data-team-uhn / NeuralCR

MIT License
26 stars 12 forks source link

Feature: multi model #16

Closed IntegralProgrammer closed 4 years ago

IntegralProgrammer commented 4 years ago

The PR allows for multiple trained models to be loaded and allows for the trained model to be selected on calls to /match/ or /annotate/. For calls to /annotate/, multiple trained models may be specified. For example, if the trained models HP, MONDO, and ICD are loaded, text can be annotated with /annotate/ using both the HP and MONDO models but not the ICD model.

Reverse-compatibility is maintained with the current version of the API as if only one model loaded and the model parameter is not specified in the /match/ or /annotate/ request, said model will be used.