Use visualizations to explore your logged data, the relationships between hyperparameters and output metrics, and dataset examples.
Here's an example of a parallel coordinates plot. In this case I've chosen four vertical axes. In this case I'm visualizing the relationship between the learning rate, learning rate decay, and IOU on different categories of point clouds. See it live →
Deselect a selection on the parallel coordinates chart by clicking outside the selected region on an axis. Here's an example:
This is likely because you have dots in your config parameter names. We flatten nested parameters using dots, and we only handle 3 levels of dots in the backend. I recommend using a different character as a separator.
Turn on the "eye" icon next to the top run in each group to visualize the maximum accuracy on the charts
You can download charts by clicking the down arrow and select a format (.png, .svg, export API or export by CSV)