Tensor visualization pipeline

[vkkhare]

Feature Description

The grid when starts to support the view for a data scientist (to some extent for data owner as well) needs to provide better context for tensors.

1. First of all tensors shouldn't just be dataset tensors but any intermediate tensor generated in the process.
2. These tensors should be searchable/filterable/categorized by data scientists in the UI (for data owner)
3. We should be able to view how these tensors are related to each other in the computation pipeline. Here a view like tensorboard graph might be very helpful

The above is a gif of how it looks for a computation graph

[tcp]

Great challenge!

Many changes are needed to create such viz and interpretation in Grid but I'm willing to invest time and effort to do something along these lines.

Looks awesome and can be extremely useful.