Training Option : Advanced Evaluation
This section introduces the different advanced evaluation options for training.
Advanced Evaluation is an option that Nexus provides to assist with visualization and manual checking of the training as it is ongoing. Advanced evaluation takes evaluations from each evaluation checkpoint. The frequency of these checkpoints is decided in Training Option : Checkpoint Strategy.
Why do we need Advanced Evaluation?
These tools can be very useful in monitoring the progress in a much more visual and meaningful manner rather than just statistics that only describe the overall progress and not necessarily the individual quality of predictions. It can also be a useful tool for spotting errors or identifying if a problem is occurring in your training early on.
Training Multiplier
Enabling Advanced Evaluation will will incur a 1.5x multiplier in the consumption of training minutes due to the addition of monitoring and computational tools that will be added to the model training process. The list of tools is shown below:
| Tool | Description |
|---|---|
| Evaluation Preview | Allows you to examine inference previews at each evaluation checkpoint to make comparisons about the training as it progresses. To see how it works, go to Evaluation Preview. |
| Confusion Matrix | Allows you to visualize differences in distributions between prediction classes and ground truth classes using color gradients. To see how it works, go to Confusion Matrix. |
| Recommendations | Provides targeted actionables to improve subsequent training runs based on your dataset and training setup. To see how it works, go to Recommendations. |
👋 Need help? Contact us via website or email
🚀 Join our Slack Community
💻 For more resources: Blog | GitHub | Tutorial Page
🛠️ Need Technical Assistance? Connect with Datature Experts or chat with us via the chat button below 👇
Updated 9 days ago