Installation

CHIA can be installed easily using pip. We recommend using CHIA in a conda environment (see the Miniconda install guide). If you plan on using any of our provided docker containers, you should pin your conda environment to Python version 3.10.19, in order to match the python version used inside the Dockerized workers.

The easiest way to get that is a dedicated conda environment. We recommend naming the conda environment the same thing on all machines you plan on using CHIA for, as it makes the setup of CHIA clusters easier. We will assume throughout these docs that this environment is called chia_env.

conda create -n chia_env python=3.10.19
conda activate chia_env

Clone the CHIA repository onto any machines you plan on using

git clone https://github.com/ucb-bar/chia.git

Then install the package in editable mode from your clone:

pip install -e /path/to/chia

This installs the chia package and it’s dependencies. We plan on releasing a PyPI CHIA release in the future.

Optional extras

Extra

Installs

Use for

tensorboard

tensorboardX, tensorboard

metrics logging

wandb

wandb

metrics logging

metrics

both of the above

metrics logging

postgres

psycopg[binary]

a Postgres-backed DatabaseNode

 
# For one extra
pip install -e "/path/to/chia[metrics]"

# For multiple extras
pip install -e "/path/to/chia[metrics,postgres]"

Core dependencies

Chia pins ray[default]==2.54.0 along with, mcp, pydantic, fastapi, boto3, and the Google GenAI / Vertex client libraries. See pyproject.toml for the full pinned set.

Next steps