AI & Machine Learning 3d ago 5 views 3 min read

Installing PyTorch 2.3 with CUDA on Ubuntu 24.04

Get PyTorch 2.3.0 installed on Ubuntu 24.04 with CUDA 12.1 support. Follow these steps to configure the environment and verify GPU access.

A
Arjun M.
Updated 3h ago
Sponsored

Cloud VPS — scale in minutes

Instantly deploy SSD cloud VPS with guaranteed resources, snapshots and per-hour billing. Pay only for what you use.

Install PyTorch 2.3.0 with CUDA support on Ubuntu 24.04. This guide covers downloading the official installer, setting environment variables, and verifying GPU access. The steps target a fresh Ubuntu 24.04 LTS system with an NVIDIA GPU.

Prerequisites

  • Ubuntu 24.04 LTS (Noble Numbat) installed and updated.
  • NVIDIA driver version 550 or newer installed.
  • CUDA Toolkit 12.1 or newer installed.
  • Python 3.10 or newer installed.
  • pip 22.0 or newer installed.
  • sudo privileges to install system packages.

Step 1: Update system packages

Ensure the package manager has the latest repository data before installing dependencies. Run the update command to pull the latest versions of core system tools.

sudo apt update && sudo apt upgrade -y
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.

Step 2: Install system dependencies

Install the build essentials and Python development libraries required to compile PyTorch extensions. This step ensures the environment has the necessary compilers and headers.

sudo apt install -y build-essential python3-dev python3-pip libopenblas-dev liblapack-dev libomp-dev
Reading package lists... Done
Building dependency tree... Done
0 newly installed, 0 to remove and 0 not upgraded.
Need to get 0 B of archives.

Step 3: Install PyTorch with CUDA

Download the official PyTorch installer script from the official website. Use the specific command that targets CUDA 12.1 and Python 3.10 for Ubuntu 24.04. The script automatically handles the pip installation and environment setup.

curl -o get-pytorch.sh https://download.pytorch.org/whl/cu121/torch_stable.sh
% Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  25.6M  100  25.6M    0  0   1234k      0  0:00:20  0:00:20 --:--:0 1234k
bash get-pytorch.sh
Collecting torch==2.3.0+cu121
  Downloading https://download.pytorch.org/whl/cu121/torch-2.3.0%2Bcu121-cp310-cp310-manylinux2014_x86_64.whl (800.9 MB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Collecting torchvision==0.18.0+cu121
Collecting torchaudio==2.3.0+cu121
Collecting filelock
Collecting typing-extensions
Collecting sympy
Collecting networkx
Collecting jinja2
Collecting fsspec
Collecting triton==3.0.0
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12
Collecting nvidia-cuda-nvrtc-cu12
Collecting nvidia-cuda-runtime-cu12
Collecting nvidia-cufft-cu12
Collecting nvidia-curand-cu12
Collecting nvidia-cusolver-cu12
Collecting nvidia-cusparse-cu12
Collecting nvidia-nccl-cu12
Collecting nvidia-nvtx-cu12
Collecting nvidia-nvjitlink-cu12
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12==12.1.3.1
Collecting nvidia-cuda-cupti-cu12==12.1.105
Collecting nvidia-cublas-cu12==12.1.3.1
Collecting nvidia-cufft-cu12==11.0.0.100
Collecting nvidia-cudnn-cu12==9.1.0.70
Collecting nvidia-cusparse-cu12==12.1.3.1
Collecting nvidia-nccl-cu12==2.20.5
Collecting nvidia-nvtx-cu12==12.1.105
Collecting nvidia-nvjitlink-cu12==12.5.80
Collecting nvidia-curand-cu12==10.3.2.109
Collecting nvidia-cusolver-cu12==11.4.1.48
Collecting nvidia-cuda-runtime-cu12==12.1.3.1
Collecting nvidia-cuda-nvrtc-cu12== Python
Sponsored

Powerful Dedicated Servers — Linux & Windows

Bare-metal performance with SSD storage, DDoS protection and 24/7 expert support. Ideal for production workloads, databases and high-traffic sites.
Tags: UbuntuGPUPyTorchDeep LearningCUDA
0
Was this helpful?

Related tutorials

AI & Machine Learning
How to quantize a large language model for edge devices

Convert a 7B parameter model to 4-bit GGUF format for inference on Raspberry Pi 4 or Jetson Nano using llama.c…

12
AI & Machine Learning
How to set up a RAG pipeline with LangChain and Weaviate

Build a retrieval-augmented generation system using Python, LangChain, and Weaviate vector database. Install d…

6
AI & Machine Learning
How to quantize a Llama 3 model for inference on a Raspberry Pi

Reduce memory usage and speed up Llama 3 inference on Raspberry Pi by quantizing the model weights to 4-bit or…

6

Comments 0

Login to leave a comment.

No comments yet — be the first to share your thoughts.