Can not reproduce regular time series datasets results

[RaptorMai]

Thanks for releasing the code. I tried to reproduce the regular time series dataset results but I found that I could not get similar number as the reported ones in the paper. I wonder if you could share the script to reproduce the regular data results.

For example, for EthanolConcentration, this is the script I ran:

python3 run_ImgCLS.py \ --model swin\ --seed 1799 \ --save_total_limit 1 \ --dataset EthanolConcentration \ --dataset_prefix differ_-*1_**1_2*2_256*256_images \ --train_batch_size 48 \ --eval_batch_size 128 \ --logging_steps 20 \ --save_steps 20 \ --epochs 20 \ --learning_rate 1e-4 \ --n_runs 1 \ --n_splits 5 \ --do_train

I only got Accuracy = 33.0 +/- 0.8, which is quite different from the reported 45.6.