Skip to content

LLM inference in C/C++

C++ ggml

model: add NextN/MTP speculative decoding support for GLM_DSA (GLM-5.2) (#25980)

Satinder Grewal7be2c65dc9ad
diff --git a/conversion/glm.py b/conversion/glm.py
index d85268a621..cc34cddbf8 100644
--- a/conversion/glm.py
+++ b/conversion/glm.py
@@ -1,6 +1,8 @@
from __future__ import annotations
-from typing import Iterable, TYPE_CHECKING
+import re
+
+from typing import Callable, Iterable, TYPE_CHECKING
import torch
@@ -213,12 +215,47 @@ class Glm4MoeLiteModel(DeepseekV2Model):
class GlmMoeDsaModel(DeepseekV2Model):
model_arch = gguf.MODEL_ARCH.GLM_DSA
skip_mtp = False
+ supports_mtp_export = True
+
+ # Trunk layer count, stashed before indexing so the classmethod
+ # filter_tensors can identify the appended NextN/MTP block (mirrors
+ # HYV3Model / Step35Model).
+ _n_main_layers: int | None = None
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
- self.block_count = self.hparams["num_hidden_layers"] + self.hparams.get("num_nextn_predict_layers", 0)
+ self.block_count = self.hparams["num_hidden_layers"]
+ if not self.no_mtp:
+ self.block_count += self.hparams.get("num_nextn_predict_layers", 0)
self.tensor_map = gguf.get_tensor_name_map(self.model_arch, self.block_count)
+ def index_tensors(self, remote_hf_model_id: str | None = None):
+ type(self)._n_main_layers = self.hparams["num_hidden_layers"]
+ return super().index_tensors(remote_hf_model_id=remote_hf_model_id)
+
+ @classmethod
+ def filter_tensors(cls, item: tuple[str, Callable[[], Tensor]]) -> tuple[str, Callable[[], Tensor]] | None:
+ if (titem := super().filter_tensors(item)) is None:
+ return None
+ name, gen = titem
+
+ # GLM-5.2 appends the NextN/MTP block past num_hidden_layers
+ # (model.layers.78 -> blk.78 in the 79-block file).
+ assert cls._n_main_layers is not None
+ is_mtp = (m := re.match(r"model\.layers\.(\d+)\.", name)) is not None and int(m.group(1)) >= cls._n_main_layers
+
+ # --no-mtp: drop the appended NextN block entirely.
+ if is_mtp and cls.no_mtp:
+ return None
+ # --mtp: keep ONLY NextN-block tensors plus the shared embeddings/
+ # norm/lm_head (so the resulting GGUF carries just the draft head).
+ if cls.mtp_only and not is_mtp and name not in (
+ "model.embed_tokens.weight", "model.norm.weight", "lm_head.weight",
+ ):
+ return None
+
+ return name, gen
+
def set_vocab(self):
return self._set_vocab_glm()
@@ -230,7 +267,7 @@ class GlmMoeDsaModel(DeepseekV2Model):
self.gguf_writer.add_rope_dimension_count(int(rope_dim * partial_rotary_factor))
# NextN/MTP prediction layers
- if (num_nextn_predict_layers := self.hparams.get("num_nextn_predict_layers")) is not None:
+ if not self.no_mtp and (num_nextn_predict_layers := self.hparams.get("num_nextn_predict_layers")) is not None:
self.gguf_writer.add_nextn_predict_layers(num_nextn_predict_layers)
# DSA indexer parameters
diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 7a70585fa4..a8422ff605 100644
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -2072,7 +2072,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
res = nullptr;
} break;
case LLM_ARCH_DEEPSEEK32:
- case LLM_ARCH_GLM_DSA:
{
res = new llama_kv_cache_dsa(
*this,
@@ -2089,6 +2088,56 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
nullptr,
nullptr);
} break;
+ case LLM_ARCH_GLM_DSA:
+ {
+ if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP && hparams.n_layer_nextn > 0) {
+ // The NextN/MTP draft head runs dense MLA (no DSA indexer), so the
+ // MTP context uses a plain attention KV cache holding only the
+ // nextn layer(s) - same pattern as the hybrid Qwen3.5 MTP context.
