BERT-for-Unity

Bidirectional Encoder Representations from Transformers technique for Unity game engine using huggingface implementation. This is a server-based interfaces for huggingface transformers "Pipeline" objects. Pipeline are high-level objects which automatically handle tokenization, running your data through a transformers model and outputting the result in a structured object.

Install

1. setup virtualenv and activate your environment
2. install transformers
3. clone this repository and install the dependencies

pip install flask
pip install flask_cors
pip install waitress

Usage

Server

start app

cd BERT-for-Unity/
python3 app.py

server is tested on Python 3.5+, PyTorch 1.0.0+

Unity

Supported pipeline objects

• next-sentence : Provide the next N sentences for the input sequence, it will consider the return as the new input during iteration.
• fill-mask : Takes an input sequence containing a masked token (e.g. ) and return list of most probable filled sequences, with their probabilities.
• question-answering : Provided some context and a question refering to the context, it will extract the answer to the question in the context.
• sentiment-analysis : Gives the polarity (positive / negative) of the whole input sequence.
• feature-extraction : Generates a tensor representation for the input sequence

StartCoroutine(transformers.task("next_sentence","I never thought it would be this hard to create #3",flask_url,next_sentence_queue));

StartCoroutine(transformers.task("fill_mask","I never thought it would be this <mask> to build a house",flask_url,next_sentence_queue));

StartCoroutine(transformers.task("question_answering","Who was Jim Henson?#Jim Henson was a nice puppet",flask_url,q_a_queue));

StartCoroutine(transformers.task("sentiment_analysis","i love you",flask_url,sentiment_analysis_queue));

StartCoroutine(transformers.task("feature_extraction","i love you",flask_url,feature_extraction_queue));

for more details see unity/Assets/Simple_BERT_Usage.cs. Open unity/Assets/bert_example_scene.unity to use the example scene.

Example BERT webgl app WIP

cd BERT-for-Unity/
python3 app.py -webgl true

and visit http://localhost:5000

Current status, February 25th:

this is the first protype so there are no tests available.