Greg DeVos

Official repo https://github.com/ServiceNow/tactis

I think it would look something like this ```python nll = -out_distr.log_prob(target).mean() if self.beta > 0.0: variance = out_distr.variance nll = nll * (variance.detach() ** self.beta) return nll ```

An open source implementation can be found here: https://github.com/AdityaLab/PROFHiT

Running without a GPU will be slow. To know if it is working, I would reduce your batch size from 512 to 2, which will reduce the time to report...

Im currently look for an apartment, I will address these comments when I have a bit more free time

@hrzn I originally wanted to add it because it looked to do reasonably well on [benchmarks](https://paperswithcode.com/paper/long-term-series-forecasting-with-query). I have run a bunch of tests and haven't been impressed. I mostly get...