Weisu Yin

Using camera is not allowed with http connection. Therefore, demo page won't work on the staging page

Did some dig into this issue: The fundamental issue is that a fresh colab gpu env has `torch==1.11` and `torchtext==0.12.0`. The installation of `autogluon` only updates `torch` to 1.10.2 while...

I just verified local env doesn't have this issue because there's no torchtext pre-installed. Torchtext is the one that's causing problem. It fails even when I do not do `pip3...

Yes. I had a look at their dependency too. Torchtext won't be installed because we specified `pytorch_lightning` as our dependency instead of `pytorch_lightning[extra]` However, even we put [extra](we should not...

Instead of downgrading `protobuf`. Make sure to update the python toolchains: ```python pip3 install -U pip pip3 install -U setuptools wheel ``` then install autogluon. Tested on both a cpu...

Maybe we can do it to 0.5 release since it's close and this issue seems to happen only on certain environment.

Hi, I wasn't able to reproduce this issue with both autogluon==0.3.1 and current master branch. Is it possible for you to share the dataset so that I can try to...

OK. I tested multi modality with text predictor being disabled and rmse worked fine. This should be improved for the next release. The usage of both mxnet(text) and torch(image) backend...

+1

There are duplicate code in terms of extracting auxiliary paths because I find it confusing to put the logic of extracting auxiliary and regular train/test paths in a single function....