Bug Reports

Repro steps: * Load a dataset with mutliple columns (>10) * Create a model where all columns are used * Start training the model Expected outcome: The statistics view should be able to handle multiple inputs without breaking. Actual outcome: You are not able to see all inputs without zooming out the browser.

Repro steps: * Create any model * Start training the model * Pause the training Expected outcome: The training should now have been paused in the kernel and the Pause icon should have changed to a Play icon Actual outcome: The training does not pause in the kernel (you can see it keep training) and the icon does not change in the frontend.

Repro steps: * Create any model * Start training it * Stop the training * Go to the Modeling Tool quickly after Expected outcome: The Run button should say Run. Actual outcome: In the Modeling Tool you'll see Stop instead of Run still since you moved to it before the Kernel status has been updated (as stopping can take a moment).

Repro steps: * Create any model * Open Data Settings from within the workspace * Change the dataset path to another dataset (can be an identical dataset just copied somewhere else on the computer) * Press Save to apply the change * Press on the Input or Target component Expected Outcome: The path in the component settings should be the new path you just changed to in the Data Settings. Actual Outcome: The path in the component settings is the old path rather than the new one. Noticed in release version: 0.12.24

This would make it a lot easier to share models and datasets from PerceptiLabs with other people.

The CSV goes off-screen, and selecting the Input/Target column causes it to jump around, if the number of columns (combined with their name length) is larger than what fits in the modal.

Running >10k images through the pre-processing pipeline seemingly takes much longer than it should. Need to look into if it's just the Resize option which does this, or if there is something else we can optimize.

Manually building a U-Net is a good example of this, where it will start going slower until finally it crashes and the frontend changes to a white screen.

If there is more than 2 inputs they won't properly connect to the merge component.