jackbkennedy/sentence-encoder-lambda

Access Tensorflow's Universal sentence encoder model via a Dockerized Lambda function.

Universal Sentence Encoder Lambda

Run locally

1. Build Docker container

docker build --tag sentence-encoder-lambda .

2. Run container

docker run -p 9000:8080  sentence-encoder-lambda:latest 

3. Make a cURL request

curl -XPOST "http://localhost:9000/2015-03-31/functions/function/invocations" -d '{"text":["are you open?"]}'

Result

The endpoint returns a 512 dimension dense vector, which can be used when implementing e.g. a similarity search

{"vector":"[-0.01935427635908127, ....]"}

Resources

1. Tensorflow Hub - Universal Sentence Encoder
2. Lambda Function Handlers - Python
3. Testing Locally - Lambda Containers