Helomin

Both use the sevir dataset to predict for two hours Using the adamw optimizer with a learning rate of 1e-4, none of them used the task learning rate optimization strategy...