ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-14 15:57:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

WIP: loads and runs, but not correct

Browse Source 
Very high PPL, empty TG.
This commit is contained in:
Kawrakow
2026-02-19 06:58:36 +00:00
parent b2cb4512c5
commit 8737d3d924
12 changed files with 353 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llama-arch.cpp
View File

@@ -30,6 +30,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_QWEN3NEXT, "qwen3next" },
{ LLM_ARCH_QWEN3VL, "qwen3vl" },
{ LLM_ARCH_QWEN3VLMOE, "qwen3vlmoe" },
{ LLM_ARCH_QWEN35MOE, "qwen35moe" },
{ LLM_ARCH_PHI2, "phi2" },
{ LLM_ARCH_PHI3, "phi3" },
{ LLM_ARCH_PLAMO, "plamo" },
@@ -159,6 +160,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_ATTENTION_INDEXER_HEAD_COUNT, "%s.attention.indexer.head_count" },
{ LLM_KV_ATTENTION_INDEXER_KEY_LENGTH, "%s.attention.indexer.key_length" },
{ LLM_KV_ATTENTION_INDEXER_TOP_K, "%s.attention.indexer.top_k" },
{ LLM_KV_FULL_ATTENTION_INTERVAL, "%s.full_attention_interval" },
{ LLM_KV_ROPE_DIMENSION_COUNT, "%s.rope.dimension_count" },

5
src/llama-arch.h
View File

@@ -29,6 +29,7 @@ enum llm_arch {
LLM_ARCH_QWEN3NEXT,
LLM_ARCH_QWEN3VL,
LLM_ARCH_QWEN3VLMOE,
LLM_ARCH_QWEN35MOE,
LLM_ARCH_PHI2,
LLM_ARCH_PHI3,
LLM_ARCH_PLAMO,
@@ -152,7 +153,7 @@ enum llm_kv {
LLM_KV_ATTENTION_INDEXER_HEAD_COUNT,
LLM_KV_ATTENTION_INDEXER_KEY_LENGTH,
LLM_KV_ATTENTION_INDEXER_TOP_K,
LLM_KV_FULL_ATTENTION_INTERVAL,
LLM_KV_ROPE_DIMENSION_COUNT,
LLM_KV_ROPE_DIMENSION_COUNT_PER_LAYER,
@@ -285,6 +286,8 @@ enum llm_tensor {
LLM_TENSOR_SSM_NORM,
LLM_TENSOR_SSM_OUT,
LLM_TENSOR_SSM_BETA_ALPHA,
LLM_TENSOR_SSM_ALPHA,
LLM_TENSOR_SSM_BETA,
LLM_TENSOR_ATTN_Q_A,
LLM_TENSOR_ATTN_Q_B,
LLM_TENSOR_ATTN_KV_A_MQA,

