---
id: fasttext-numpy2
version: "0.10.4"
license: MIT
license_treatment: permissive
maintenance: dormant
---

# fasttext-numpy2 — fasttext Python bindings, fixed numpy 2 compatibiliy

License: permissive · Maintenance: dormant · Downloads: 1.0M/mo

## What it is and what it does

fasttext-numpy2 is a fork of Facebook's fastText library that adds compatibility with numpy 2, addressing breaking changes in the original package. It provides Python bindings for training unsupervised word representation models (skipgram and CBOW) and supervised text classifiers. The package handles UTF-8 encoded text input, tokenizes it, and learns dense vector representations of words or classifies documents based on labeled training data.

The library depends on pybind11 for Python-C++ interop, setuptools for building, and numpy for numerical operations. It ships precompiled wheels for Python 3.6 through 3.13 on Linux x86_64, reducing installation friction for those platforms. Model objects can be saved, loaded, and quantized to reduce file size with minimal accuracy loss.

Use it for:

- Train word embeddings on raw text corpora for downstream NLP tasks like similarity search or clustering.
- Build and evaluate text classifiers on labeled datasets with built-in precision/recall metrics.
- Compress trained supervised models using quantization to deploy smaller, faster inference artifacts.
- Migrate existing fastText code to numpy 2 environments without rewriting bindings.
- Extract word vectors for a given vocabulary to use as features in other machine learning pipelines.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

fasttext-numpy2 provides Python bindings for fastText, a library for learning word representations and training text classifiers, with compatibility fixes for numpy 2.

Yes, if you need fastText with numpy 2 support and accept the dormant maintenance status. The package is stable (0.10.4 release), has no known vulnerabilities, and carries a permissive MIT license. However, the repository has not been updated in 644 days and is archived; consider whether you can tolerate lack of future bug fixes or feature updates. For active development or long-term production use, evaluate whether the original fastText package or an actively maintained alternative better fits your risk tolerance.

## Install

pip install fasttext-numpy2
uv add fasttext-numpy2
poetry add fasttext-numpy2

## Installing fasttext-numpy2

Before you install: Medium install friction due to compiled C++ dependencies (pybind11, setuptools) and binary wheels for multiple Python versions (3.6–3.13). Last release was 644 days ago; repository is archived and dormant, suggesting limited ongoing maintenance.

License in practice: MIT license is permissive and poses no restrictions on commercial or private use, modification, or redistribution.

Quickstart:

pip install fasttext-numpy2

import fasttext

# Train unsupervised word vectors
model = fasttext.train_unsupervised('data.txt', model='skipgram')
print(model['king'])  # retrieve word vector

Requires numpy and a C++11-capable compiler; text input must be UTF-8 encoded.

Verify before relying:

- Whether numpy 2 compatibility extends to all fastText features or only core functionality.
- Performance or accuracy differences compared to the original fastText package.
- Active maintenance status and responsiveness to bug reports given the dormant repository state.

## Package facts

- License: MIT (permissive)
- Python support: unspecified
- Install friction: medium
- Maintenance: dormant
- Downloads: 1.0M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

word embeddings python, text classification fasttext, word vectors library, fasttext numpy 2 compatible, sentence embeddings, unsupervised word representation, text classification model training, word-embeddings, text-classification, numpy2-compat

[View on SkillFed](https://skillfed.io/packages/fasttext-numpy2) · [View on PyPI](https://pypi.org/project/fasttext-numpy2/)