In your implementaion , like tt.py: line 61 out = tf.reshape(out, [mat_ranks[i] * inp_modes[i], -1]).
It seems the different Infos from different batch_size dimensions will be mixed with others. Could it be OK?
Since usually different infos from different batch_size dimensions will be processed seperately.
In your implementaion , like tt.py: line 61 out = tf.reshape(out, [mat_ranks[i] * inp_modes[i], -1]).
It seems the different Infos from different batch_size dimensions will be mixed with others. Could it be OK?
Since usually different infos from different batch_size dimensions will be processed seperately.