Training Neural Network Model from DGX Spark using tensorflow and Keras

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
1

Hello Yall ..

I am trying to train couple of NN Models using tensorflow and keras . I am expecting to see the GPU utilization in the dashboard when I perform the training . Unfortunately looks like the GPU is not utilized.

What am I missing .

Thanks in Advance.

2

The easiest way, I read, to utilize the GPU is to work within a Docker container that can properly access the GPU. See this thread: Unable to work with TensorFlow in Docker on DGX Spark - #3 by ejd