Skip to content

GPU allocations during each forward pass (causing memory strain) #40

Description

@IanButterworth

If you run the following on a CUDA system, you'll see gradual increase in gpu memory allocation through the repeated runs. This causes big slowdowns on memory constrained systems when the swap file starts to be used.

using ObjectDetector
ObjectDetector.benchmark(select=[3]) #run yolov3-tiny

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions