Something about seeing these joblib files made me wonder if it could be helpful to switch to onnx format sometime and explore the possibilities. You could use sklearn-onnx to implement the models and perhaps onnxruntime to run the models. There could be performance and other benefits to making this shift. Hat tip to @MattsonCam, who mentioned ONNX to me a bit ago.
Originally posted by @d33bs in #17 (comment)
Something about seeing these
joblibfiles made me wonder if it could be helpful to switch toonnxformat sometime and explore the possibilities. You could usesklearn-onnxto implement the models and perhapsonnxruntimeto run the models. There could be performance and other benefits to making this shift. Hat tip to @MattsonCam, who mentioned ONNX to me a bit ago.Originally posted by @d33bs in #17 (comment)