Proposal: Adding SegRNN Code to the Library

[lss-1138]

Hi, thank you for your efforts in contributing to the advancement of the community, and I have been following your work for quite some time. I noticed that your repository does not include any baseline methods based on RNNs, which is understandable given that traditional RNNs struggle to model very long sequences.

To address this, we have introduced SegRNN (repository link: https://github.com/lss-1138/SegRNN), which significantly improves the performance of RNN methods in long time series prediction tasks and currently stands as the state-of-the-art RNN-based method. It demonstrates outstanding performance regardless of the length of the look-back length, whether it be longer (720) or shorter (96).

The performance of SegRNN on the Weather dataset with a look-back length of 720：

The performance of SegRNN on the Weather dataset with a look-back length of 96：

We would greatly appreciate it if you could consider incorporating SegRNN into your repository. Once again, thank you for your efforts.

[wuhaixu2016]

Hi, many thanks for your recommendation.
Due to the time limitation, we recently mainly focus on fixing bug. Would you please kindly propose a pull request?

[lss-1138]

Ok. We have proposed a pull request. Thank you for your efforts. @wuhaixu2016