Thanks for your Code!
When I run [runTrain_CUB_INC.sh] which you provided, I found loss_d has 'Nan', because the 'count' of TSA.samples_split is 0!!!
I want to know "sample_num_source=32, sample_num_target=32" in TSA is big? and, how to avoid 'count' is 0.
And if I let loss_d =0 when 'count' is 0, unfortunately, it does much influence the evaluation scores.
Thanks for your Code! When I run [runTrain_CUB_INC.sh] which you provided, I found loss_d has 'Nan', because the 'count' of TSA.samples_split is 0!!! I want to know "sample_num_source=32, sample_num_target=32" in TSA is big? and, how to avoid 'count' is 0. And if I let loss_d =0 when 'count' is 0, unfortunately, it does much influence the evaluation scores.