in task manager? It will be more convenient.

It shows that the python process is using the GPU, so I would say that it is most likely being used, yes. I'm not sure why GPU utilization is not being shown for your GPU though.

app.php/rules#rule-4b
Nothing happens when I click on output system information. Asking for future references, where is this saved?

The information should be output to the console part of the GUI (the bottom text area). I don't know why this wouldn't be output. You can try the cli method by opening an anaconda prompt, activating your faceswap environment, entering the faceswap directory and entering the command listed. ie:

Start > Anaconda Prompt (cmd)

conda activate faceswap
cd faceswap
python -c "from lib.sysinfo import sysinfo ; print(sysinfo)"

Make sure you change to the correct folder for the cd faceswap command

Are these the guides and faq's that you are referring to for the masker, alignment and optimum settings or are there any other too?
app.php/faqpage
viewtopic.php?t=146

Yes. Also:

• Windows Installation Guide:: https://faceswap.dev/forum/viewtopic.php?f=4&t=20
• Extraction Guide : viewtopic.php?f=25&t=27
• Training Guide: viewtopic.php?f=27&t=146
• Convert Guide: viewtopic.php?f=24&t=1083

torzdf wrote: ↑Thu Nov 10, 2022 9:58 am

billubakra wrote: ↑Wed Nov 09, 2022 4:56 pm

Also where exactly to change the gpu setting in task manager to cuda usage? Can't find it there too.

I don't run Windows, but under the performance settings in Task Manager (the page which shows the graphs) there was a pull down item which let you select Cuda instead of 3D. This was Windows 10. Things may have changed. Possibly related, not tested: https://answers.microsoft.com/en-us/win ... 2666bdd87f

Is the gpu being used? Please let me know how to change the gpu settings to cuda in task manager? It will be more convenient.

It shows that the python process is using the GPU, so I would say that it is most likely being used, yes. I'm not sure why GPU utilization is not being shown for your GPU though.

app.php/rules#rule-4b
Nothing happens when I click on output system information. Asking for future references, where is this saved?

The information should be output to the console part of the GUI (the bottom text area). I don't know why this wouldn't be output. You can try the cli method by opening an anaconda prompt, activating your faceswap environment, entering the faceswap directory and entering the command listed. ie:

Start > Anaconda Prompt (cmd)

Code: Select all

conda activate faceswap
cd faceswap
python -c "from lib.sysinfo import sysinfo ; print(sysinfo)"

Make sure you change to the correct folder for the cd faceswap command

Are these the guides and faq's that you are referring to for the masker, alignment and optimum settings or are there any other too?
app.php/faqpage
viewtopic.php?t=146

Yes. Also:

• Windows Installation Guide:: https://faceswap.dev/forum/viewtopic.php?f=4&t=20
• Extraction Guide : viewtopic.php?f=25&t=27
• Training Guide: viewtopic.php?f=27&t=146
• Convert Guide: viewtopic.php?f=24&t=1083

Thank you for replying

From what I have read on various forums the task manager in windows 11 does not show cuda values/parameters. Does this prove that FaceSwap is using the gpu? Ran this while FaceSwap was running.

Will get back with the output file in a bit.

Was checking some other image swapping app, though it is very inferior to FaceSwap but at the same time it uses less resources. Anyways it needed to run some commands with Conda. Conda was installed our system during the installation of FaceSwap only. Now a command of that app has broken that app and as per Github Conda should be reinstalled. The current Conda version installed by FaceSwap in our system is Miniconda3 py39_4.12.0 (Python 3.9.12 64-bit) If this is uninstalled manually and then the latest version is installed from here https://www.anaconda.com/products/distribution then will it cause any problems to FaceSwap or with our current projects?

Thank you for all the guide links. Will be checking them out.

Is FaceSwap available with Google Collab for users having low end machines?

Say we save a project called Project 1.fsw with 1000000 iterations and when we open FaceSwap the next time we start a new project and save it as Project 2.fsw in a different folder. Now can we switch back and forth for training? Is it possible, if Face B is same in both projects, to use the already trained Face B for 100000 iterations in Project 2? If this info is in the guides then please say so, will definitely check it later.

billubakra wrote: ↑Thu Nov 10, 2022 4:14 pm

From what I have read on various forums the task manager in windows 11 does not show cuda values/parameters. Does this prove that FaceSwap is using the gpu? Ran this while FaceSwap was running.

