search

salesforce / CoST

PyTorch code for CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting (ICLR 2022)
BSD 3-Clause "New" or "Revised" License
214 stars 43 forks source link

--eval problem #1

Closed zhangkexin1126 closed 2 years ago

zhangkexin1126 commented 2 years ago

Thank for sharing the code. I found that the value of parameter "padding" in forecasting.py (line 24 or the first line in function eval_forecasting()) should be equal to args.max-train-length, if not, the evaluation process is not available.

gorold commented 2 years ago

Thanks for catching that, it was hardcoded into the script since all experiments in the paper used that same setting.

msonah commented 2 years ago

Thank for sharing the code. I found that the value of parameter "padding" in forecasting.py (line 24 or the first line in function eval_forecasting()) should be equal to args.max-train-length, if not, the evaluation process is not available.

The author set the parameter 'kernels' (Kernel sizes for mixture of AR experts module) in the code to 'None', when I run the code, it will report an error (Rearrange can't be applied to an empty list), please ask' How should kernels' be set up, may you give me some advice?

gorold commented 2 years ago

Fixed issue: 3c4e765