145
src/llama-build-context.cpp
View File

@@ -142,7 +142,7 @@ ggml_cgraph * llm_build_context::build_k_shift() {
ggml_set_input(lctx.inp_K_shift);
for (int il = 0; il < n_layer; ++il) {
if (model.arch == LLM_ARCH_QWEN3NEXT && hparams.is_recurrent(il)) {
if ((model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) && hparams.is_recurrent(il)) {
continue;
}
if (kv_self.k_l[il] == nullptr) {
@@ -241,7 +241,7 @@ ggml_cgraph * llm_build_context::build_defrag(const std::vector<uint32_t> & ids)
}
for (int il = 0; il < n_layer; ++il) {
if (model.arch == LLM_ARCH_QWEN3NEXT && hparams.is_recurrent(il)) {
if ((model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) && hparams.is_recurrent(il)) {
continue;
}
if (kv_self.k_l[il] == nullptr) {
@@ -4478,6 +4478,143 @@ ggml_cgraph * llm_build_context::build_qwen3next() {
return gf;
}
ggml_cgraph * llm_build_context::build_qwen35moe() {
static constexpr int QWEN3NEXT_CHUNK_SIZE = 64;
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, llama_model_max_nodes(model), false);
delta_net delta(lctx, batch);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
int sections[4];
std::copy(std::begin(hparams.rope_sections), std::begin(hparams.rope_sections) + 4, sections);
auto build_layer_attn = [&](ggml_tensor * cur, ggml_tensor * inp_pos, ggml_tensor * KQ_mask, int il) -> ggml_tensor * {
auto Qaux = llm_build_lora_mm(lctx, ctx0, model.layers[il].wq, cur);
auto Kcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wk, cur);
auto Vcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wv, cur);
cb(Qaux, "Qaux", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
ggml_build_forward_expand(gf, Qaux);
ggml_build_forward_expand(gf, Kcur);
ggml_build_forward_expand(gf, Vcur);
Qaux = ggml_reshape_3d(ctx0, Qaux, n_embd_head * 2, n_head, n_tokens);
auto Qcur = ggml_cont(ctx0, ggml_view_3d(ctx0, Qaux, n_embd_head, n_head, n_tokens, Qaux->nb[1], Qaux->nb[2], 0));
auto gate = ggml_cont_2d(ctx0, ggml_view_3d(ctx0, Qaux, n_embd_head, n_head, n_tokens, Qaux->nb[1], Qaux->nb[2], n_embd_head*ggml_element_size(Qaux)), n_embd_head*n_head, n_tokens);
cb(Qcur, "Qcur", il);
cb(gate, "gate", il);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
Qcur = llm_build_norm(ctx0, Qcur, hparams, model.layers[il].attn_q_norm, nullptr, LLM_NORM_RMS, cb, il);
cb(Qcur, "Qcur_normed", il);
Kcur = llm_build_norm(ctx0, Kcur, hparams, model.layers[il].attn_k_norm, nullptr, LLM_NORM_RMS, cb, il);
cb(Kcur, "Kcur_normed", il);
Qcur = ggml_rope_multi(ctx0, Qcur, inp_pos, nullptr,
n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
Kcur = ggml_rope_multi(ctx0, Kcur, inp_pos, nullptr,
n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
cb(Qcur, "Qcur_roped", il);
cb(Kcur, "Kcur_roped", il);
ggml_tensor * attn = llm_build_kv(ctx0, lctx, kv_self, gf, nullptr, nullptr,
Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv,
hparams.f_attention_scale == 0.0f ? 1.0f / sqrtf(float(n_embd_head)) : hparams.f_attention_scale, cb, il);
