The official implementation of STD-PLM.

Train and Test the STD-PLM

cd ./code/STD-PLM/src

#prediction
bash '../scripts/pems03.sh'

#imputation
bash '../scripts/pems08_sctc30.sh'

#zero-shot
bash '../scripts/pems03<-PEMS07.sh'

#few-shot
bash '../scripts/pems04_few.sh'

Arguments

--lora : Fine-tune the PLM with lora
--ln_grad : Train the layernorm of PLM
--wo_conloss  : Removing the constraint loss function
--sandglassAttn : Introducing the SGA module.
--time_token : Add time token
--model plm : Specify the PLM to use
--llm_layers layers : Specify the layers of PLM
--few_shot ratio    :   Specify the ratio of few-shot
--zero_shot :   zero-shot
--from_pretrained_model model.pth   : Specify the trained model weights
--task task : Specify the task