How to Install Ultralytics YOLO with Conda#

This guide walks through setting up a Conda environment for your Ultralytics projects. Conda is an open-source package and environment management system that offers an excellent alternative to pip for installing packages and dependencies. Its isolated environments make it particularly well-suited for data science and machine learning work. For more details, visit the Ultralytics Conda package on Anaconda and check out the Ultralytics feedstock repository for package updates on GitHub.

This guide covers how to create an environment with Ultralytics, run inference, use the Conda Docker image, and speed up installs with libmamba.

Prerequisites#

You should have Anaconda or Miniconda installed on your system. If not, download and install it from Anaconda or Miniconda.

Installing Ultralytics#

Create a new Conda environment with Ultralytics from the conda-forge channel, then activate it. Install Python, Ultralytics, and torchvision (not a dependency of the conda-forge package) in the same command so conda resolves current, compatible versions together:

conda create --name ultralytics-env -c conda-forge python=3.11 ultralytics torchvision -y
conda activate ultralytics-env

conda create --name ultralytics-env -c conda-forge python=3.11 ultralytics torchvision pytorch-gpu -y
conda activate ultralytics-env

Note on ROCm Environment (AMD GPU, Linux)#

For AMD GPU acceleration on Linux using ROCm, PyTorch ROCm wheels are not available via Conda and must be installed with pip. Use a Conda environment for isolation, then install PyTorch ROCm and Ultralytics via pip:

conda create --name ultralytics-rocm python=3.12 -y
conda activate ultralytics-rocm
pip install "torch[device-all]" "torchvision[device-all]" --index-url https://stable.repo.amd.com/rocm/whl-next/
pip install ultralytics

See the PyTorch Get Started page for supported ROCm versions and the AMD GPU guide for MIGraphX inference setup, which requires Python 3.11 or newer on Linux x86_64.

Using Ultralytics#

With Ultralytics installed, you can now start using its robust features for object detection, instance segmentation, and more. For example, to predict an image, you can run:

from ultralytics import YOLO

model = YOLO("yolo26n.pt")  # initialize model
results = model("path/to/image.jpg")  # perform inference
results[0].show()  # display results for the first image

Ultralytics Conda Docker Image#

The Conda Dockerfile uses conda-forge PyTorch 2.13, torchvision 0.28, and CUDA 13.0. Conda-forge releases can lag the PyTorch pip releases used by the standard GPU image. Automated publishing of latest-conda is disabled, so build the current image locally:

git clone https://github.com/ultralytics/ultralytics
cd ultralytics
t=ultralytics-conda:local
sudo docker build -f docker/Dockerfile-conda -t $t .

Run the image:

# Run the Ultralytics image in a container with GPU support
sudo docker run -it --ipc=host --device nvidia.com/gpu=all $t                         # all GPUs
sudo docker run -it --ipc=host --device nvidia.com/gpu=2 --device nvidia.com/gpu=3 $t # specify GPUs

On Linux, CDI device requests require Docker >= 28.2.0 and NVIDIA Container Toolkit >= 1.18. The legacy --gpus all flag can lose GPU access after host daemon reloads, so upgrade older Linux hosts and use --device instead. See the Docker Quickstart Guide for details.

Speeding Up Installation with Libmamba#

libmamba is a fast, cross-platform, dependency-aware solver that replaces Conda's classic solver. Conda 23.10 and later already use libmamba as the default solver, so most installations are faster out of the box.

If you're on an older Conda version, you can enable libmamba manually:

1. First, install the conda-libmamba-solver package:

   conda install conda-libmamba-solver

2. Next, configure Conda to use libmamba as the solver:

   conda config --set solver libmamba

You have successfully set up a Conda environment, installed the Ultralytics package, and are now ready to explore its features. For more advanced tutorials and examples, see the Ultralytics documentation.

FAQ#