cb(attn, "attn_pregate", il);
gate = ggml_sigmoid(ctx0, gate);
cb(gate, "gate_sigmoid", il);
attn = ggml_mul(ctx0, attn, gate);
cb(attn, "attn_gated", il);
attn = llm_build_lora_mm(lctx, ctx0, model.layers[il].wo, attn);
cb(attn, "attn_output", il);
return attn;
};
ggml_tensor * inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
ggml_tensor * inp_pos = build_inp_pos();
ggml_tensor * inp_out_ids = n_tokens > 1 ? build_inp_out_ids() : nullptr;
ggml_tensor * KQ_mask = build_inp_KQ_mask();
lctx.inp_s_seq_qnext = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, 1, n_tokens);
cb(lctx.inp_s_seq_qnext, "inp_s_seq_qnext", -1);
ggml_set_input(lctx.inp_s_seq_qnext);
ggml_tensor * causal_mask = nullptr;
ggml_tensor * identity = nullptr;
ggml_tensor * diag_mask = nullptr;
causal_mask = ggml_tri(ctx0,
ggml_fill_inplace(ctx0, ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, QWEN3NEXT_CHUNK_SIZE, QWEN3NEXT_CHUNK_SIZE), 1.0f),
GGML_TRI_TYPE_LOWER);
identity = ggml_diag(ctx0, ggml_fill_inplace(ctx0, ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, QWEN3NEXT_CHUNK_SIZE), 1.0f));
diag_mask = ggml_add(ctx0, causal_mask, identity);
ggml_build_forward_expand(gf, causal_mask);
ggml_build_forward_expand(gf, identity);
ggml_build_forward_expand(gf, diag_mask);
ggml_tensor * cur = nullptr;
for (int il = 0; il < n_layer; ++il) {
ggml_tensor * inpSA = inpL;
cur = llm_build_norm(ctx0, inpL, hparams, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
if (hparams.is_recurrent(il)) {
cur = delta.build_layer_attn_linear(ctx0, gf, cur, causal_mask, identity, diag_mask, il, cb);
} else {
cur = build_layer_attn(cur, inp_pos, KQ_mask, il);
}
if (il == n_layer - 1 && inp_out_ids) {
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
cur = ggml_add(ctx0, cur, inpSA);
cb(cur, "attn_residual", il);
cur = llm_build_std_moe_ffn(ctx0, lctx, model.layers[il].ffn_norm, cur,
model.layers[il].ffn_gate_inp, nullptr,
model.layers[il].ffn_up_exps, nullptr,
model.layers[il].ffn_gate_exps, nullptr,
model.layers[il].ffn_down_exps, nullptr,
nullptr,
model.layers[il].ffn_up_shexp, nullptr, // we don't have shared expert biases?
model.layers[il].ffn_gate_shexp, nullptr,
model.layers[il].ffn_down_shexp, nullptr,
n_expert, n_expert_used,
LLM_FFN_SILU, true, false, 0.0f,
LLM_EXPERT_GATING_FUNC_SOFTMAX,
LLM_FFN_SILU, cb, il, gf, true, model.layers[il].ffn_up_gate_exps, nullptr, model.layers[il].ffn_gate_inp_shexp);
cur = lctx.cvec.apply_to(ctx0, cur, il);
cb(cur, "l_out", il);
inpL = cur;
}
cur = build_output(lctx, ctx0, inpL, model.output, model.output_norm, cb);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
ggml_cgraph * llm_build_context::build_qwen3vl() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, llama_model_max_nodes(model), false);
@@ -9508,6 +9645,10 @@ ggml_cgraph * llm_build_context::llama_build_graph(
{
result = llm.build_qwen3next();
} break;
case LLM_ARCH_QWEN35MOE:
{
result = llm.build_qwen35moe();
} break;
case LLM_ARCH_QWEN3VL:
{
result = llm.build_qwen3vl();