+ llama_kv_cache::layer_filter_cb filter =
+ [&](uint32_t il) { return il >= hparams.n_layer(); };
+
+ res = new llama_kv_cache(
+ *this,
+ hparams,
+ params.type_k,
+ params.type_v,
+ !cparams.flash_attn,
+ cparams.offload_kqv,
+ cparams.kv_unified,
+ cparams.n_ctx_seq,
+ cparams.n_seq_max,
+ 1,
+ hparams.n_swa,
+ hparams.swa_type,
+ nullptr,
+ filter,
+ nullptr,
+ nullptr);
+ } else {
+ // Main context: DSA cache for the trunk layers only - the nextn
+ // layer(s) are never attended by the trunk graph.
+ llama_kv_cache::layer_filter_cb filter = nullptr;
+ if (hparams.n_layer_nextn > 0) {
+ filter = [&](uint32_t il) { return il < hparams.n_layer(); };
+ }
+
+ res = new llama_kv_cache_dsa(
+ *this,
+ params.type_k,
+ params.type_v,
+ !cparams.flash_attn,
+ cparams.offload_kqv,
+ cparams.kv_unified,
+ cparams.n_ctx_seq,
+ cparams.n_seq_max,
+ 1,
+ hparams.n_swa,
+ hparams.swa_type,
+ filter,
+ nullptr);
+ }
+ } break;
// Models that need standard caching should rely on recurrent/hybrid
// checks
default:
@@ -2194,7 +2243,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
filter = [&](uint32_t il) { return il >= hparams.n_layer(); };
}
- if ((arch == LLM_ARCH_STEP35 || arch == LLM_ARCH_HY_V3) && hparams.n_layer_nextn > 0) {
+ if ((arch == LLM_ARCH_STEP35 || arch == LLM_ARCH_HY_V3 || arch == LLM_ARCH_GLM_DSA) && hparams.n_layer_nextn > 0) {
if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP) {
filter = [&](uint32_t il) { return il >= hparams.n_layer(); };
} else {
diff --git a/src/models/glm-dsa.cpp b/src/models/glm-dsa.cpp
index df190e1f63..bd1c4df212 100644
--- a/src/models/glm-dsa.cpp
+++ b/src/models/glm-dsa.cpp
@@ -72,15 +72,27 @@ void llama_model_glm_dsa::load_arch_hparams(llama_model_loader & ml) {
ml.get_key_or_arr(LLM_KV_ATTENTION_INDEXER_TYPES, hparams.is_indexer_full_impl, hparams.n_layer(), false);
switch (hparams.n_layer()) {
- case 78: type = LLM_TYPE_744B_A40B; break;
+ case 78: // GGUF with NextN/MTP metadata: n_layer() excludes the nextn layer
+ case 79:
+ type = LLM_TYPE_744B_A40B; break;
default: type = LLM_TYPE_UNKNOWN;
}
}
-void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) {
+void llama_model_glm_dsa::load_arch_tensors(llama_model_loader & ml) {
LLAMA_LOAD_LOCALS;
const int64_t n_expert_shared = hparams.n_expert_shared;
+ // MTP-only: the GGUF carries only the NextN/MTP block(s) (user split target/draft).
+ const bool mtp_only = (hparams.n_layer_nextn > 0) && (ml.get_weight("blk.0.attn_norm.weight") == nullptr);
+ // Trunk-only: the GGUF declares MTP layers in metadata but the actual MTP
+ // tensors live in a separate file (or were stripped at conversion). Mark
+ // MTP tensors NOT_REQUIRED so the trunk loads cleanly.
+ const std::string mtp_probe = "blk." + std::to_string(n_layer) + ".nextn.eh_proj.weight";
+ const bool trunk_only = (hparams.n_layer_nextn > 0) && (ml.get_weight(mtp_probe.c_str()) == nullptr);
+ const int trunk_flags = mtp_only ? TENSOR_NOT_REQUIRED : 0;
+ const int mtp_flags = trunk_only ? TENSOR_NOT_REQUIRED : 0;
+
const bool is_mla = hparams.is_mla();
if (!is_mla) {
throw std::runtime_error("GLM_DSA architecture requires MLA");
@@ -109,12 +121,9 @@ void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) {
}
for (int i = 0; i < n_layer_all; ++i) {
- int flags = 0;
- if (i >= n_layer) {
- // skip all tensors in the NextN layers
- // TODO @ngxson : TENSOR_NOT_REQUIRED was a hack, need to remove it later
- flags |= TENSOR_SKIP | TENSOR_NOT_REQUIRED;
- }
+ // NextN/MTP layers (i >= n_layer) are full decoder blocks used by the
+ // LLM_GRAPH_TYPE_DECODER_MTP draft head; load them like qwen35moe/step35/hy_v3.
