https://github.com/tna0y/gml

Run any process with a GPU memory limit
https://github.com/tna0y/gml

docker gpu nvidia nvidia-docker

Last synced: 7 days ago
JSON representation

Run any process with a GPU memory limit

• Host: GitHub
• URL: https://github.com/tna0y/gml
• Owner: tna0y
• Created: 2021-06-15T19:25:13.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2021-06-18T23:00:49.000Z (over 3 years ago)
• Last Synced: 2024-11-29T15:53:52.268Z (2 months ago)
• Topics: docker, gpu, nvidia, nvidia-docker
• Language: Go
• Homepage:
• Size: 10.7 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# GML - GPU memory limiter

![build](https://github.com/tna0y/gml/actions/workflows/go.yml/badge.svg)

[![GitHub release](https://img.shields.io/github/release/tna0y/gml.svg)](https://GitHub.com/tna0y/gml/releases/)

Impose a GPU memory limit on any process.

Utility was originally designed to provide a workaround for absence of GPU memory limit in `nvidia-docker` but works for any

process on Linux.

## How it works

GML runs any command and monitors its GPU memory usage. If the limit is exceded GML sends a signal (default: SIGKILL) to

the process and waits until it returns.

## Install

Pre-compiled binary is available for Linux / x86_64. Install it with following commands:

```shell script

curl -L -o gml https://github.com/tna0y/gml/releases/download/v0.1/gml-linux-x86_64

chmod +x gml && sudo cp gml /usr/local/bin/gml

```

## Usage

```

Usage: gml [--limit LIMIT] [--signal SIGNUM] -- COMMAND

      --help            show help message

  -l, --limit string    total GPU memory limit for the command (default "1MB")

  -s, --signal string   signal that will be sent to the process if the memory exceeds the limit (default "SIGKILL")

```

### Example

```shell script

gml --limit 1GB --signal SIGINT -- ./my-gpu-app arg1 arg2

```