2
src/llama-build-context.h
View File

@@ -206,6 +206,8 @@ struct llm_build_context {
ggml_cgraph * build_qwen3next();
ggml_cgraph * build_qwen35moe();
ggml_cgraph * build_phi2();
ggml_cgraph * build_phi3();

55
src/llama-delta-net.cpp
View File

@@ -92,6 +92,9 @@ std::pair<ggml_tensor *, ggml_tensor *> delta_net::build_delta_net_chunking(ggml
GGML_ASSERT(v->ne[2] == n_tokens);
GGML_ASSERT(k->ne[2] == n_tokens);
GGML_ASSERT(g->ne[0] == H_v && g->ne[1] == n_tokens && g->ne[2] == n_seqs);
if (beta->ne[0] != H_v || beta->ne[2] != n_tokens || beta->ne[3] != n_seqs) {
printf("beta: %ld x %ld x %ld, expected %ld x %ld x %ld\n", beta->ne[0], beta->ne[2], beta->ne[3], H_v, n_tokens, n_seqs);
}
GGML_ASSERT(beta->ne[0] == H_v && beta->ne[2] == n_tokens && beta->ne[3] == n_seqs);
GGML_ASSERT(state->ne[0] == S_v && state->ne[1] == S_v && state->ne[2] == H_v && state->ne[3] == n_seqs);
GGML_ASSERT(H_k == H_v);
@@ -320,10 +323,6 @@ std::pair<ggml_tensor *, ggml_tensor *> delta_net::build_delta_net_autoregressiv
GGML_ASSERT(H_k == H_v);
GGML_ASSERT(state->ne[0] == S_v && state->ne[1] == S_v && state->ne[2] == H_v && state->ne[3] == n_seqs);
//const float eps_norm = hparams.f_norm_rms_eps;
//q = ggml_l2_norm(ctx0, q, eps_norm);
//k = ggml_l2_norm(ctx0, k, eps_norm);
const float scale = 1.0f / sqrtf(S_v);
q = ggml_scale(ctx0, q, scale);
@@ -464,35 +463,45 @@ ggml_tensor * delta_net::build_layer_attn_linear_core(ggml_context * ctx0, ggml_
const uint32_t qnext_state_slots = llm_build_context::llama_kv_qnext_state_slots(kv_self);
GGML_ASSERT(qnext_state_slots > 0);
const int64_t n_tok = cur->ne[1];
const int64_t n_seqs = 1;
const int64_t n_seq_tokens = n_tok;
auto qkvz = build_qkvz(ctx0, cur, il, cb);
ggml_tensor * qkv_mixed = qkvz.first;
ggml_tensor * z = qkvz.second;
ggml_tensor * mixed_ba = llm_build_context::llm_build_lora_mm(lctx, ctx0, model.layers[il].ssm_beta_alpha, cur);
cb(mixed_ba, "linear_attn_mixed_ba", il);
ggml_tensor *alpha, *beta;
if (model.layers[il].ssm_beta_alpha) {
ggml_tensor * mixed_ba = llm_build_context::llm_build_lora_mm(lctx, ctx0, model.layers[il].ssm_beta_alpha, cur);
cb(mixed_ba, "linear_attn_mixed_ba", il);
int64_t ba_new_dim = 2 * num_v_heads / num_k_heads;
ggml_tensor * mixed_ba_reshaped = ggml_reshape_4d(ctx0, mixed_ba, ba_new_dim, num_k_heads, n_tok, 1);
int64_t ba_new_dim = 2 * num_v_heads / num_k_heads;
ggml_tensor * mixed_ba_reshaped = ggml_reshape_4d(ctx0, mixed_ba, ba_new_dim, num_k_heads, n_tok, 1);
int64_t split_sizes_ba[2] = {
num_v_heads / num_k_heads,
num_v_heads / num_k_heads
};
int64_t split_sizes_ba[2] = {
num_v_heads / num_k_heads,
num_v_heads / num_k_heads
};
ggml_tensor * b = ggml_view_4d(ctx0, mixed_ba_reshaped, split_sizes_ba[0], num_k_heads, n_tok, 1,
mixed_ba_reshaped->nb[1], mixed_ba_reshaped->nb[2], mixed_ba_reshaped->nb[3], 0);
cb(b, "b", il);
ggml_tensor * b = ggml_view_4d(ctx0, mixed_ba_reshaped, split_sizes_ba[0], num_k_heads, n_tok, 1,
mixed_ba_reshaped->nb[1], mixed_ba_reshaped->nb[2], mixed_ba_reshaped->nb[3], 0);
cb(b, "b", il);
ggml_tensor * a = ggml_view_4d(ctx0, mixed_ba_reshaped, split_sizes_ba[1], num_k_heads, n_tok, 1,
mixed_ba_reshaped->nb[1], mixed_ba_reshaped->nb[2], mixed_ba_reshaped->nb[3],
split_sizes_ba[0] * ggml_element_size(mixed_ba_reshaped));
cb(a, "a", il);
ggml_tensor * a = ggml_view_4d(ctx0, mixed_ba_reshaped, split_sizes_ba[1], num_k_heads, n_tok, 1,
mixed_ba_reshaped->nb[1], mixed_ba_reshaped->nb[2], mixed_ba_reshaped->nb[3],
split_sizes_ba[0] * ggml_element_size(mixed_ba_reshaped));
cb(a, "a", il);
ggml_tensor * beta = ggml_cont_4d(ctx0, b, num_v_heads, 1, n_tok, 1);
ggml_tensor * alpha = ggml_cont_3d(ctx0, a, num_v_heads, n_tok, 1);
beta = ggml_cont_4d(ctx0, b, num_v_heads, 1, n_tok, 1);
alpha = ggml_cont_3d(ctx0, a, num_v_heads, n_tok, 1);
} else {
beta = llm_build_context::llm_build_lora_mm(lctx, ctx0, model.layers[il].ssm_beta, cur);
beta = ggml_reshape_4d(ctx0, beta, num_v_heads, 1, n_tok, 1);
alpha = llm_build_context::llm_build_lora_mm(lctx, ctx0, model.layers[il].ssm_alpha, cur);
// Why???
alpha = ggml_cont_3d(ctx0, alpha, num_v_heads, n_seq_tokens, n_seqs);
}
cb(beta, "beta", il);
cb(alpha, "alpha", il);
@@ -645,7 +654,7 @@ ggml_tensor * delta_net::build_layer_attn_linear(ggml_context * ctx0, ggml_cgrap
GGML_ASSERT(model.layers[il].ssm_conv1d != nullptr);
GGML_ASSERT(model.layers[il].ssm_dt != nullptr);
GGML_ASSERT(model.layers[il].ssm_a != nullptr);
GGML_ASSERT(model.layers[il].ssm_beta_alpha != nullptr);
GGML_ASSERT(model.layers[il].ssm_beta_alpha != nullptr || (model.layers[il].ssm_alpha != nullptr && model.layers[il].ssm_beta != nullptr));
GGML_ASSERT(model.layers[il].ssm_norm != nullptr);
GGML_ASSERT(model.layers[il].ssm_out != nullptr);
GGML_ASSERT(model.layers[il].wqkv != nullptr || model.layers[il].ssm_in != nullptr);

