The aim was to optimize the GPU usage after the profiling of the model and to figure out the minimum change in the architecture that would lead to a gain in performance and decrease training time
The GPU usage went from 9% to 60%
The aim was to optimize the GPU usage after the profiling of the model and to figure out the minimum change in the architecture that would lead to a gain in performance and decrease training time on CIFAR10 dataset on CNN architechture
Python
The aim was to optimize the GPU usage after the profiling of the model and to figure out the minimum change in the architecture that would lead to a gain in performance and decrease training time
The GPU usage went from 9% to 60%