google-deepmind/tree

Health v2 · maintenance only

Adoption not part of the score

1022 stars · 71 forks observed · 2026-08-28

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

tree (dm-tree) is a Python library for working with nested data structures, generalizing the builtin map function to apply functions to leaves while preserving structure. It is backed by an optimized C++ implementation suitable for demanding machine learning workloads.

Use cases

• flatten nested python structures into flat lists
• apply a function to every leaf of a nested structure
• manipulate model parameters stored in nested dicts and tuples
• process nested data in machine learning training pipelines
• compare or zip nested structures elementwise
• unflatten flat sequences back into nested structures

When to choose

• you need fast, C++-backed manipulation of nested data structures
• you work with ML frameworks that represent parameters as nested trees
• you want a JAX-style pytree utility for TensorFlow or NumPy code

When to avoid

• you only need simple flat list transformations with builtin map
• you need serialization to disk or network formats rather than in-memory structure ops
• your project is not Python-based

Facets

library · maturity stable

serialization data-science machine-learning machine-learning data-science developer-tools python cpp cross-platform nested-structures pytree map-structure flatten deepmind

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/tree")