nkqiaolin comments

Repositories
Issues
Comments

Results 2 comments of


                                            nkqiaolin

In DeepFm.py line 102, why the second order term's size is (None, K) rather than (None, )?

IMO, the original paper said each of three components should all be one scalar, then sum together to fed to sigmod() . But you can also treat the sum of...

Why not use updated parent beam's embedding as next input when inferring?

I checked the fairseq-py, which will re-order the input for every beam search step: def reorder_buffer(self, new_order): if self.input_buffer is not None: self.input_buffer = self.input_buffer.index_select(0, new_order)