35
src/llama-hparams.cpp
View File

@@ -85,8 +85,8 @@ void llm_load_hparams(
std::fill(hparams.n_ff_arr.begin(), hparams.n_ff_arr.end(), 0);
std::fill(hparams.recurrent_layer_arr.begin(), hparams.recurrent_layer_arr.end(), false);
ml.get_key_or_arr(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff_arr, hparams.n_layer);
ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head_arr, hparams.n_layer);
ml.get_key_or_arr(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff_arr, hparams.n_layer, false);
ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head_arr, hparams.n_layer, false);
// n_head_kv is optional, default to n_head
hparams.n_head_kv_arr = hparams.n_head_arr;
@@ -476,6 +476,37 @@ void llm_load_hparams(
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_QWEN35MOE:
{
ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false);
ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, true);
// Load linear attention (gated delta net) parameters
ml.get_key(LLM_KV_SSM_CONV_KERNEL, hparams.ssm_d_conv);
ml.get_key(LLM_KV_SSM_INNER_SIZE, hparams.ssm_d_inner);
ml.get_key(LLM_KV_SSM_STATE_SIZE, hparams.ssm_d_state);
ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
ml.get_key(LLM_KV_SSM_GROUP_COUNT, hparams.ssm_n_group);
// Mark recurrent layers (linear attention layers)
{
uint32_t full_attn_interval = 4;
ml.get_key(LLM_KV_FULL_ATTENTION_INTERVAL, full_attn_interval, false);
for (uint32_t i = 0; i < hparams.n_layer; ++i) {
hparams.recurrent_layer_arr[i] = ((i + 1) % full_attn_interval != 0);
}
}
switch (hparams.n_layer) {
case 28: model.type = e_model::MODEL_80B_A3B; break;
case 48: model.type = e_model::MODEL_80B_A3B; break;
case 60: model.type = e_model::MODEL_397B_A17B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_QWEN3VLMOE:
{
ml.get_key(LLM_KV_NUM_DEEPSTACK_LAYERS, hparams.n_deepstack_layers, false);

