pluto-ml-nightly 0.0.4.dev20260207105511

Pluto ML - Machine Learning Operations Framework

Pluto is an experiment tracking platform. It provides self-hostable superior experimental tracking capabilities and lifecycle management for training ML models. To take an interactive look, try out our demo environment or get an account with us today!

See it in action

https://github.com/user-attachments/assets/6aff6448-00b6-41f2-adf4-4b7aa853ede6

🚀 Getting Started

Install the pluto-ml sdk

pip install -Uq "pluto-ml[full]"

import pluto

pluto.init(project="hello-world")
pluto.log({"e": 2.718})
pluto.finish()

• Self-host your very own Pluto instance using the Pluto Server & get started in just 3 commands with docker-compose

git clone --recurse-submodules https://github.com/Trainy-ai/pluto-server.git; cd pluto-server
cp .env.example .env
sudo docker-compose --env-file .env up --build

You may also learn more about Pluto by checking out our documentation.

Migration

Neptune

Want to move your run data from Neptune to Pluto. Checkout the official docs from the Neptune transition hub here.

Before committing to Pluto, you want to see if there's parity between your Neptune and Pluto views? See our compatibility module documented here. Log to both Neptune and Pluto with a single import statement and no code changes.

🛠️ Development Setup

Want to contribute? Here's the quickest way to get the local toolchain (including the linters used in CI) running:

git clone https://github.com/Trainy-ai/pluto.git
cd pluto
python -m venv .venv && source .venv/bin/activate   # or use your preferred environment manager
python -m pip install --upgrade pip
pip install -e ".[full]"

Linting commands (mirrors .github/workflows/lint.yml):

bash format.sh

Run these locally before sending a PR to match the automation that checks on every push and pull request.

🫡 Vision

Pluto is a platform built for and by ML engineers, supported by our community! We were tired of the current state of the art in ML observability tools, and this tool was born to help mitigate the inefficiencies - specifically, we hope to better inform you about your model performance and training runs; and actually save you, instead of charging you, for your precious compute time!

🌟 Be sure to star our repos if they help you ~