Will get back with the output file in a bit.

I couldn't say. The only way I will be able to know for sure is with the system info output I asked for.

Was checking some other image swapping app, though it is very inferior to FaceSwap but at the same time it uses less resources. Anyways it needed to run some commands with Conda. Conda was installed our system during the installation of FaceSwap only. Now a command of that app has broken that app and as per Github Conda should be reinstalled. The current Conda version installed by FaceSwap in our system is Miniconda3 py39_4.12.0 (Python 3.9.12 64-bit) If this is uninstalled manually and then the latest version is installed from here https://www.anaconda.com/products/distribution then will it cause any problems to FaceSwap or with our current projects?

Should be fine. I would remove the currently installed MiniConda and re-run the faceswap installer pointing at your new (ana)Conda location when asked to be sure (this will then create a new Faceswap environment in your new Conda).

Is FaceSwap available with Google Collab for users having low end machines?

No, because of this: https://www.unite.ai/google-has-banned- ... -in-colab/

Say we save a project called Project 1.fsw with 1000000 iterations and when we open FaceSwap the next time we start a new project and save it as Project 2.fsw in a different folder. Now can we switch back and forth for training? Is it possible, if Face B is same in both projects, to use the already trained Face B for 100000 iterations in Project 2? If this info is in the guides then please say so, will definitely check it later.

The .fsw just saves the settings for training a model. You would also need load in/freeze the weights. Covered in training guide.

torzdf wrote: ↑Fri Nov 11, 2022 9:46 am

billubakra wrote: ↑Thu Nov 10, 2022 4:14 pm

From what I have read on various forums the task manager in windows 11 does not show cuda values/parameters. Does this prove that FaceSwap is using the gpu? Ran this while FaceSwap was running.

Will get back with the output file in a bit.

I couldn't say. The only way I will be able to know for sure is with the system info output I asked for.

Was checking some other image swapping app, though it is very inferior to FaceSwap but at the same time it uses less resources. Anyways it needed to run some commands with Conda. Conda was installed our system during the installation of FaceSwap only. Now a command of that app has broken that app and as per Github Conda should be reinstalled. The current Conda version installed by FaceSwap in our system is Miniconda3 py39_4.12.0 (Python 3.9.12 64-bit) If this is uninstalled manually and then the latest version is installed from here https://www.anaconda.com/products/distribution then will it cause any problems to FaceSwap or with our current projects?

Should be fine. I would remove the currently installed MiniConda and re-run the faceswap installer pointing at your new (ana)Conda location when asked to be sure (this will then create a new Faceswap environment in your new Conda).

Is FaceSwap available with Google Collab for users having low end machines?

No, because of this: https://www.unite.ai/google-has-banned- ... -in-colab/

Say we save a project called Project 1.fsw with 1000000 iterations and when we open FaceSwap the next time we start a new project and save it as Project 2.fsw in a different folder. Now can we switch back and forth for training? Is it possible, if Face B is same in both projects, to use the already trained Face B for 100000 iterations in Project 2? If this info is in the guides then please say so, will definitely check it later.

The .fsw just saves the settings for training a model. You would also need load in/freeze the weights. Covered in training guide.

Thanks. I will get back to you with the output file tomorrow.
Reinstalled Anaconda from their website, for single users it is installed in C:\Users\username\anaconda3 but now when I click on FaceSwap it does not open, nothing happens. Tried running the FaceSwap installer but it says this

How to rectify this? Really don't want to go through the re installation process again.

You have to remove the folder. Ultimately, you deleted the faceswap environment when you deleted your old conda.

You can always set a different location for your new faceswap. The download and population of the faceswap folder is not the part which takes a lot of time. It is the setting up of the environment which does.

torzdf wrote: ↑Sat Nov 12, 2022 11:38 am

You have to remove the folder. Ultimately, you deleted the faceswap environment when you deleted your old conda.

You can always set a different location for your new faceswap. The download and population of the faceswap folder is not the part which takes a lot of time. It is the setting up of the environment which does.

Got it. As of now I have set up an environment for some other app from Github in the Anaconda installed manually using the commands that they have shared. If I remove that FaceSwap folder and resintall FaceSwap, then will both these apps work fine in Conda? Isn't there a way to manually add the FaceSwap environment in Conda via some command, rather than deleting everything and then reinstalling?

Cuda in task manager in Windows 11 starts to show up after following these steps
Go to Settings -> System -> Display -> Graphics settings (down the bottom of page) -> Hardware-accelerated GPU scheduling -> Switch to "Off". After reboot the CUDA option had returned to the graphs in the Task Manager GPU window.

billubakra wrote: ↑Sat Nov 12, 2022 12:50 pm

Got it. As of now I have set up an environment for some other app from Github in the Anaconda installed manually using the commands that they have shared. If I remove that FaceSwap folder and resintall FaceSwap, then will both these apps work fine in Conda? Isn't there a way to manually add the FaceSwap environment in Conda via some command, rather than deleting everything and then reinstalling?

Yes: https://github.com/deepfakes/faceswap/b ... al-install

torzdf wrote: ↑Mon Nov 14, 2022 10:15 am

billubakra wrote: ↑Sat Nov 12, 2022 12:50 pm

Got it. As of now I have set up an environment for some other app from Github in the Anaconda installed manually using the commands that they have shared. If I remove that FaceSwap folder and resintall FaceSwap, then will both these apps work fine in Conda? Isn't there a way to manually add the FaceSwap environment in Conda via some command, rather than deleting everything and then reinstalling?

Yes: https://github.com/deepfakes/faceswap/b ... al-install

Followed like half of those steps but left in the middle as if we mess up something then we might mess up with the other app too, whose environment is working in Conda as of now. If we delete the faceswap folder from C:Users folder and then re-run the FaceSwap installer file, then will it redownload and overwrite the existing Conda? Doing this can impact the working of the other app too as that app requires manual installation and is very very hard to set up. If the Faceswap installer won't reinstall Conda but will run the needful commands etc. in the background to setup a working Environment for FaceSwap then it would be great.
Have set up FaceSwap in a machine with a 2060 gpu, will post the results on this forum.

billubakra wrote: ↑Mon Nov 14, 2022 10:51 am

If we delete the faceswap folder from C:Users folder and then re-run the FaceSwap installer file, then will it redownload and overwrite the existing Conda?

Not if you pay attention to the installer prompts. Most likely existing Conda install will be detected, but if not:

from: viewtopic.php?t=20

torzdf wrote: ↑Tue Nov 15, 2022 9:52 am

billubakra wrote: ↑Mon Nov 14, 2022 10:51 am

If we delete the faceswap folder from C:Users folder and then re-run the FaceSwap installer file, then will it redownload and overwrite the existing Conda?

Not if you pay attention to the installer prompts. Most likely existing Conda install will be detected, but if not:
conda.jpg
from: viewtopic.php?t=20

Hi,
FaceSwap is crashing in one of the machines. Have shared with you the crash report file, please check your direct message.

No algorithm worked!  Error messages:
  Profiling failure on CUDNN engine 1#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 69210240 bytes.
  Profiling failure on CUDNN engine 1: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 69210240 bytes.
  Profiling failure on CUDNN engine 0#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 16777216 bytes.
  Profiling failure on CUDNN engine 0: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 16777216 bytes.
  Profiling failure on CUDNN engine 2#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 29884416 bytes.
  Profiling failure on CUDNN engine 2: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 29884416 bytes.
  Profiling failure on CUDNN engine 5#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 2405433344 bytes.
  Profiling failure on CUDNN engine 5: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 2405433344 bytes.
	 [[{{node dfaker/encoder/conv_1024_1_conv2d/Conv2D_1}}]] [Op:__inference_train_function_9766]

search.php?keywords=RESOURCE_EXHAUSTED

torzdf wrote: ↑Wed Nov 16, 2022 10:21 am

Code: Select all

No algorithm worked!  Error messages:
  Profiling failure on CUDNN engine 1#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 69210240 bytes.
  Profiling failure on CUDNN engine 1: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 69210240 bytes.
  Profiling failure on CUDNN engine 0#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 16777216 bytes.
  Profiling failure on CUDNN engine 0: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 16777216 bytes.
  Profiling failure on CUDNN engine 2#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 29884416 bytes.
  Profiling failure on CUDNN engine 2: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 29884416 bytes.
  Profiling failure on CUDNN engine 5#TC: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 2405433344 bytes.
  Profiling failure on CUDNN engine 5: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 2405433344 bytes.
	 [[{{node dfaker/encoder/conv_1024_1_conv2d/Conv2D_1}}]] [Op:__inference_train_function_9766]

search.php?keywords=RESOURCE_EXHAUSTED

Thanks. Please check your DM.