78
src/llama-load-tensors.cpp
View File

@@ -75,6 +75,8 @@ struct create_tensors_helper : public create_tensors_helper_interface {
bool create_qwen3next_tensors(const LLM_TN & tn);
bool create_qwen35moe_tensors(const LLM_TN & tn);
bool create_phi2_tensors(const LLM_TN & tn);
bool create_phi3_tensors(const LLM_TN & tn);
@@ -1387,6 +1389,80 @@ bool create_tensors_helper::create_qwen3next_tensors(const LLM_TN & tn) {
return use_mmap_buffer;
}
bool create_tensors_helper::create_qwen35moe_tensors(const LLM_TN & tn) {
LOADING_PRELUDE
model.tok_embd = create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED);
if (model.output == NULL) {
model.output = create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED);
}
}
const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
const int64_t head_k_dim = hparams.ssm_d_state;
const int64_t head_v_dim = hparams.ssm_d_state;
const int64_t n_k_heads = hparams.ssm_n_group;
const int64_t n_v_heads = hparams.ssm_dt_rank;
const int64_t key_dim = head_k_dim * n_k_heads;
const int64_t value_dim = head_v_dim * n_v_heads;
const int64_t conv_dim = key_dim * 2 + value_dim;
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
layer.attn_post_norm = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), { n_embd }, 0);
layer.ffn_norm = layer.attn_post_norm;
if (!hparams.is_recurrent(i)) {
// Attention layers
layer.wq = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), { n_embd, n_embd_head_k * n_head * 2 }, 0);
layer.wk = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), { n_embd, n_embd_k_gqa }, 0);
layer.wv = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), { n_embd, n_embd_v_gqa }, 0);
layer.wo = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
// Q/K normalization for attention layers
layer.attn_q_norm = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), { n_embd_head_k }, 0);
layer.attn_k_norm = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), { n_embd_head_k }, 0);
} else {
// Linear attention (gated delta net) specific tensors
// Create tensors with calculated dimensions
layer.wqkv = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), { n_embd, key_dim * 2 + value_dim }, llama_model_loader::TENSOR_NOT_REQUIRED);
layer.wqkv_gate = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_GATE, "weight", i), { n_embd, value_dim }, llama_model_loader::TENSOR_NOT_REQUIRED);
layer.ssm_conv1d = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_CONV1D, "weight", i), { hparams.ssm_d_conv, conv_dim }, 0);
layer.ssm_dt = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_DT, "bias", i), { hparams.ssm_dt_rank }, 0);
layer.ssm_a = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_A_NOSCAN, i), { hparams.ssm_dt_rank }, 0);
layer.ssm_beta = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_BETA, "weight", i), { n_embd, n_v_heads }, 0);
layer.ssm_alpha = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_ALPHA, "weight", i), { n_embd, n_v_heads }, 0);
layer.ssm_norm = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_NORM, "weight", i), { head_v_dim }, 0);
layer.ssm_out = create_tensor(ctx_split, tn(LLM_TENSOR_SSM_OUT, "weight", i), { value_dim, n_embd }, 0);
}
layer.ffn_gate_inp = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), { n_embd, n_expert }, 0);
layer.ffn_gate_exps = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert }, 0);
layer.ffn_down_exps = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff_exp, n_embd, n_expert }, 0);
layer.ffn_up_exps = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert }, 0);
// Shared experts
const int64_t n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff;
layer.ffn_gate_inp_shexp = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_INP_SHEXP, "weight", i), { n_embd }, 0);
layer.ffn_gate_shexp = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), { n_embd, n_ff_shexp }, 0);
layer.ffn_up_shexp = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), { n_embd, n_ff_shexp }, 0);
layer.ffn_down_shexp = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_shexp, n_embd }, 0);
}
return use_mmap_buffer;
}
bool create_tensors_helper::create_mimo2_tensors(const LLM_TN & tn) {
LOADING_PRELUDE
@@ -3319,6 +3395,8 @@ bool create_tensors_helper::create_tensors() {
use_mmap_buffer = create_qwen3_moe_tensors(tn); break;
case LLM_ARCH_QWEN3NEXT:
use_mmap_buffer = create_qwen3next_tensors(tn); break;
case LLM_ARCH_QWEN35MOE:
use_mmap_buffer = create_qwen35moe_tensors(tn); break;
case LLM_ARCH_PHI2:
use_mmap_buffer = create_phi2_tensors(tn); break;
case LLM_ARCH_PHI3:

