Stay organized with collections
Save and categorize content based on your preferences.
Prerequisites and system recommendations
Python 3.11 or 3.12 is required to use Meridian.
We also recommend using a minimum of 1 GPU.
Install
To install Meridian, run the following command to automatically install the most
recent published version from PyPI.
Linux (GPU)
python3-mpipinstall--upgrade'google-meridian[and-cuda]'# Verify the installation:python3-c"import meridian; print(meridian.__version__)"
macOS
python3-mpipinstall--upgrade'google-meridian'# Verify the installation:python3-c"import meridian; print(meridian.__version__)"
CPU
python3-mpipinstall--upgrade'google-meridian'# Verify the installation:python3-c"import meridian; print(meridian.__version__)"
Latest
python3-mpipinstall--upgradegit+https://github.com/google/meridian.git
# Verify the installation:python3-c"import meridian; print(meridian.__version__)"
How to use the Meridian library
You can run the Meridian code programmatically using sample data with
the Getting started colab.
The Meridian model uses a holistic MCMC sampling approach called
No U Turn Sampler (NUTS)
which can be compute intensive. To help with this, GPU support has been
developed across the library (out-of-the-box) using tensors. We recommend
running your Meridian model on GPUs to get real time optimization results and
significantly reduce training time.
[null,null,["Last updated 2025-06-24 UTC."],[[["\u003cp\u003eMeridian requires Python 3.11 or 3.12 to operate.\u003c/p\u003e\n"],["\u003cp\u003eIt is recommended to use at least 1 GPU for optimal performance, with testing performed on V100 and T4 GPUs with 16 GB of RAM.\u003c/p\u003e\n"],["\u003cp\u003eMeridian can be installed using the command: \u003ccode\u003epip install --upgrade google-meridian\u003c/code\u003e.\u003c/p\u003e\n"],["\u003cp\u003eMeridian utilizes a compute-intensive No U Turn Sampler (NUTS) approach, and leveraging GPUs is advised to speed up optimization and reduce training times.\u003c/p\u003e\n"]]],["Meridian requires Python 3.11 or 3.12 and recommends at least 1 GPU (V100 or T4 with 16GB RAM tested). Installation via `pip` uses: `python3 -m pip install --upgrade 'google-meridian[and-cuda]'` (Linux/GPU), `google-meridian` (macOS/CPU). CPU-only install also uses `google-meridian`. To verify the install, run: `python3 -c \"import meridian; print(meridian.__version__)\"`. The library, which uses No U-Turn Sampler, is compute-intensive, thus GPU usage is recommended for real-time optimization and faster training.\n"],null,["# Install Meridian\n\nPrerequisites and system recommendations\n----------------------------------------\n\nPython 3.11 or 3.12 is required to use Meridian.\n\nWe also recommend using a minimum of 1 GPU.\n| **Note:** This project has been tested on V100 and T4 GPU using 16 GB of RAM.\n\nInstall\n-------\n\nTo install Meridian, run the following command to automatically install the most\nrecent published version from PyPI. \n\n### Linux (GPU)\n\n**Note:** CUDA toolchain and a compatible GPU device is necessary for `[and-cuda]` extra to activate. For CPU-only environment, see: \"CPU\". \n\n python3 -m pip install --upgrade 'google-meridian[and-cuda]'\n # Verify the installation:\n python3 -c \"import meridian; print(meridian.__version__)\"\n\n### macOS\n\n**Note:** There is no official GPU support for macOS. \n\n python3 -m pip install --upgrade 'google-meridian'\n # Verify the installation:\n python3 -c \"import meridian; print(meridian.__version__)\"\n\n### CPU\n\n python3 -m pip install --upgrade 'google-meridian'\n # Verify the installation:\n python3 -c \"import meridian; print(meridian.__version__)\"\n\n### Latest\n\n python3 -m pip install --upgrade git+https://github.com/google/meridian.git\n # Verify the installation:\n python3 -c \"import meridian; print(meridian.__version__)\"\n\nHow to use the Meridian library\n-------------------------------\n\nYou can run the Meridian code programmatically using sample data with\nthe [Getting started colab](/meridian/notebook/meridian-getting-started).\n\nThe Meridian model uses a holistic MCMC sampling approach called\n[No U Turn Sampler (NUTS)](https://www.tensorflow.org/probability/api_docs/python/tfp/experimental/mcmc/NoUTurnSampler)\nwhich can be compute intensive. To help with this, GPU support has been\ndeveloped across the library (out-of-the-box) using tensors. We recommend\nrunning your Meridian model on GPUs to get real time optimization results and\nsignificantly reduce training time."]]
