search

HKUDS / FlashST

[ICML'2024] "FlashST: A Simple and Universal Prompt-Tuning Framework for Traffic Prediction"
https://arxiv.org/abs/2405.17898
Apache License 2.0
52 stars 2 forks source link

Reproduction of results on Table4 #3

Open RWLinno opened 1 month ago

RWLinno commented 1 month ago

How the w/o fine-tuning results in Table 4 are derived? I tried to use -mode ori to generate a pre-trained model of stgcn on PEMS07M, but the pretrain mode in the code only yields results after enhancement with FlashST.

LZH-YS1998 commented 1 month ago

Hello. You can use the same pre-training data as FlashST to pre-train on different baselines, and then proceed to testing. If you use the "ori" mode, you can assess the original performance of the baseline models, rather than the performance enhanced by FlashST.