Evaluate a Model

Learn how to evaluate models in the platform

Once the model and code integration are uploaded, we can start a model evaluation.

The Evaluate Model operation inferences the dataset throughout the model, and collects performance metrics and metadata for each sample. It runs the Tensorleap explainability engine over an extracted latent space to find issues within the model and dataset and to provide insights on how to diagnose and correct these. After it is finished, the collected data is presented on the Dashboard.

More than one evaluation can be performed on a model, since the same model could be configured to use different code integration scripts which would effect data loading and metrics computations.

The Evaluation panel

To run an evaluation:

1. Click the button at the top right to open the Evaluate menu. More info about the properties below:

2. Select a Model Version

3. Provide a Model Run Name

4. (Optional) provide a description of the Run

5. Set the Batch Size.

6. Click .

Model Evaluation Video Tutorial