Leo (Hsiu-Wei) Yang

icon indicating an email [email protected]

icon company Thomson Reuters icon location Waterloo, ON icon location Data Scientist

Results 1 comments of Leo (Hsiu-Wei) Yang

Thanks for your code ,I have a question that how to reduce the 768 embedding to 300，and maintain semantic similarity，I don't find it in your code

Hi, sorry for late reply. Basically, there's one more FC layer to reduce the dimensionality: https://github.com/Impavidity/relogic/blob/d8197cdfcf035c156fd0f91c23195cdef6fbea03/relogic/logickit/modules/representation_module.py#L10