Hi all, I created OPENNLP-1515 to change the ONNX Runtime dependency from onnxruntime-gpu to onnxruntime. This change will remove GPU support and cause OpenNLP to always use CPU for inference. The reason for this change is the onnxruntime dependency supports Linux, Windows, and Mac x64, and the onnxruntime-gpu dependency only supports Linux and Windows. ( https://onnxruntime.ai/docs/get-started/with-java.html) I think OpenNLP should support the most operating systems out of the box instead of favoring GPU. Please take a look at the pull request: https://github.com/apache/opennlp/pull/551
This change is partially proposed to support OpenNLP's ONNX integration in Apache Solr: https://github.com/apache/solr/pull/1999 I think GPU support in OpenNLP should be easily accessible to users, so I wrote OPENNLP-1516 to capture that with a link to one possible method. If the above PR is merged, a user can still enable GPU in OpenNLP by manually replacing the onnxruntime.jar with onnxruntime-gpu.jar on their classpath until OPENNLP-1516 is resolved. All comments/suggestions are welcome! Thanks, Jeff
