train 1 and 1a models

[jsadler2]

we've trained the 0_baseline_LSTM and 2_multitask_dense models with the updated input drivers (#177 , #62) but we haven't trained the 1_metab_multitask or the 1a_multitask_do_gpp_er models. Once those are trained, we can add their metrics to the combined metrics files (#185)