37
src/llama-model.cpp
View File

@@ -462,6 +462,40 @@ static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NA
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
},
},
{
LLM_ARCH_QWEN35MOE,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_GATE, "blk.%d.attn_gate" },
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
{ LLM_TENSOR_SSM_A_NOSCAN, "blk.%d.ssm_a" },
{ LLM_TENSOR_SSM_BETA, "blk.%d.ssm_beta" },
{ LLM_TENSOR_SSM_ALPHA, "blk.%d.ssm_alpha" },
//{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
{ LLM_TENSOR_SSM_NORM, "blk.%d.ssm_norm" },
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
{ LLM_TENSOR_FFN_GATE_INP_SHEXP, "blk.%d.ffn_gate_inp_shexp" },
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
},
},
{
LLM_ARCH_QWEN3VL,
{
@@ -1690,9 +1724,10 @@ const char * llama_model_type_name(e_model type) {
case MODEL_310B_A15B: return "310B.A15B";
case MODEL_300B_A47B: return "300B.A47B";
case MODEL_355B_A32B: return "355B.A32B";
case MODEL_397B_A17B: return "397B.A17B";
case MODEL_744B_A40B: return "744B.A40B";
case MODEL_E2B: return "E2B";
case MODEL_E4B: return "E4B";
default: return "?B";
default: return "?B";
}
}

3
src/llama-model.h
View File

@@ -116,6 +116,7 @@ enum e_model {
MODEL_310B_A15B,
MODEL_300B_A47B, // Ernie MoE big
MODEL_355B_A32B,
MODEL_397B_A17B, // Qwen-3.5-MoE
MODEL_744B_A40B,
MODEL_E2B,
MODEL_E4B,
@@ -292,6 +293,8 @@ struct llama_layer {
struct ggml_tensor * ssm_out = nullptr;
struct ggml_tensor * ssm_norm = nullptr;
struct ggml_tensor * ssm_beta_alpha = nullptr;
struct ggml_tensor * ssm_alpha = nullptr;
struct ggml_tensor * ssm_beta = nullptr;
// mamba
struct ggml_tensor * ssm_conv1d = nullptr;

11
src/llama-vocab.cpp
View File

@@ -360,6 +360,13 @@ struct llm_tokenizer_bpe : llm_tokenizer {
"(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
};
break;
case LLAMA_VOCAB_PRE_TYPE_QWEN35:
regex_exprs = {
// original regex from tokenizer.json
// "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+"
"(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
};
break;
case LLAMA_VOCAB_PRE_TYPE_PORO:
case LLAMA_VOCAB_PRE_TYPE_BLOOM:
case LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH:
@@ -1885,6 +1892,10 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
tokenizer_pre == "deepseek-r1-qwen") {
pre_type = LLAMA_VOCAB_PRE_TYPE_QWEN2;
clean_spaces = false;
} else if (
tokenizer_pre == "qwen35") {
pre_type = LLAMA_VOCAB_PRE_TYPE_QWEN35;
clean_spaces = false;
} else if (
tokenizer_pre == "stablelm2") {
pre_type = LLAMA_VOCAB_PRE_TYPE_STABLELM2;

1
src/llama-vocab.h
View File

@@ -50,6 +50,7 @@ enum llama_vocab_pre_type {
LLAMA_VOCAB_PRE_TYPE_GROK_2 = 39,
LLAMA_VOCAB_PRE_TYPE_MINIMAX_M2 = 40,
LLAMA_VOCAB_PRE_TYPE_GRANITE_DOCLING = 41,
LLAMA_VOCAB_PRE_TYPE_QWEN35 = 46,
};
struct LLM_KV;

