soco-encoders 0.2.9.2

Neural encoders for embedding diverse objects

Encoders

A library of neural encoders to convert any types of data into embeddings, dense or sparse.

Build a package

python setup.py bdist_wheel
twine upload dist/*

Use Locally

model = EncoderLoader.load_model('pretrain-models', <model_id>, use_gpu=True, region='cn')
model.encode(['I am a good man'], show_progress_bar=True, batch_size=batch_size,)

Use as RESTful API

Start a server

uvicorn soco_encoders.http.main:app --host 0.0.0.0 --port 8000 --workers 4

Start a client

res = requests.post(url='http://localhost:8000/encoder/v1/encode',
                      json={
                          "model_id": model_id,
                          "text": [x1] * batch_size,
                          'batch_size': batch_size,
                          "mode": "default",
                          "kwargs": {}
                      })

Use as GRPC API

Start a server

python -m soco_encoders.grpc.server --host 0.0.0.0 --port 8000 --workers 4

Start a client

   check out example in bench_grpc.py