Model size and performance

[kdoh0914]

Hi. Can you share your

• model size(number of parameters),
• training epoch or steps when you wrote the paper
• and inference speed when you forecast 48 steps ahead as you mentioned in your paper?

Thanks a lot!

[MountVoom]

Hi. Thanks for your concern. You can calculate model size by this code:

for _, p in model.named_parameters():
        paras += p.numel()

As for train epoch, you can refer to main_informer.py. Inference speed is different in different operating environments, you can do a simple test using time.time().