cde27bc0dad15b0bef4b0568f6ac1778920447e5,models/TransformerModel.py,TransformerModel,_prepare_feature,#TransformerModel#Any#Any#Any#,339

Before Change

        att_feats = pack_wrapper(self.att_embed, att_feats, att_masks)

        if att_masks is None:
            att_masks = att_feats.new_ones(att_feats.shape[:2], dtype=torch.long)
        att_masks = att_masks.unsqueeze(-2)

        if seq is not None:
            // crop the last one
            seq = seq[:,:-1]
            seq_mask = (seq.data > 0)
            seq_mask[:,0] += 1

            seq_mask = seq_mask.unsqueeze(-2)
            seq_mask = seq_mask & subsequent_mask(seq.size(-1)).to(seq_mask)
        else:
            seq_mask = None

        return att_feats, seq, att_masks, seq_mask

    def _forward(self, fc_feats, att_feats, seq, att_masks=None):
        att_feats, seq, att_masks, seq_mask = self._prepare_feature(att_feats, att_masks, seq)

After Change

        att_feats, seq, att_masks, seq_mask = self._prepare_feature_forward(att_feats, att_masks)
        memory = self.model.encode(att_feats, att_masks)

        return fc_feats[...,:1], att_feats[...,:1], memory, att_masks

    def _prepare_feature_forward(self, att_feats, att_masks=None, seq=None):
        att_feats, att_masks = self.clip_att(att_feats, att_masks)

Instances