+ const int flags = (i >= n_layer) ? mtp_flags : trunk_flags;
auto & layer = layers[i];
@@ -167,7 +176,7 @@ void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) {
layer.ffn_up_shexp = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, flags);
}
- // NextN/MTP tensors (preserved but unused) - conditionally load for last n_layer_nextn
+ // NextN/MTP tensors - the NextN-specific wiring around the extra decoder block
if (i >= n_layer) {
layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
@@ -182,6 +191,9 @@ void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) {
}
std::unique_ptr<llm_graph_context> llama_model_glm_dsa::build_arch_graph(const llm_graph_params & params) const {
+ if (params.gtype == LLM_GRAPH_TYPE_DECODER_MTP) {
+ return std::make_unique<graph_mtp>(*this, params);
+ }
return std::make_unique<graph>(*this, params);
}
@@ -469,7 +481,9 @@ llama_model_glm_dsa::graph::graph(const llama_model & model, const llm_graph_par
Qcur, Kcur, Vcur, nullptr, nullptr, model.layers[il].wv_b, top_k, kq_scale, il);
}
}
- if (il == n_layer - 1 && inp_out_ids) {
+ // when unmasked nextn embeddings are requested, t_h_nextn must keep all rows,
+ // so the early output masking has to be skipped (it is applied after the final norm instead)
+ if (il == n_layer - 1 && inp_out_ids && (!cparams.embeddings_nextn || cparams.embeddings_nextn_masked)) {
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
@@ -532,6 +546,14 @@ llama_model_glm_dsa::graph::graph(const llama_model & model, const llm_graph_par
cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+ // post-norm hidden state feeds the NextN/MTP draft head
+ cb(cur, "h_nextn", -1);
+ res->t_h_nextn = cur;
+
+ if (cparams.embeddings_nextn && !cparams.embeddings_nextn_masked && inp_out_ids) {
+ cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+ }
+
cb(cur, "result_norm", -1);
res->t_embd = cur;
@@ -543,3 +565,242 @@ llama_model_glm_dsa::graph::graph(const llama_model & model, const llm_graph_par
ggml_build_forward_expand(gf, cur);
}
+
+// LLM_GRAPH_TYPE_DECODER_MTP draft head for GLM-5.2 (GLM_DSA).
+// Semantics mirror the deepseek-family NextN/MTP layer:
+// enorm(embed) + hnorm(prev_hidden) -> concat(e, h) -> eh_proj ->
+// full glm_dsa decoder block (dense MLA attention + sigmoid-gated MoE FFN
+// with shared expert, exactly as the trunk deepseek2 graph builds it) ->
+// shared_head_norm (fallback output_norm) -> shared LM head.
+// The DSA indexer is not used at runtime (same as the trunk graph).
+llama_model_glm_dsa::graph_mtp::graph_mtp(const llama_model & model, const llm_graph_params & params)
+ : llm_graph_context(params) {
+ GGML_ASSERT(hparams.n_layer_nextn > 0 && "GLM_DSA MTP requires n_layer_nextn > 0");
+ GGML_ASSERT(hparams.n_layer_nextn == 1 && "GLM_DSA MTP currently only supports a single MTP block");
+ GGML_ASSERT(hparams.is_mla() && "GLM_DSA MTP requires MLA");
+
+ const int il = hparams.n_layer() + cparams.nextn_layer_offset;
+ GGML_ASSERT(cparams.nextn_layer_offset >= 0 &&
+ cparams.nextn_layer_offset < (int) hparams.n_layer_nextn &&
+ "nextn_layer_offset out of range [0, n_layer_nextn)");
+ const auto & layer = model.layers[il];
+
+ GGML_ASSERT(layer.nextn.eh_proj && "MTP block missing nextn.eh_proj");
+ GGML_ASSERT(layer.nextn.enorm && "MTP block missing nextn.enorm");
+ GGML_ASSERT(layer.nextn.hnorm && "MTP block missing nextn.hnorm");
+ GGML_ASSERT(layer.ffn_gate_inp && "MTP block missing ffn_gate_inp");
+
+ // note: these are the actual head sizes you get when treating as MHA or after "decompression" using wv_b for MLA
+ const int64_t n_embd_head_k = hparams.n_embd_head_k_mla();
+
+ const int64_t n_embd_head_qk_rope = hparams.n_rot();
+ const int64_t n_embd_head_qk_nope = n_embd_head_k - n_embd_head_qk_rope;
+
+ const uint32_t kv_lora_rank = hparams.n_lora_kv;
+
+ // We have to pre-scale kq_scale and attn_factor to make the YaRN RoPE work correctly.
