If customer has mounted model artifacts to /opt/ml/model, and also specified HF_MODEL_ID=/opt/ml/model, then the container fails to startup with
ERROR ModelServer Invalid configuration
ai.djl.serving.http.BadRequestException: Workflow model is already registered.
at ai.djl.serving.models.ModelManager.registerWorkflow(ModelManager.java:92) ~[serving-0.28.0.jar:?]
at ai.djl.serving.ModelServer.initModelStore(ModelServer.java:444) ~[serving-0.28.0.jar:?]
at ai.djl.serving.ModelServer.start(ModelServer.java:208) ~[serving-0.28.0.jar:?]
at ai.djl.serving.ModelServer.startAndWait(ModelServer.java:174) ~[serving-0.28.0.jar:?]
at ai.djl.serving.ModelServer.main(ModelServer.java:143) [serving-0.28.0.jar:?]
Issue is that when adding to the list or urls, we returned the local path which is different than the url so it looks like 2 different models, but at workflow register level they are the same
Description
If customer has mounted model artifacts to /opt/ml/model, and also specified HF_MODEL_ID=/opt/ml/model, then the container fails to startup with
Issue is that when adding to the list or urls, we returned the local path which is different than the url so it looks like 2 different models, but at workflow register level they are the same