multi-covariate finetune problem of ETTh1 dataset

[handongke]

Does the default fine-tuning example use multiple variables? How can I set it up during fine-tuning to only use the specified multiple variable data columns, such as HUFL, HULL, MUFL, MULL, to predict the target column OT?

ETTh1 dataset

date,HUFL,HULL,MUFL,MULL,LUFL,LULL,OT
2016-07-01 00:00:00,5.827000141143799,2.009000062942505,1.5989999771118164,0.4620000123977661,4.203000068664552,1.3400000333786009,30.5310001373291
2016-07-01 01:00:00,5.692999839782715,2.075999975204468,1.4919999837875366,0.4259999990463257,4.142000198364259,1.371000051498413,27.78700065612793
2016-07-01 02:00:00,5.1570000648498535,1.741000056266785,1.2790000438690186,0.35499998927116394,3.776999950408936,1.218000054359436,27.78700065612793
2016-07-01 03:00:00,5.0900001525878915,1.9420000314712524,1.2790000438690186,0.3910000026226044,3.806999921798706,1.2790000438690186,25.04400062561035
...

[LSY-Cython]

Dear authors, thanks for your work! I am also interested in this question, since I found that the publicized finetuning code can support mode "S" (i.e. univariate input and univariate output) and mode "M" (i.e. multivariate input and multivariate output), without the support of such mode: multivariate input and univariate output. So how can I achieve this mode? Looking forward to your reply! Thanks a lot!