I finally managed to figure out the issue with the second flux. Everything was working during training and validation but during testing, the regressor was comparing the first flux value to the second fluxes predicted value which of course made the test loss very large.
I finally managed to figure out the issue with the second flux. Everything was working during training and validation but during testing, the regressor was comparing the first flux value to the second fluxes predicted value which of course made the test loss very large.