Weisu Yin

@sebhrusen @PGijsbers

@sebhrusen Hey Seb, happy holidays! Can you review the code when you have time?

@sebhrusen Hi, can you review this PR when you have time? Thanks!

I think issue 2 is partly fixed by the adding of soft time limit by https://github.com/awslabs/autogluon/pull/2121.

We just installed it within orbit. @sxjscience Is that correct?

So I think tensorflow comes with tensorboard, and the gpu docker image I deployed has tensorflow. We were trying exactly that tutorial and ended up getting timeout error.

After the env got deployed, orbit build a GPU image once. Then, make some change to the image and rebuild it. The image might have to be large enough to...

This feature has been added in the mainline and will be available after 0.8 release. An example usage would be `predictor = TabularPredictor(label=label).fit(train_data, included_model_types=["KNN"])`

This issue seems only related to ray lightning backend, which is being deprecated by the Ray team. We will likely switch to the new backend and revisit this problem later.

Mark as resolved as `text` module is being deprecated and just tested AG multimodal 0.6.2 on colab. This issue doesn't exist anymore