+ // See the deepseek2 trunk graph for the detailed explanation - this must match it EXACTLY.
+ GGML_ASSERT(ext_factor >= 0.0f);
+ const float attn_factor_org = attn_factor * (1.0f + 0.1f * logf(1.0f / freq_scale));
+
+ const float mscale = attn_factor_org * (1.0f + 0.1f * hparams.rope_yarn_log_mul * logf(1.0f / freq_scale));
+ const float kq_scale = 1.0f * mscale * mscale / sqrtf(float(n_embd_head_k));
+
+ // TODO: extract in a common llm_graph_context::build_inp_embd_h()
+ auto inp = std::make_unique<llm_graph_input_embd_h>(hparams.n_embd);
+
+ inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
+ ggml_set_input(inp->tokens);
+
+ inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd_inp(), n_tokens);
+ ggml_set_input(inp->embd);
+
+ ggml_tensor * tok_embd;
+ if (ubatch.token) {
+ ggml_tensor * tok_embd_w = layer.nextn.embed_tokens ? layer.nextn.embed_tokens : model.tok_embd;
+
+ tok_embd = ggml_get_rows(ctx0, tok_embd_w, inp->tokens);
+ } else {
+ tok_embd = inp->embd;
+ }
+ cb(tok_embd, "mtp_tok_embd", il);
+
+ inp->h = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, hparams.n_embd, n_tokens);
+ ggml_set_input(inp->h);
+ ggml_set_name(inp->h, "mtp_h_input");
+
+ ggml_tensor * h_embd = inp->h;
+
+ res->add_input(std::move(inp));
+
+ ggml_tensor * inp_pos = build_inp_pos();
+ ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+ // MLA with the absorption optimization uses a K-only cache (V is a view of K)
+ auto * inp_attn = build_attn_inp_k();
+
+ ggml_tensor * h_norm = build_norm(h_embd, layer.nextn.hnorm, nullptr, LLM_NORM_RMS, il);
+ cb(h_norm, "mtp_hnorm", il);
+
+ ggml_tensor * e_norm = build_norm(tok_embd, layer.nextn.enorm, nullptr, LLM_NORM_RMS, il);
+ cb(e_norm, "mtp_enorm", il);
+
+ ggml_tensor * concat = ggml_concat(ctx0, e_norm, h_norm, /*dim=*/ 0);
+ cb(concat, "mtp_concat", il);
+
+ ggml_tensor * cur = build_lora_mm(layer.nextn.eh_proj, concat, layer.nextn.eh_proj_s);
+ cb(cur, "mtp_eh_proj", il);
+
+ ggml_tensor * inpSA = cur;
+
+ cur = build_norm(cur, layer.attn_norm, nullptr, LLM_NORM_RMS, il);
+ cb(cur, "mtp_attn_norm", il);
+
+ // self-attention: dense MLA, same construction as the deepseek2 trunk graph
+ {
+ ggml_tensor * q = ggml_mul_mat(ctx0, layer.wq_a, cur);
+ cb(q, "mtp_q", il);
+
+ q = build_norm(q, layer.attn_q_a_norm, nullptr, LLM_NORM_RMS, il);
+ cb(q, "mtp_q", il);
+
+ q = ggml_mul_mat(ctx0, layer.wq_b, q);
+ cb(q, "mtp_q", il);
+
+ // split into {n_embd_head_qk_nope, n_head, n_tokens}
+ ggml_tensor * q_nope =
+ ggml_view_3d(ctx0, q, n_embd_head_qk_nope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+ ggml_row_size(q->type, n_embd_head_k) * n_head, 0);
+ cb(q_nope, "mtp_q_nope", il);
+
+ // and {n_embd_head_qk_rope, n_head, n_tokens}
