What Cv strategy do to use ?

Just a baseline fastai image regression model with

fit_one_cycle(), EarlyStoppingCallback etc with only 2kfold to test run for now.

image regression

hey, is it mse loss?, because I get 130 ~ 140 mse loss. Finally, when you report LB score is the average predictions of your 5 folds or retraining model using all data?.

MSE loss is not the metric for this comp, the metric for this comp is MAE loss. Yes i submit preds of 5 fold

which loss function your using?