smartonfhir/cnlp-transformers repository overview

⁠Background

This image provides CNLP-transformers⁠ pre-built for CPU/NVIDIA CUDA GPU support for inclusion in dockerized applications. See the repository for more details on configuration.

There is one image per model, which is ready to use with minimal additional configuration. The example compose.yaml⁠ shows how this can be run as part of a network, and how a different model could be specified. You can also run the image directly with docker run smartonfhir/cnlp-transformers:[model_name]. The REST service is exposed on port 8000 by default.

⁠Using GPU acceleration

To leverage an NVIDIA GPU for CUDA acceleration, you will need to do some configuration to expose the GPU to the docker engine from the underlying host system. We've tested this on CentOS7/Ubuntu 22 on a G4dn EC2 box. You can use the following links to find the specific steps for your platform:

• https://docs.nvidia.com/datacenter/tesla/tesla-installation-notes/⁠
• https://docs.nvidia.com/cuda/cuda-installation-guide-linux/⁠
• https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html⁠