FrankWork / fudan_mtl_reviews

TensorFlow implementation of the paper `Adversarial Multi-task Learning for Text Classification`
178 stars 40 forks source link

diff_loss中的转置是否存在问题? #17

Closed alkaideemo closed 5 years ago

alkaideemo commented 5 years ago

https://github.com/FrankWork/fudan_mtl_reviews/blob/3cb081d07167a2d97bfab2cf950c52acae3b0c2e/src/models/mtl_model.py#L71-L72 task_feat的size是(bsz, feature_size), shared_feat的size也是(bsz, feature_size), 这样乘出来的结果岂不是(feature, bsz) x (bsz, feature) --> (feature, feature)。 但是我个人感觉这个size应该是(bsz, bsz)更合理些?

是我搞错了, task_feat & shared_feat 的feature_size不一定相同。