# CyberZHG/keras-bert

Implementation of BERT that could load official pre-trained models for feature extraction and prediction

Repository: https://github.com/CyberZHG/keras-bert
Canonical: https://ross.abutalabs.com/products/keras-bert
Language: Python
License: MIT
License Family: permissive
Topics: keras, bert, language-model
Archived: true
Last push: 2022-01-22T10:33:11+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2875, "days_push": 1684, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2419, forks 498 (observed 2026-08-28T04:06:50.477407+00:00)

## What it is
A Keras implementation of BERT that can load official Google pre-trained checkpoints for feature extraction and prediction. It includes a tokenizer, model builder, and training utilities for fine-tuning on tasks like classification.

## Use cases
- load official pretrained BERT models in Keras
- extract BERT embeddings from text
- fine-tune BERT for text classification
- predict masked words in a sentence
- tokenize text into BERT wordpiece indices
- run BERT on TPU

## When to choose
- you're on a Keras/TensorFlow stack and need BERT with official checkpoint compatibility
- you want a lightweight pip-installable BERT for feature extraction or fine-tuning

## When to avoid
- you need modern transformer features or active maintenance
- you're using PyTorch or Hugging Face Transformers
- you need newer models like RoBERTa or DeBERTa

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, nlp, sdk
- domain: deep-learning, machine-learning
- platform: python
- tags: bert, keras, transformers, pretrained-models, language-model, feature-extraction, tokenizer, natural-language-processing

## Member repositories
- CyberZHG/keras-bert (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:50.477407+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:33:05.538167+00:00, confidence not recorded.
  - readme: https://github.com/CyberZHG/keras-bert (fetched 2026-08-28T04:06:50.477407+00:00, sha 908ddac041d1)
  - registry_pypi: https://pypi.org/pypi/keras-bert/json (fetched 2026-08-29T10:13:28.374137+00:00, sha d0bf4e289dca)
- Data as of 2026-08-30T08:39:29.467469+00:00.