Training a model

After having created a model, you can initiate the training of it. Navigate to the model details page and click on the Start training button to initate training.

A virtual machine with either a powerful Graphics Processing Unit (GPU) or a Tensor Processing Unit (TPU) will provisioned to handle the training of the model. Provisioning this virtual machine (VM) typically takes between 3 and 5 minutes. Immediately after provisioning, the data required for training the model will be downloaded onto the VM. Downloading the data can take anywhere between a couple of minutes and a few hours depending on the size of your dataset(s). After the data has been downloaded, the training phase will initiate. You can follow the training process by staying on the model details page. Training updates from the VM will automatically be sent to your browser so you can follow the process in real time.

After the training phase is complete, a model may post some evaluation data on the performance of the model. Whether the model will post evaluation data as well a the nature of such evaluation data is entirely dependent on the model type. You can learn more about what type of evaluation data you can expect from the model by reading the instructions of the model type.

In the case of the Image Classification model type, one part of the evaluation data produced by the model includes a Facets Dive presentation of the data. You can learn more about Facets Dive in the article describing how to use Facets Dive to explore datasets. See the screenshot below.

Once the model has concluded training, the timestamp for when the model has completed in the Details section will be updated. In this specific example, we can see that the entire process took about 8 minutes to complete.

This concludes the article on training models in Criterion AI.

Still need help? Contact Us Contact Us