68f970aca1f95cddbc1c5fc4e4c7e6b6bffb2293,models/AttModel.py,AttModel,_forward,#AttModel#Any#Any#Any#Any#,98

Before Change

        state = self.init_hidden(batch_size)

        // outputs = []
        outputs = Variable(fc_feats.data.new(batch_size, seq.size(1) - 1, self.vocab_size+1).zero_())

        fc_feats, att_feats, p_att_feats = self._prepare_feature(fc_feats, att_feats, att_masks)

        for i in range(seq.size(1) - 1):
            if self.training and i >= 1 and self.ss_prob > 0.0: // otherwiste no need to sample
                sample_prob = fc_feats.data.new(batch_size).uniform_(0, 1)
                sample_mask = sample_prob < self.ss_prob
                if sample_mask.sum() == 0:
                    it = seq[:, i].clone()
                else:
                    sample_ind = sample_mask.nonzero().view(-1)
                    it = seq[:, i].data.clone()
                    //prob_prev = torch.exp(outputs[-1].data.index_select(0, sample_ind)) // fetch prev distribution: shape Nx(M+1)
                    //it.index_copy_(0, sample_ind, torch.multinomial(prob_prev, 1).view(-1))
                    // prob_prev = torch.exp(outputs[-1].data) // fetch prev distribution: shape Nx(M+1)
                    prob_prev = torch.exp(outputs[:, i-1].data) // fetch prev distribution: shape Nx(M+1)
                    it.index_copy_(0, sample_ind, torch.multinomial(prob_prev, 1).view(-1).index_select(0, sample_ind))
                    it = Variable(it, requires_grad=False)
            else:

After Change

        state = self.init_hidden(batch_size)

        // outputs = []
        outputs = fc_feats.new_zeros(batch_size, seq.size(1) - 1, self.vocab_size+1)

        fc_feats, att_feats, p_att_feats = self._prepare_feature(fc_feats, att_feats, att_masks)

        for i in range(seq.size(1) - 1):
            if self.training and i >= 1 and self.ss_prob > 0.0: // otherwiste no need to sample
                sample_prob = fc_feats.new(batch_size).uniform_(0, 1)
                sample_mask = sample_prob < self.ss_prob
                if sample_mask.sum() == 0:
                    it = seq[:, i].clone()
                else:
                    sample_ind = sample_mask.nonzero().view(-1)
                    it = seq[:, i].data.clone()
                    //prob_prev = torch.exp(outputs[-1].data.index_select(0, sample_ind)) // fetch prev distribution: shape Nx(M+1)
                    //it.index_copy_(0, sample_ind, torch.multinomial(prob_prev, 1).view(-1))
                    // prob_prev = torch.exp(outputs[-1].data) // fetch prev distribution: shape Nx(M+1)
                    prob_prev = torch.exp(outputs[:, i-1].detach()) // fetch prev distribution: shape Nx(M+1)
                    it.index_copy_(0, sample_ind, torch.multinomial(prob_prev, 1).view(-1).index_select(0, sample_ind))
            else:
                it = seq[:, i].clone()          

Instances