deepseek_vl2/models/modeling_deepseek.py [448:462]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            )  # [n, n_group]
            group_idx = torch.topk(
                group_scores, k=self.topk_group, dim=-1, sorted=False
            )[
                1
            ]  # [n, top_k_group]
            group_mask = torch.zeros_like(group_scores)  # [n, n_group]
            group_mask.scatter_(1, group_idx, 1)  # [n, n_group]
            score_mask = (
                group_mask.unsqueeze(-1)
                .expand(
                    bsz * seq_len, self.n_group, self.n_routed_experts // self.n_group
                )
                .reshape(bsz * seq_len, -1)
            )  # [n, e]
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



deepseek_vl2/models/modeling_deepseek.py [472:486]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            )  # [n, n_group]
            group_idx = torch.topk(
                group_scores, k=self.topk_group, dim=-1, sorted=False
            )[
                1
            ]  # [n, top_k_group]
            group_mask = torch.zeros_like(group_scores)  # [n, n_group]
            group_mask.scatter_(1, group_idx, 1)  # [n, n_group]
            score_mask = (
                group_mask.unsqueeze(-1)
                .expand(
                    bsz * seq_len, self.n_group, self.n_routed_experts // self.n_group
                )
                .reshape(bsz * seq_len, -1)
            )  # [n, e]
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -