Doc comment on text2vec-transformers (/modules/retriever-vectorizer-modules/text2vec-transformers)

[giscus[bot]]

Doc comment on text2vec-transformers (/modules/retriever-vectorizer-modules/text2vec-transformers)

Introduction

https://weaviate.io/developers/weaviate/modules/retriever-vectorizer-modules/text2vec-transformers

[scepeda78]

if a custom vectorized is used, then for the query WCS would need that we have the local server with the vectorized running, to vectorize the query?

[hbsardhara09]

How to create a vector using "text2vec-transformers" module? I want to use "nearVector" method but there is no way to create vector using "text2vec-transformers" module. I am using this module as my default vectorizer.

[peterhessey]

When trying to run the docker compose file listed under "Option 1: Use a pre-built transformer model container", I get the following error if try to set ENABLE_CUDA: 1:

embeddings-t2v-transformers-1  | RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
embeddings-t2v-transformers-1  | 
embeddings-t2v-transformers-1  | ERROR:    Application startup failed. Exiting.

Any ideas on what is causing this? NVIDIA and CUDA are set up on properly on my machine (confired with nvidia-smi.

[peterhessey]

confirmed with nvidia-smi)*

[zachtrong]

Try this, expose gpus from host to docker container, it worked for me with docker on Windows WSL.
https://stackoverflow.com/questions/70761192/docker-compose-equivalent-of-docker-run-gpu-all-option

[daviibf]

Within a same schema, a have two different topic-related classes that are referenced one by the other. I was willing to vectorize one of those classes using a pre-built text2vec transformer image, at the same time I vectorize the other using a custom text2vec transformer image. Is it possible to do it ?

[mikeloh77]

I am following the steps for Option 2. Unfortunately when testing the vectorizer

curl localhost:9090/vectors -H 'Content-Type: application/json' -d '{"text": "foo bar"}'

I get a 404 error

{"code":404,"message":"path /vectors was not found"}%       

I think there must be a bug in the custom image or the documentation is out of date. As I have no way of inspecting the available routes I can not confirm this or figure out a solution. Please help.

[dudanogueira]

Hi! You need to add

    ports:
    - 9090:8080

to the t2v-transformers docker service, Like so:

  t2v-transformers:
    image: semitechnologies/transformers-inference:sentence-transformers-multi-qa-MiniLM-L6-cos-v1
    environment:
      ENABLE_CUDA: 0 # set to 1 to enable
    ports:
    - 9090:8080

[khaddeep143]

Works fine but if run on many data, I am getting below error {'error': [{'message': 'update vector: send POST request: Post "http://t2v-transformers:8080/vectors": dial tcp 172.19.0.2:8080: connect: connection refused'}]}