Skip to content

A PaddlePaddle implementation of CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting.

Notifications You must be signed in to change notification settings

aptx1231/CoST_Paddle

Repository files navigation

CoST_Paddle

【论文复现赛】CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting

队伍名:aptx1231

方法介绍

见文件CoST

数据集

参考原作者的仓库,下载数据后放在dataset/目录即可。

环境配置

主要依赖为PaddlePaddle=2.3.0,Python=3.7.13。

依赖环境已放在requirements.txt中。

训练与测试

多变量

# ETTh1
python -u train.py ETTh1 forecast_multivar --alpha 0.0005 --kernels 1 2 4 8 16 32 64 128 --max-train-length 201 --batch-size 128 --archive forecast_csv --repr-dims 320 --max-threads 8 --seed 0 --gpu 0 --eval > etth1_m.log
# ETTh2
python -u train.py ETTh2 forecast_multivar --alpha 0.0005 --kernels 1 2 4 8 16 32 64 128 --max-train-length 201 --batch-size 128 --archive forecast_csv --repr-dims 320 --max-threads 8 --seed 0 --gpu 1 --eval > etth2_m.log
# ETTm1
python -u train.py ETTm1 forecast_multivar --alpha 0.0005 --kernels 1 2 4 8 16 32 64 128 --max-train-length 201 --batch-size 128 --archive forecast_csv --repr-dims 320 --max-threads 8 --seed 0 --gpu 3 --eval > ettm1_m.log

单变量

# ETTh1
python -u train.py ETTh1 forecast_univar --alpha 0.0005 --kernels 1 2 4 8 16 32 64 128 --max-train-length 201 --batch-size 128 --archive forecast_csv_univar --repr-dims 320 --max-threads 8 --seed 0 --gpu 0 --eval > etth1_s.log
# ETTh2
python -u train.py ETTh2 forecast_univar --alpha 0.0005 --kernels 1 2 4 8 16 32 64 128 --max-train-length 201 --batch-size 128 --archive forecast_csv_univar --repr-dims 320 --max-threads 8 --seed 0 --gpu 1 --eval > etth2_s.log
# ETTm1
python -u train.py ETTm1 forecast_univar --alpha 0.0005 --kernels 1 2 4 8 16 32 64 128 --max-train-length 201 --batch-size 128 --archive forecast_csv_univar --repr-dims 320 --max-threads 8 --seed 0 --gpu 3 --eval > ettm1_s.log

参考实现

本代码主要参考以下两个仓库:

About

A PaddlePaddle implementation of CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages