Questions about hyperparameters

[hana842]

Hello, When I followed the method published on GitHub, I obtained the following results.
When auxi_lambda is 0.8 rather than 0, I expected MAE/MSE to improve, but the performance actually got worse. If my hyperparameters differ from the paper, could you share the ones reported there?

result------
long_term_forecast_ETTh1_96_192_iTransformer_ETTh1_ftM_sl96_ll48_pl192_dm128_nh8_el2_dl1_df128_fc3_ebtimeF_dtTrue_ax0.0_rl1.0_axlMAE_mf1_LTFAll_0
mse:0.4458734393119812, mae:0.4395802915096283

long_term_forecast_ETTh1_96_336_iTransformer_ETTh1_ftM_sl96_ll48_pl336_dm128_nh8_el2_dl1_df128_fc3_ebtimeF_dtTrue_ax0.0_rl1.0_axlMAE_mf1_LTFAll_0
mse:0.49356234073638916, mae:0.46669265627861023

long_term_forecast_ETTh1_96_720_iTransformer_ETTh1_ftM_sl96_ll48_pl720_dm128_nh8_el2_dl1_df128_fc3_ebtimeF_dtTrue_ax0.0_rl1.0_axlMAE_mf1_LTFAll_0
mse:0.5168138742446899, mae:0.500756561756134

long_term_forecast_ETTh1_96_192_iTransformer_ETTh1_ftM_sl96_ll48_pl192_dm128_nh8_el2_dl1_df128_fc3_ebtimeF_dtTrue_ax0.8_rl0.2_axlMAE_mf1_LTFAll_0
mse:0.47631266713142395, mae:0.458281010389328

long_term_forecast_ETTh1_96_336_iTransformer_ETTh1_ftM_sl96_ll48_pl336_dm128_nh8_el2_dl1_df128_fc3_ebtimeF_dtTrue_ax0.8_rl0.2_axlMAE_mf1_LTFAll_0
mse:0.5614497065544128, mae:0.504854679107666

long_term_forecast_ETTh1_96_720_iTransformer_ETTh1_ftM_sl96_ll48_pl720_dm128_nh8_el2_dl1_df128_fc3_ebtimeF_dtTrue_ax0.8_rl0.2_axlMAE_mf1_LTFAll_0
mse:0.5730749368667603, mae:0.5308745503425598

[Master-PLC]

Thank you for your interest in our work and for reporting this issue. Regarding the hyperparameters, we have provided the exact hyperparameter settings used in our experiments in the script files located at scripts/ltf_overall/{dataset}/iTransformer.sh. You can refer to these scripts as a reference for reproducing our results. It's worth noting that experimental results may vary slightly due to differences in platform environments, hardware configurations and software versions.