Keras on a Radeon RX 550X

[Mohamed]

Hi all,

I have a Radeon RX 550X GPU in my Thinkpad laptop. I am wondering whether I can make use of it to run Keras codes on it. 

Best regards,

Mohamed Elwekeil.

[Eduardo A. Flores Verduzco]

Hi Mohamed, AFAIK Tensorflow does not officialy support ATI GPUs but they are working on this project ROCm which would support TF:

https://rocmdocs.amd.com/en/latest/Current_Release_Notes/Current-Release-Notes.html

https://rocmdocs.amd.com/en/latest/Deep_learning/Deep-learning.html

In the docs they mention support for TF1.15 which is quite old by today.

Hope this heps!

[Eduardo A. Flores Verduzco]

BTW, you can use TF CPU-only of course

[Mohamed]

Thank you Eduardo,

It seems from the link you shared that using TF on AMD GPUs can be done on Linux. I will appreciate that if you inform me whether this can be done on Windows as well.

[Mohamed]

Thanks,

 

I am using the TF CPU. I wanted to know the possibility of using the GPU version.

[Eduardo A. Flores Verduzco]

I really don't know if it works on Windows, I have not used it, I'm just aware it exists. I would recommend use Google Colab  or Kaggle Notebooks with free GPUs, it may be enough for many use cases, like learning of using not private data. If not, AWS, Azure or GCP as they usually offer free credits for starters. Spot or preemtible instances offer even more discount, yet they may be a little harder to setup.

Best!

[Mohamed]

Thanks for the recommendation. 

Best regards

[Angelo Varlotta]

You may also want to check PlaidAI, developed by Intel and tested on some AMD GPUs

https://www.intel.com/content/www/us/en/artificial-intelligence/plaidml.html

https://github.com/plaidml/plaidml

[Mohamed]

Hi Angelo,

Thank you very much for letting me know about the plaidML. I will check it.

[Mohamed]

Hi Angelo,

I have gone through the PlaidML website. I have managed to train CNN on the Radeon GPU. However, the training time on the GPU with PlaidML is slightly increased compared to the CPU training. I will appreciate that if you inform me what might be the reasn.

Best regards,

On Saturday, 20 June 2020 10:27:32 UTC+2, Angelo Varlotta wrote:

[Angelo Varlotta]

Hi Mohamed,
Wish I could help, but I know as much as you. PlaidAI is currently in development and new features are currently added to the development branch, but it's still very much a work in progress.

Cheers,
Angelo

--
You received this message because you are subscribed to the Google Groups "Keras-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to keras-users...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/30a61386-1326-4aed-bc08-84e575c8afcbo%40googlegroups.com.

[Mohamed Elwekeil]

Hi Angelo,

Thank you very much for your reply. I really appreciate that.

This PlaidML is interesting. I wish they will be able to provide more advances in the future.

Thanks again for letting me know about it.

All the best,

Mohamed.

--

Best regards,

Mohamed Elwekeil

[Angelo Varlotta]

You might find some help from this article:

Analyzing MacBook GPU Performance; visualizing with MLflow and Elastic Kibana - Towards Data Science

A comparison of the various MacBook PlaidML configurations for accelerating Deep Learning with AMD GPU, Intel UHD, and Intel CPU; visualized with Databricks MLflow and with Elastic Kibana.

[Mohamed Elwekeil]

Thank you very much for sharing this information. I really appreciate it.

Best regards,

Mohamed.