ETSformer hyper-parameter mismatch during NNI tuning

[WenjieDu]

1. System Info

PyPOTS v0.5

2. Information

• The official example scripts
• My own created scripts

3. Reproduction

Tune ETSformer hyperparameters in PyPOTS with NNI.

4. Expected behavior

2024-05-28 15:18:24 [INFO]: Have set the random seed as 2022 for numpy and pytorch.
2024-05-28 15:18:24 [INFO]: The tunner assigns a new group of params: {'n_steps': 48, 'n_features': 35, 'epochs': 100, 'patience': 10, 'n_e_layers': 2, 'n_d_layers': 3, 'd_model': 128, 'd_ffn': 1024, 'n_heads': 8, 'top_k': 3, 'dropout': 0.4, 'lr': 0.008927675977610391}
Traceback (most recent call last):
  File "/home/linglong/data/linglong/.conda/envs/pypots/bin/pypots-cli", line 8, in <module>
    sys.exit(main())
  File "/home/linglong/data/linglong/.conda/envs/pypots/lib/python3.10/site-packages/pypots/cli/pypots_cli.py", line 37, in main
    service.run()
  File "/home/linglong/data/linglong/.conda/envs/pypots/lib/python3.10/site-packages/pypots/cli/tuning.py", line 249, in run
    raise RuntimeError(
RuntimeError: Hyperparameters do not match. Mismatched hyperparameters (in the tuning configuration but not in the given model's arguments): ['top_k', 'd_ffn', 'n_heads', 'n_d_layers', 'n_e_layers', 'dropout']