15
src/llama.cpp
View File

@@ -569,7 +569,7 @@ bool llama_context::can_reuse_graph(const llama_batch & u_batch) {
bool llama_context::update_cache_copies() {
int n_layer = model.hparams.n_layer - model.hparams.nextn_predict_layers; //cache_copies.size()/2;
auto layer_has_attention_kv = [&](int il) {
return !(model.arch == LLM_ARCH_QWEN3NEXT && model.hparams.is_recurrent(il));
return !((model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) && model.hparams.is_recurrent(il));
};
if ((int)kv_self.k_l.size() != n_layer) return false;
if (!(kv_self.v_l.empty() || (int)kv_self.v_l.size() == n_layer)) return false;
@@ -656,7 +656,7 @@ static inline bool llama_qwen3next_is_recurrent_layer(
const llama_model & model,
const llama_hparams & hparams,
uint32_t il) {
return model.arch == LLM_ARCH_QWEN3NEXT && hparams.is_recurrent(il);
return (model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) && hparams.is_recurrent(il);
}
static inline uint32_t llama_kv_v_row_embd(
@@ -665,7 +665,7 @@ static inline uint32_t llama_kv_v_row_embd(
uint32_t il) {
// qwen3next recurrent state is stored in a dedicated V-cache tail (per sequence),
// so per-token V rows include only attention values.
if (model.arch == LLM_ARCH_QWEN3NEXT) {
if (model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) {
return hparams.n_embd_v_gqa(il);
}
@@ -724,7 +724,7 @@ static bool llama_kv_cache_init(
cache.recurrent = model.arch == LLM_ARCH_MAMBA;
// qwen3next uses hybrid recurrent+attention cache semantics. Keep V rows in
// standard layout to match the mainline hybrid path when flash attention is off.
cache.v_trans = !cache.recurrent && !cparams.flash_attn && model.arch != LLM_ARCH_QWEN3NEXT;
cache.v_trans = !cache.recurrent && !cparams.flash_attn && model.arch != LLM_ARCH_QWEN3NEXT && model.arch != LLM_ARCH_QWEN35MOE;
cache.head = 0;
cache.size = kv_size;
@@ -736,7 +736,7 @@ static bool llama_kv_cache_init(
cache.cells.clear();
cache.cells.resize(kv_size);
if (cache.recurrent || model.arch == LLM_ARCH_QWEN3NEXT) {
if (cache.recurrent || model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) {
// init state copy sources
for (uint32_t i = 0; i < cache.size; ++i) {
cache.cells[i].src = i;
@@ -821,7 +821,7 @@ static bool llama_kv_cache_init(
std::vector<size_t> mem_split(model.splits.size(), 0);
const uint32_t qnext_state_slots = llama_qwen3next_state_slots(cparams, kv_size);
if (model.arch == LLM_ARCH_QWEN3NEXT && qnext_state_slots < std::max<uint32_t>(1, cparams.n_seq_max)) {
if ((model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) && qnext_state_slots < std::max<uint32_t>(1, cparams.n_seq_max)) {
LLAMA_LOG_WARN("%s: reducing qwen3next state slots from %u to %u to fit KV cache size\n",
__func__, std::max<uint32_t>(1, cparams.n_seq_max), qnext_state_slots);
}
@@ -3144,7 +3144,7 @@ static int llama_decode_internal(
auto tim1 = ggml_time_us();
#endif
uint32_t n_tokens = std::min(n_ubatch, n_tokens_all - cur_token);
if (model.arch == LLM_ARCH_QWEN3NEXT &&
if ((model.arch == LLM_ARCH_QWEN3NEXT || model.arch == LLM_ARCH_QWEN35MOE) &&
n_tokens > 1 &&
batch_all.n_seq_id != nullptr &&
batch_all.seq_id != nullptr) {
@@ -5227,6 +5227,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
case LLM_ARCH_QWEN3VL:
case LLM_ARCH_QWEN3VLMOE:
case LLM_ARCH_QWEN35MOE:
return LLAMA_ROPE_TYPE_IMROPE;
// all model arches should be listed explicitly here