I attempted to load the model from model registry but failed. Here is the details.

I create the experiment using mlflow.set_experiment()

experiment = mlflow.create_experiment(
        name=exp_name, 
        artifact_location="artifact_location", 
        tags={"env":"dev", "version":"1.0.0"}
    )

I used artifact_location as my artifact directory. Which is located outside of /mlruns.

Then I registered my best model in model registry with

best_run_id = best_run.run_id
model_uri = f"runs:/{best_run_id}/model"

mlflow.register_model(model_uri=model_uri, 
                      name="cat_classifier")

After that I try to load the registered model with

model_name = "cat_classifier"
model_version = 1
model_uri = f"models:/{model_name}/{model_version}"
load_model = mlflow.pyfunc.load_model(model_uri)

but failed with the following error:

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
Cell In[366], line 4
      2 model_version = 1
      3 model_uri = f"models:/{model_name}/{model_version}"
----> 4 load_model = mlflow.pyfunc.load_model(model_uri)

........

OSError: No such file or directory: '/home/User/workplace/cat/artifact_location/0eb40601413841e78f8b7b89fbf82be7/artifacts/models/.'

Then I checked /home/User/workplace/cat/artifact_location/0eb40601413841e78f8b7b89fbf82be7 which I did not found the /artifacts/models/., but I found

|-- model 
     |-- conda.yaml
     |-- MLModel
     |-- model.pkl
     |-- python_env.yaml
     |-- requirement.txt

Do I have done wrong by using artifact_location as the new path, which cause the failure of loading the registered model?

MLFlow UI showed the registered model, which I assumed it existed physically. I even successfully did alias for it, but still failed to load.

How can I solved this path issue?