Xingjian Shi

@jturner116 are you loading a saved TabularPredictor? We had other users that have trained the model on GPU instances and load the model to run inference on CPU.

@Innixma @jturner116 I mean, if the instance has only CPUs, we will be using CPU inference by default and there is nothing else you need to do. This is the...

Sorry for the late reply as I just see this issue. The "price" column has been marked as "ignored" in https://github.com/sxjscience/automl_multimodal_benchmark/blob/77edc53ef7e8a91ea4cfdaa336b353af66e3db00/multimodal_text_benchmark/src/auto_mm_bench/datasets.py#L480-L485 .

Yeah, thanks for pointing out this bug. I haven't tried using multiple GPUs for the MovingMNIST experiments.

Seems to be related to the latest MXNet

@pflashgary Thanks for the question, I haven't run the source code for a while and the bug seems to be related to MXNet. Which version of MXNet are you currently...

@pflashgary and @sulisetyowidodo I currently do not have bandwidth to check which version of MXNet works for the latest CUDA. We have switched the development to PyTorch and you may...

This should be related to ImagePredictor hasn't switched default backend to MXNet.

I think it is partially resolved via the PR https://github.com/autogluon/autogluon/pull/2859

Resolved since we now have tensorboard instructions in .fit()