explosion/thinc
🔮 A refreshing functional take on deep learning, compatible with your favorite libraries
Lightweight library using functional composition and type-checked APIs to build models that seamlessly wrap PyTorch, TensorFlow, and MXNet layers. Features an integrated config system for hyperparameter management and model serialization, enabling both standalone development and use as an interface layer for multi-framework model architectures.
2,893 stars. Used by 9 other packages. Actively maintained with 29 commits in the last 30 days. Available on PyPI.
Stars
2,893
Forks
294
Language
Python
License
MIT
Category
Last pushed
Feb 09, 2026
Commits (30d)
29
Dependencies
12
Reverse dependents
9
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/explosion/thinc"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related frameworks
google-deepmind/optax
Optax is a gradient processing and optimization library for JAX.
patrick-kidger/equinox
Elegant easy-to-use neural networks + scientific computing in JAX. https://docs.kidger.site/equinox/
google/grain
Library for reading and processing ML training data.
extropic-ai/thrml
Thermodynamic Hypergraphical Model Library in JAX
patrick-kidger/optimistix
Nonlinear optimisation (root-finding, least squares, ...) in JAX+Equinox....