An offical implementation of PatchTST: "A Time Series is Worth 64 Words: Long-term Forecasting with Transformers." (ICLR 2023) https://arxiv.org/abs/2211.14730
Apache License 2.0
1.37k
stars
248
forks
source link
Bug about the parameter "feature" in self-supervised PatchTST #77
In self-supervised PatchTST, if i set the parameter as "MS", which means i want to predict single variable. But in the _all_batch_train function, the losses are calculated from multivariate's output. On the contrary, This is correct in supervised version.
In self-supervised PatchTST, if i set the parameter as "MS", which means i want to predict single variable. But in the _all_batch_train function, the losses are calculated from multivariate's output. On the contrary, This is correct in supervised version.