+ ggml_tensor * q_pe = ggml_view_3d(
+ ctx0, q, n_embd_head_qk_rope, n_head, n_tokens, ggml_row_size(q->type, n_embd_head_k),
+ ggml_row_size(q->type, n_embd_head_k) * n_head, ggml_row_size(q->type, n_embd_head_qk_nope));
+ cb(q_pe, "mtp_q_pe", il);
+
+ ggml_tensor * kv_cmpr_pe = ggml_mul_mat(ctx0, layer.wkv_a_mqa, cur);
+ cb(kv_cmpr_pe, "mtp_kv_cmpr_pe", il);
+
+ // split into {kv_lora_rank, n_tokens}
+ ggml_tensor * kv_cmpr =
+ ggml_view_2d(ctx0, kv_cmpr_pe, kv_lora_rank, n_tokens,
+ ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope), 0);
+ cb(kv_cmpr, "mtp_kv_cmpr", il);
+
+ // and {n_embd_head_qk_rope, 1, n_tokens}
+ ggml_tensor * k_pe = ggml_view_3d(ctx0, kv_cmpr_pe, n_embd_head_qk_rope, 1, n_tokens,
+ ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+ ggml_row_size(kv_cmpr_pe->type, kv_lora_rank + n_embd_head_qk_rope),
+ ggml_row_size(kv_cmpr_pe->type, kv_lora_rank));
+ cb(k_pe, "mtp_k_pe", il);
+
+ q_pe = ggml_rope_ext(ctx0, q_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+ ext_factor, attn_factor, beta_fast, beta_slow);
+ cb(q_pe, "mtp_q_pe", il);
+
+ k_pe = ggml_rope_ext(ctx0, k_pe, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+ ext_factor, attn_factor, beta_fast, beta_slow);
+ cb(k_pe, "mtp_k_pe", il);
+
+ kv_cmpr = build_norm(kv_cmpr, layer.attn_kv_a_norm, nullptr, LLM_NORM_RMS, il);
+ cb(kv_cmpr, "mtp_kv_cmpr", il);
+
+ // {n_embd_head_qk_nope, n_tokens, n_head}
+ q_nope = ggml_permute(ctx0, q_nope, 0, 2, 1, 3);
+ cb(q_nope, "mtp_q_nope_perm", il);
+
+ // {n_embd_head_qk_nope, kv_lora_rank, n_head} x {n_embd_head_qk_nope, n_tokens, n_head}
+ ggml_tensor * q_nope_absorbed = ggml_mul_mat(ctx0, layer.wk_b, q_nope);
+ cb(q_nope_absorbed, "mtp_q_nope_absorbed", il);
+
+ // {kv_lora_rank, n_head, n_tokens}
+ q_nope_absorbed = ggml_permute(ctx0, q_nope_absorbed, 0, 2, 1, 3);
+ cb(q_nope_absorbed, "mtp_q_nope_absorbed_perm", il);
+
+ // {n_embd_head_qk_rope + kv_lora_rank, n_head, n_tokens}
+ // note: rope must go first for in-place context shifting in build_rope_shift()
+ ggml_tensor * Qcur = ggml_concat(ctx0, q_nope_absorbed, q_pe, 0);
+ cb(Qcur, "mtp_Qcur", il);
+
+ kv_cmpr = ggml_reshape_3d(ctx0, kv_cmpr, kv_lora_rank, 1, n_tokens);
+ cb(kv_cmpr, "mtp_kv_cmpr_reshape", il);
+
+ // {n_embd_head_qk_rope + kv_lora_rank, 1, n_tokens}
+ ggml_tensor * Kcur = ggml_concat(ctx0, kv_cmpr, k_pe, 0);
+ cb(Kcur, "mtp_Kcur", il);
+
+ // {kv_lora_rank, 1, n_tokens}
+ ggml_tensor * Vcur = kv_cmpr;
+ cb(Vcur, "mtp_Vcur", il);
+
+ // note: MLA with the absorption optimization converts into MQA (ie: GQA with 1 group)
+ cur = build_attn(inp_attn,
+ layer.wo, NULL, layer.wo_s,
+ Qcur, Kcur, Vcur, nullptr, nullptr, layer.wv_b, kq_scale, il);
+ cb(cur, "mtp_attn_out", il);
+ }
+
+ ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+ cb(ffn_inp, "mtp_ffn_inp", il);
+
+ cur = build_norm(ffn_inp, layer.ffn_norm, NULL, LLM_NORM_RMS, il);
+ cb(cur, "mtp_ffn_norm", il);
+
+ // MoE FFN with shared expert - same construction as the deepseek2 trunk graph
+ ggml_tensor * moe_out = build_moe_ffn(cur,
+ layer.ffn_gate_inp,
+ layer.ffn_up_exps,
+ layer.ffn_gate_exps,
+ layer.ffn_down_exps,
+ layer.ffn_exp_probs_b,
+ n_expert, n_expert_used,
+ LLM_FFN_SILU, hparams.expert_weights_norm,
+ hparams.expert_weights_scale,
+ (llama_expert_gating_func_type) hparams.expert_gating_func,
+ il,
+ nullptr,
+ layer.ffn_gate_up_exps,
+ layer.ffn_up_exps_s,
+ layer.ffn_gate_exps_s,
+ layer.ffn_down_exps_s);
+ cb(moe_out, "mtp_ffn_moe_out", il);
+
+ // FFN shared expert
+ ggml_tensor * ffn_shexp =
+ build_ffn(cur,
+ layer.ffn_up_shexp, NULL, layer.ffn_up_shexp_s,
+ layer.ffn_gate_shexp, NULL, layer.ffn_gate_shexp_s,
+ layer.ffn_down_shexp, NULL, layer.ffn_down_shexp_s,
+ NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
+ cb(ffn_shexp, "mtp_ffn_shexp", il);
+
+ cur = ggml_add(ctx0, moe_out, ffn_shexp);
+ cb(cur, "mtp_ffn_out", il);
+
+ cur = ggml_add(ctx0, cur, ffn_inp);
+ cb(cur, "mtp_post_ffn", il);
+
+ // shared_head_norm applied after the decoder block, before the shared LM head.
+ // The post-norm hidden state seeds the next MTP step.
+ ggml_tensor * head_norm_w = layer.nextn.shared_head_norm
+ ? layer.nextn.shared_head_norm
+ : model.output_norm;
+ GGML_ASSERT(head_norm_w && "GLM_DSA MTP: missing both nextn.shared_head_norm and output_norm");
+ cur = build_norm(cur, head_norm_w, nullptr, LLM_NORM_RMS, -1);
+
+ cb(cur, "h_nextn", -1);
+ res->t_h_nextn = cur;
+
+ cur = ggml_get_rows(ctx0, cur, inp_out_ids);
+ cb(cur, "mtp_shared_head_norm", -1);
+
+ ggml_tensor * head_w = layer.nextn.shared_head_head ? layer.nextn.shared_head_head : model.output;
+ ggml_tensor * head_s = layer.nextn.shared_head_head ? layer.nextn.shared_head_head_s : model.output_s;
+ GGML_ASSERT(head_w && "GLM_DSA MTP: missing LM head (nextn.shared_head_head or model.output)");
+ cur = build_lora_mm(head_w, cur, head_s);
+ cb(cur, "result_output", -1);
+
+ res->t_logits = cur;
+ ggml_build_forward_expand(gf, cur);
+}
diff --git a/src/models/models.h b/src/models/models.h
index 92ebfafa1e..c73136f3bd 100644
--- a/src/models/models.h
+++ b/src/models/models.h
@@ -1237,6 +1237,10 @@ struct llama_model_glm_dsa : public llama_model_base {
graph(const llama_model & model, const llm_graph_params & params);
};
+ struct graph_mtp : public llm_graph_context {
+ graph_mtp(const llama_model & model, const llm_graph_params & params);
+ };
+
std::unique_ptr<llm_graph_context> build_arch_graph(const llm_graph_params & params) const override;
};