What Is MegaMolBART gRPC service?

MegaMolBART is a model trained on SMILES string and this container deploys MegaMolBART model for infrencing. The container bring up a gRPC service available in port 500051.

Getting Started

MegaMolBART gRPC service is currently used with Cheminformatics Demo application (also available as a container). Please refer to Cheminformatics resource for bringing up the complete suite.

Please following these steps to start MegaMolBART gRPC service in stand-alone mode.

• Download MegaMolBART model or execute the following command to do so

ngc registry model download-version "nvidia/clara/megamolbart:0.1.2"

• Start an instance of the Docker image using the following command:

# E.g. for Docker version 19.03 or later
docker run \
--gpus all \
--rm \
-v $(pwd)/megamolbart_v0.1/:/models/megamolbart \
nvcr.io/nvidia/clara/megamolbart:0.1.2

# E.g. for Docker version 19.02 or later
docker run \
--runtime nvidia \
--rm \
-v $(pwd)/megamolbart_v0.1/:/models/megamolbart \
nvcr.io/nvidia/clara/megamolbart:0.1.2

Interface

Protobuf definition of this service is available in Cheminformatics resource .

Please use the following command to generate language specific stubs.

python -m grpc_tools.protoc -I./grpc/ \
 --<>_out=generated \
 --experimental_allow_proto3_optional \
 --grpc_python_out=generated \
 ./grpc/generativesampler.proto

The service has three functions:

• SmilesToEmbedding: Returns the latent space embedding of an input SMILES string.
• FindSimilars: Generates SMILES using the model by probing the vicinity of input SMILES's position in latent space.
• Interpolate: Generates SMILES using the model by probing the latent space between two input SMILES.
• GetIteration: Returns the checkpoint iteration version of the deployed model.

Current usage

• Cheminformatics

License

By pulling and using the container, you accept the terms and conditions of Apache 2.0 license