Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

[Feature Request]: allow to specify which gpu to use

See original GitHub issue

As far as I could see, there is no such option. To clarify, this is useful when running a script on a machine with more than one GPU, having a flag such as --gpu 1 or --device cuda:1 would be of great help!

Thanks for all the work!

Issue Analytics

State:
Created 4 years ago
Comments:6 (5 by maintainers)

Top GitHub Comments

2reactions

sw005320commented, Mar 3, 2020

The GPU ID specification is different from the environment, and we did not prepare such options (we use to have that though). The safe way is to do that through CUDA_VISIBLE_DEVICES, e.g., CUDA_VISIBLE_DEVICES=0,1,2 ./run.sh --ngpu 3

0reactions

kamo-naoyukicommented, Mar 4, 2020

Oh, this is my fault and must be a bug, but It’s just giving args.ngpu with None to train(), so simply the program will be aborted.

Top Results From Across the Web

Detecting GPU Features and Metal Software Versions

Detecting GPU Features and Metal Software Versions. Use the device object's properties to determine how you perform tasks in Metal.

Schedule GPUs - Kubernetes

Configure and schedule GPUs for use as a resource by nodes in a cluster.

Enabling GPU access with Compose - Docker Documentation

Enabling GPU access to service containers . Docker Compose v1.28.0+ allows to define GPU reservations using the device structure defined in the...

Using Graphical Processing Units (GPUs) - GitLab Docs

GitLab Runner supports the use of Graphical Processing Units (GPUs). The following section describes the required configuration to enable GPUs for various ...

A guide to GPU implementation and activation - TechTarget

There are two main ways to enable GPU resources. The first approach is to install the GPU subsystem as an aftermarket upgrade onto...