Deactivated due to the usage of training data generator.

[PikaCat]

By using the release executable of the training data generator to generate training data（https://github.com/fairy-stockfish/variant-nnue-tools）, I get deactivated, so sad.

[AthiraM_Intel]

Hi,

Thank you for posting in Intel Communities.

Could you please let us know which DevCloud (oneAPI DevCloud/ DevCloud for Edge/ FPGA DevCloud) you are using?

Thanks.

[PikaCat]

It's on Intel® DevCloud: Edge Workloads -> Connect and Create -> Connect to JupyterLab. The one with 80 core CPU.

[Hari_B_Intel]

Hi PikaCat, 

Thank for contacting us, could you provide some details on how you import the fairy-stockfish in to DevCloud for edge Jupyter Notebook ? 

From there we can identify the issue you are facing and provide some helps. 

For your information, generating training data using DevCloud for edge might not be a good idea because DevCloud for edge is design for inference Engine simulation use. 

Hope this information Help

Thank you  

[PikaCat]

Hi, 

I first try to upload the compiled version of the fairy-stockfish training data generator executable into the Jupyter and run it with !./fairy-stockfish. But I found that it says that 'GLIBC 2.29 is not found', and just after that, I get disconnected forcibly from it and my DevCloud enrollment gets deactivated.

I only spend 3 minutes there and this is the first time I use it.

Because my computer only has 4 cores for data generation. It would take about 240 hours to generate enough data for training, which is kind of unacceptable and slow as hell.

After knowing that Intel provides 80 core Jupyter notebook CPU, I come up with the idea to use that for data generation.

BTW, if it is banned to generate training data like this on Intel compute, I'll just leave and think of other ways to do it. But as a student, I'm kind of hopeless because I don't have much money to rent a high-end CPU for doing that.

Once the generation is done(about 9 hours maybe, using 80 cores), I'll not do it again since it's enough data for the training process.

If I'm able to use the Intel compute again, I'll try to solve the GLIBC 2.29 problem by trying to download the source code from fairy-stockfish and compiling it on JupyterLab using CMake. I don't know if that's going to violate something or not.

Best regards

[Hari_B_Intel · Accepted solution]

Hi PikaCat

I am afraid due to INTEL® DEVCLOUD TERMS OF USE, and users are not allowed to install or modify the systems, that you might not be able to install GLIBC 2.29

Also, for your information, Intel DevCloud is only used for developing, prototyping, benchmarking, and testing AI inference applications on a broad range of Intel hardware. So, Intel DevCloud for edge might not be suitable for your current usage.

Below is the Intel Devcloud Usages policy for your reference.

https://www.intel.com/content/www/us/en/developer/articles/license/devcloud-edge-terms-and-conditions.html

I hope this information helps

Thank you

[Hari_B_Intel]

This thread will no longer be monitored. If you need any additional information from us, please submit a new question.