Caner Turkmen

Thanks for raising this issue @Alexwenner0 , a fix is already being worked on at #1993. This error is due to many finance time series having irregular timestamps, when the...

@Alexwenner0 We recently merged #1993 . You can now initialize a `TimeSeriesPredictor(ignore_time_index=True)` to make sure time indices are ignored during training and testing. I will close this for now, but...

👍 I would think that depends entirely on the use case. For example, in finance, I believe the irregular times would correspond to days when the market is open and...

I was able to reproduce this issue before but cannot anymore, with `torch==1.12.1` and `mxnet==1.9.0`. Could you perhaps check again @yinweisu ? @sxjscience for timeseries we hope to make it...

Hi @abdulfatir, could you please make sure the style/type checks are OK for this new directory? @lostella, this is part of Fatir's NeurIPS paper. Can we go ahead and merge...

Hi @abdulfatir, in line with Jasper's comments, could you move the tests to `test/nursery` and make sure they run in the test runner?

OK fair enough. @jaheba, will merge by EOD today (camera-ready deadline) if no further questions from your end.

Hey @abdulfatir , It looks like the doc tests are still failing. could you take a look, potentially with removing the tests altogether?

Hi @SamuelDietz thank you for your question. AutoGluon-TimeSeries went out with v0.5, and multi-window backtesting capability was added with v0.6. see https://auto.gluon.ai/stable/tutorials/timeseries/forecasting-indepth.html#multi-window-backtesting We are looking forward to adding past-only covariates...

Hi @SamuelDietz, Past only covariate support was also added with #2680 . Please let us know if this helps and your experience with it. Closing this for now.