Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

keras-bert relies on features removed in keras-transformer 0.34.0

See original GitHub issue

Describe the Bug

The get_model function in the keras_bert.bert module passes a use_adapter parameter to the keras_transformers.get_encoders function: https://github.com/CyberZHG/keras-bert/blob/26bdfe3c36e77fa0524902f31263a920ccd62efb/keras_bert/bert.py#L124

However, the latest commit to keras-transformer, corresponding to the v0.34.0 release, removes this parameter from the API: https://github.com/CyberZHG/keras-transformer/commit/60f1a0968caba6adb914a14b743f0ffbdc176a18

keras-bert specifies its keras-transformer dependency as >=0.30.0. This means that new or upgraded installs will automatically use 0.34.0 despite the breaking change.

This means that calling get_model fails with the following error:

TypeError: get_encoders() got an unexpected keyword argument 'use_adapter'

The problem can be worked around by pinning the keras-transformer version at v0.33.0.

Version Info

I’m using the latest version

Minimal Codes To Reproduce

The problem can be seen by running the load_and_predict.py demo.

Issue Analytics

State:
Created 3 years ago
Comments:7 (2 by maintainers)

Top GitHub Comments

2reactions

magrathjcommented, Jun 1, 2020

Experiencing the same error when calling predict function - followed advise and changed keras-transformer==0.33.0, worked after this.

1reaction

CyberZHGcommented, Jun 11, 2020

@joseph-wakeling-frequenz Added one: https://pypi.org/project/keras-bert/0.81.1/

Top Results From Across the Web

Pretraining BERT with Hugging Face Transformers - Keras

BERT makes use of Transformer, an attention mechanism that learns contextual relations between words (or subwords) in a text.

BERT Text Classification using Keras | by Swatimeena - Medium

The BERT (Bidirectional Encoder Representations from Transformers) model was proposed in BERT: Pre-training of Deep Bidirectional Transformers for Language ...

Simple Text Multi Classification Task Using Keras BERT!

BERT relies on a Transformer (the attention mechanism that learns contextual relationships between words in a text).

COMP4901K BERT-BILSTM-CRF-best_0.001 | Kaggle

!pip install keras-bert !pip install bert-for-tf2 !pip install tf2crf ... (from keras-bert) (2.4.3) Collecting keras-transformer>=0.38.0 Downloading ...

Posit AI Blog: BERT from R - RStudio

Today, we're happy to feature a guest post written by Turgut Abdullayev, ... There are several methods to install keras-bert in Python.