DeepSeek-V3.2-Exp · Expert scores

61 transformer blocks › 58 MoE blocks › Mixture of experts › Router

Expert scores

learnedlinear(x.float(), self.weight.float())

One learned score vector per expert: 256 scores per position.

{\htmlData{light=logits scores}{z}}={\htmlData{light=input}{x_{\mathrm{ffn}}}}W_{r}

in: $x_{\mathrm{ffn}}$ [B batch, T positions, 7168 d]
out: $z$ [B batch, T positions, 256 experts]
parameters: 1,835,008 · × 58 MoE blocks = 106,430,464

Held in bfloat16, the default type (generate.py 109), and applied in float32; the other linear weights in the block are FP8 (model.py 674, 687, 876).

Parameters

1,835,008 parameters per instance.

Exact per-instance accounting
Component	Calculation	Parameters
Weight matrix	$7{,}168 \times 256$	1,835,008
Per-instance total		1,835,008

This projection has no bias parameter.

The reference module holds the weight output by input, [out, in].

Across 58 independent MoE blocks: $1{,}835{,}008 \times 58 = 106{,}430{,}464$ parameters.

Derived from the configured matrix dimensions; checkpoint tensors have not been independently counted.

Codemodel.py 668–675, 687–709

model.py 668–675

        self.dim = args.dim
        self.topk = args.n_activated_experts
        self.n_groups = args.n_expert_groups
        self.topk_groups = args.n_limited_groups
        self.score_func = args.score_func
        self.route_scale = args.route_scale
        self.weight = nn.Parameter(torch.empty(args.n_routed_experts, args.dim))
        self.bias = nn.Parameter(torch.empty(args.n_routed_experts, dtype=torch.float32)) if self.dim == 7168 else None

model.py 687–709

        scores = linear(x.float(), self.weight.float())
        if self.score_func == "softmax":
            scores = scores.softmax(dim=-1)
        else:
            scores = scores.sigmoid()
        original_scores = scores
        if self.bias is not None:
            scores = scores + self.bias
        if self.n_groups > 1:
            scores = scores.view(x.size(0), self.n_groups, -1)
            if self.bias is None:
                group_scores = scores.amax(dim=-1)
            else:
                group_scores = scores.topk(2, dim=-1)[0].sum(dim=-1)
            indices = group_scores.topk(self.topk_groups, dim=-1)[1]
            mask = scores.new_ones(x.size(0), self.n_groups, dtype=bool).scatter_(1, indices, False)
            scores = scores.masked_fill_(mask.unsqueeze(-1), float("-inf")).flatten(1)
        indices = scores.topk(self.topk, dim=-1)[1]
        weights = original_scores.gather(1, indices)
        if self.score_func == "sigmoid":
            weights /= weights.sum(dim=-1, keepdim=True)
        weights *= self.route_scale
        return weights, indices

Sources2

Direct evidence

The router owns one score vector per routed expert and, at this width, a per-expert bias used only for selection.

gate-init · deepseek-ai/DeepSeek-V3.2-Exp / inference/model.py
Pinned revision 87e509a2e5a1 ↗ · lines 668–675

Read code excerpt

        self.dim = args.dim
        self.topk = args.n_activated_experts
        self.n_groups = args.n_expert_groups
        self.topk_groups = args.n_limited_groups
        self.score_func = args.score_func
        self.route_scale = args.route_scale
        self.weight = nn.Parameter(torch.empty(args.n_routed_experts, args.dim))
        self.bias = nn.Parameter(torch.empty(args.n_routed_experts, dtype=torch.float32)) if self.dim == 7168 else None

Direct evidence

Routing: sigmoid scores, bias added for selection, the best groups kept by their top-two scores, the top-8 experts chosen, their unbiased scores renormalized and scaled.

gate-forward · deepseek-ai/DeepSeek-V3.2-Exp / inference/model.py
Pinned revision 87e509a2e5a1 ↗ · lines 687–709

Read code excerpt

        scores = linear(x.float(), self.weight.float())
        if self.score_func == "softmax":
            scores = scores.softmax(dim=-1)
        else:
            scores = scores.sigmoid()
        original_scores = scores
        if self.bias is not None:
            scores = scores + self.bias
        if self.n_groups > 1:
            scores = scores.view(x.size(0), self.n_groups, -1)
            if self.bias is None:
                group_scores = scores.amax(dim=-1)
            else:
                group_scores = scores.topk(2, dim=-1)[0].sum(dim=-1)
            indices = group_scores.topk(self.topk_groups, dim=-1)[1]
            mask = scores.new_ones(x.size(0), self.n_groups, dtype=bool).scatter_(1, indices, False)
            scores = scores.masked_fill_(mask.unsqueeze(-1), float("-inf")).flatten(1)
        indices = scores.topk(self.topk, dim=-1)[1]
        weights = original_scores.gather(1, indices)
        if self.score_func == "sigmoid":
            weights /= weights.sum(dim=-1, keepdim=True)
        weights *= self.route_scale
        return weights, indices