GPU Dashboard

The GPU dashboard in OOD JupyterLab includes GPU and CPU memory and utilization tracking in real time. This is helpful for GPU selection.

Memory Usage

PyTorch

Correct GPU memory usage will be reported.

TensorFlow/ Keras

By default, TF automatically allocates ALL of the GPU memory so GPU Dashboard may show that all (or almost all) of the GPU memory is being used.

To track the amount of GPU memory actually used, you can add these lines to your python script:

import os
os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true'

More Info: https://www.tensorflow.org/guide/gpu#limiting_gpu_memory_growth

Homework for Keras users: try out GPU dashboard and see if it reports all of the GPU memory as used