Performance validation of EasyTPP models

We run the experiments on various dataset to validate the implementations: each model is trained with a max number of epochs and the best model is selected based on the performance on the valid set, then we report the results on the test set.

Simulated dataset

Conttime

Models	Loglike	RMSE	Acc	Num Training Epochs
Torch_NHP	-0.93504	0.34000	0.38656	200
Tf_NHP	-0.85774	0.34014	0.38806	200
Torch_AttNHP	-1.02001	0.33678	0.36782	200
Tf_AttNHP	-1.02315	0.33816	0.19456	200
Torch_AttNHP	-1.00593	0.33685	0.37723	500
Tf_AttNHP	-0.99827	0.33717	0.36498	500
Torch_THP	-0.99827	0.33717	0.36498	500
Tf_THP	-1.01898	0.33677	0.37875	500

## Real dataset ### Taxi