Difference between plot and summary.txt values of loss and accuracy
I have performed a training with OrcaNet in order to set up a GNN for track and shower classification.
When cross-validating the values of training loss (train_loss
) and training accuracy (train_acc
) from the summary table below with the values in the summary plots, it seems they are quite different.
For example, in the summary table, train_loss
is below 0.4 several times (in epoch 3 and 5), while never falling below 0.4 in the plot. What could cause this discrepancy between table and plot values? Or does this normally happen? val_loss
and val_acc
do seem to match the plotted values.