mars-project/mars

Health v2 · maintenance only

Adoption not part of the score

2742 stars · 324 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Mars is a tensor-based unified framework for large-scale data computation that provides NumPy-, pandas-, and scikit-learn-compatible APIs which scale to distributed and out-of-memory workloads. It lets existing Python data science code run on a local session or a Mars cluster with minimal code changes.

Use cases

• scale numpy computations beyond a single machine's memory
• run pandas dataframes on a cluster
• distribute scikit-learn model training
• process large tensors in parallel
• dask alternative for distributed dataframes
• out-of-core data processing in python

When to choose

• you have NumPy/pandas/sklearn code that outgrows one machine and want a familiar API
• you need distributed execution with lazy evaluation and explicit .execute() control
• you want a lighter-weight alternative to Spark or Dask for Python data workloads

When to avoid

• you need the most active ecosystem - Dask or Ray may have larger communities
• your data fits comfortably in memory on one machine
• you need deep Spark/Hadoop ecosystem integration

Facets

library · maturity maintenance

data-science machine-learning etl streaming concurrency data-science big-data machine-learning microservices python cloud distributed-computing numpy-compatible pandas-compatible tensor dataframe scale-up dask-alternative data-engineering linux macos docker kubernetes

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mars-project/mars")