Main Stream

1. Data set 구성
Time series(seq len, time lag, pred len)
+ noies
Train val test split

2. Dataloader parameter
shuffle True
batch size
drop last => dynamic model in batch (not fix batch size)
batch first

4. model search
model architecture (RNN, LSTM, GRU, etc.)
hidden layers & nodes
classification
regressor
eval & train

5. pytorch model
model save & best loss model

6. drop out / batch normalization

7. learning rate & epochs

8. loss functions
MSE, RMSE
cross Entropy

9. Gpu & Env setting

10. scaling