Using Registered Model to make predictions without having to deploy it in Azure ML.

Sena Aslan 25

I have trained a model in azure ml designer pipeline using "boosted decision tree regression" and "tune model hyperparameters" components. I have registered this model using "register model" button in "tune model hyperparameters" component. Now i need to get this model to make predictions in azure ml notebook. I am getting the registered model with below code, however when i try to make predictions with the model it throws : AttributeError: 'Model' object has no attribute 'predict'. So, how can i solve this without having to deploy the model ?

#get registered model
model = ml_client.models.get(name="my_model_name", version="1")
model

1 answer

dupammi 8,185 Reputation points Microsoft Vendor

2024-02-26T07:31:28.6766667+00:00

Hi @Sena Aslan

Thank you for using the Microsoft Q&A forum.

To use a registered model for making predictions in Azure ML Notebook, you need to download the model artifact and load it into your notebook. You can download the model artifact from the registered model using the Model.get_model_path() method and load the model from the .pkl file using the joblib.load() function.

Once the model is loaded, you can prepare input data and then, you can make predictions using the predict method of the loaded model.

You can find more information about - How to work with models in Azure Machine Learning, here.

I hope you understand. Thank you.

If this answers your query, do click Accept Answer and Yes for was this answer helpful.
Please sign in to rate this answer.
Sena Aslan 25 Reputation points

2024-02-26T15:30:38.9766667+00:00

Hi @dupammi , I tried the method you mentioned. However, i got the error: IsADirectoryError: [Errno 21] Is a directory: 'trained_model_outputs'. So, i know my model is in the 'trained_model_outputs' folder in artifacts. I also tried writing the model path like this : '.../trained_model_outputs/data.ilearner' . When i tried the model path like that i got : 'ModuleNotFoundError: No module named 'azureml.studio'' but i have this module in my workspace. Can you please help me ? Here is my code :

import joblib from azureml.core import Model,Workspace ws = Workspace.from_config() model_name = 'my_model_name' model_obj = Model(ws, model_name ) model_path = model_obj.download(exist_ok = True) model = joblib.load(model_path)

dupammi 8,185 Reputation points Microsoft Vendor

2024-02-26T16:38:50.6266667+00:00

Hi @Sena Aslan

The error message "IsADirectoryError: [Errno 21] Is a directory: 'trained_model_outputs' suggests that the path specified in the model_path variable is a directory instead of a file.

You can try modifying the model_path variable to point to the specific file within the directory that you want to load. For example, if the file you want to load is named model.pkl, you can modify the model_path variable as follows:

model_path = os.path.join(model_path, 'model.pkl')

This will append the file name to the directory path, creating the full path to the file. Then, you can try loading the model again using:

joblib.load(model_path)

I hope you understand. Thank you.

If this answers your query, do click Accept Answer and Yes for was this answer helpful.

Sena Aslan 25 Reputation points

2024-02-26T17:21:31.1733333+00:00

Hi @dupammi , I tried it as i mention in my previous question. I got this : ModuleNotFoundError: No module named 'azureml.studio' Trained-Best-Model folder contain the following files _meta.yaml _samples.json _schema.json conda_env.yaml data.ilearner model_spec.yaml score.py . And my model file is data.ilearner file. I guess I'am getting this error because of the format of my model. When you train a model in azure designer, model format is data.ilearner, it's not in pkl or json format. So how can i load these '.ilearner' models to get predictions before the deployment ?

dupammi 8,185 Reputation points Microsoft Vendor

2024-02-27T03:02:18.9733333+00:00

Hi @Sena Aslan Based on my understanding, you are using Azure Machine Learning Designer to train a model, and the model is saved in the .iLearner binary format. Unfortunately, it seems that this format cannot be modified directly. Please see here from documentation.

What training produces is a specific binary format, the iLearner, that encapsulates the statistical patterns learned from the data. You cannot directly modify or read this format; however, other components can use this trained model.

Per my research, the ilearn interface is supported on Azure Machine Learning Studio(classic), but I cannot find any official document indicating this is supported on Azure Machine Learning Designer as well. Sorry for the experience. However, the .iLearner format is supported in Azure Machine Learning Studio (classic), as mentioned in the documentation here. I hope this information helps. Thank you.

Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.
Sign in to comment

Share via

Using Registered Model to make predictions without having to deploy it in Azure ML.

1 answer