Download ML Package

While downloading an ML Package from AI Center in pipeline runs, what does the downloaded file contains? If I have trained model, what should it contain? If the model is not trained, then what will be the result?

Hi @aditi.kumari

Trained Model Contains:

  1. Model Artifacts: Serialized model files, which are the learned parameters and configurations of the trained model. These files can include TensorFlow SavedModel, ONNX, or other serialized formats depending on the ML framework used.
  2. Metadata: Information about the model, such as its version, training configuration, and any other relevant metadata.
  3. Evaluation Metrics (Optional): Metrics and statistics from the model evaluation process, providing insights into the model’s performance on a validation dataset.

Untrained Model Contains:

  1. The downloaded ML Package may result in an empty or incomplete package. The content may include placeholders or default configurations, but it won’t contain meaningful trained model artifacts since the training process has not been executed.

I downloaded a ML Package, a json file is getting downloaded. When opening, it gets redirected to UiPath studio. But nothing is opening.


The downloaded file is typically in JSON format, and its contents include metadata about the trained machine learning model. However, the actual trained model files, configuration files, and other artifacts may not be directly present in the JSON file. Instead, the JSON file serves as a reference or pointer to the location of these files within the AI Center environment.

How to open that json file then?


When you try to open the .json file, it might automatically open in UiPath Studio if your system is configured to associate .json files with UiPath Studio.

You can use a text editor to open the file. This way, you can inspect the metadata, configuration settings, and other details about the ML Package.

Does the downloaded file contains datasets upon which the model is being trained?

And If I want to move the trained ML package to prod environment, can I simply uploded this downloaded json file?


The downloaded file typically does not contain the training datasets.

So If we upload the downloaded file to prod server, we need to train the model again or how does that work?

It contains the information needed to recreate or load your trained model, but you don’t typically need to retrain the model unless you want to update it with new data.

The datasets used for training are usually separate data files, and the trained model is what gets saved to a file format, such as JSON.

Okay, got it. Thank you. One more thing, how can we identify if a model is trained or not in AI centre?

Check for logs or history related to the specific model to see if there are records of training sessions.

okay, thank you so much…