mirror of
https://github.com/ollama/ollama.git
synced 2026-07-10 17:54:12 +00:00
52196f1a97
Bump llama.cpp to b9493 and refresh the Laguna compat patch for upstream enum/tokenizer movement and the renamed SWA layer bitmap field.
232 lines
10 KiB
C++
Vendored
232 lines
10 KiB
C++
Vendored
#include "models/models.h"
|
|
|
|
void llama_model_laguna::load_arch_hparams(llama_model_loader & ml) {
|
|
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
|
|
|
|
// MoE
|
|
ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead, false);
|
|
ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp);
|
|
ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false);
|
|
ml.get_key(LLM_KV_EXPERT_SHARED_COUNT, hparams.n_expert_shared, false);
|
|
ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false);
|
|
ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false);
|
|
ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false);
|
|
|
|
ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false);
|
|
hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
|
|
ml.get_key_or_arr("laguna.attention.layer_types", hparams.is_swa_impl, hparams.n_layer, false);
|
|
|
|
ml.get_key("laguna.rope.swa.dimension_count", hparams.n_rot_swa, false);
|
|
ml.get_key("laguna.rope.swa.freq_base", hparams.rope_freq_base_train_swa, false);
|
|
ml.get_key("laguna.rope.scaling.beta_fast", hparams.yarn_beta_fast, false);
|
|
ml.get_key("laguna.rope.scaling.beta_slow", hparams.yarn_beta_slow, false);
|
|
|
|
type = LLM_TYPE_UNKNOWN;
|
|
}
|
|
|
|
void llama_model_laguna::load_arch_tensors(llama_model_loader &) {
|
|
LLAMA_LOAD_LOCALS;
|
|
|
|
const int64_t n_ff_exp = hparams.n_ff_exp;
|
|
const int64_t n_ff_shexp = hparams.n_ff_shexp;
|
|
|
|
tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
|
|
|
|
output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
|
|
output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
|
|
if (output == NULL) {
|
|
output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
|
|
}
|
|
|
|
for (int i = 0; i < n_layer; ++i) {
|
|
auto & layer = layers[i];
|
|
|
|
const int64_t n_head_i = hparams.n_head(i);
|
|
const int64_t n_head_kv_i = hparams.n_head_kv(i);
|
|
const int64_t n_embd_q = n_embd_head_k * n_head_i;
|
|
const int64_t n_embd_kv = n_embd_head_k * n_head_kv_i;
|
|
|
|
layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
|
|
|
|
layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_q}, 0);
|
|
layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_kv}, 0);
|
|
layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_kv}, 0);
|
|
layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_q, n_embd}, 0);
|
|
|
|
layer.wqkv_gate = create_tensor(tn(LLM_TENSOR_ATTN_GATE_LAGUNA, "weight", i), {n_embd, n_head_i}, 0);
|
|
|
|
layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
|
|
layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
|
|
|
|
layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
|
|
|
|
if (i < (int) hparams.n_layer_dense_lead) {
|
|
layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
|
|
layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd}, 0);
|
|
layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
|
|
} else {
|
|
layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
|
|
layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
|
|
|
|
layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert}, 0);
|
|
layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert}, 0);
|
|
layer.ffn_up_exps = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert}, 0);
|
|
|
|
layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0);
|
|
layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0);
|
|
layer.ffn_up_shexp = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0);
|
|
}
|
|
}
|
|
}
|
|
|
|
std::unique_ptr<llm_graph_context> llama_model_laguna::build_arch_graph(const llm_graph_params & params) const {
|
|
return std::make_unique<graph>(*this, params);
|
|
}
|
|
|
|
llama_model_laguna::graph::graph(const llama_model & model, const llm_graph_params & params) :
|
|
llm_graph_context(params) {
|
|
const int64_t n_embd_head = hparams.n_embd_head_v();
|
|
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
|
|
|
|
const float kq_scale = 1.0f / sqrtf(float(n_embd_head));
|
|
|
|
ggml_tensor * cur;
|
|
ggml_tensor * inpL;
|
|
|
|
inpL = build_inp_embd(model.tok_embd);
|
|
|
|
ggml_tensor * inp_pos = build_inp_pos();
|
|
|
|
auto * inp_attn = build_attn_inp_kv_iswa();
|
|
|
|
ggml_tensor * inp_out_ids = build_inp_out_ids();
|
|
|
|
for (int il = 0; il < n_layer; ++il) {
|
|
ggml_tensor * inpSA = inpL;
|
|
|
|
const int64_t n_head_il = hparams.n_head(il);
|
|
const int64_t n_head_kv_il = hparams.n_head_kv(il);
|
|
const bool is_swa = hparams.is_swa(il);
|
|
|
|
const int rope_n_dims = hparams.n_rot(il);
|
|
const float rope_base = is_swa ? hparams.rope_freq_base_train_swa : hparams.rope_freq_base_train;
|
|
const float rope_scale = is_swa ? hparams.rope_freq_scale_train_swa : hparams.rope_freq_scale_train;
|
|
const float rope_ext = is_swa ? 0.0f : 1.0f;
|
|
const float rope_bfast = hparams.yarn_beta_fast;
|
|
const float rope_bslow = hparams.yarn_beta_slow;
|
|
|
|
cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
|
|
cb(cur, "attn_norm", il);
|
|
|
|
// self-attention
|
|
{
|
|
ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
|
|
cb(Qcur, "Qcur", il);
|
|
ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
|
|
cb(Kcur, "Kcur", il);
|
|
ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
|
|
cb(Vcur, "Vcur", il);
|
|
|
|
ggml_tensor * gate = build_lora_mm(model.layers[il].wqkv_gate, cur);
|
|
cb(gate, "gate", il);
|
|
|
|
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head_il, n_tokens);
|
|
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv_il, n_tokens);
|
|
Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv_il, n_tokens);
|
|
|
|
Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
|
|
cb(Qcur, "Qcur_normed", il);
|
|
Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
|
|
cb(Kcur, "Kcur_normed", il);
|
|
|
|
Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr,
|
|
rope_n_dims, rope_type, hparams.n_ctx_orig_yarn, rope_base, rope_scale,
|
|
rope_ext, hparams.rope_attn_factor, rope_bfast, rope_bslow);
|
|
Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr,
|
|
rope_n_dims, rope_type, hparams.n_ctx_orig_yarn, rope_base, rope_scale,
|
|
rope_ext, hparams.rope_attn_factor, rope_bfast, rope_bslow);
|
|
|
|
cb(Qcur, "Qcur", il);
|
|
cb(Kcur, "Kcur", il);
|
|
cb(Vcur, "Vcur", il);
|
|
|
|
cur = build_attn(inp_attn,
|
|
nullptr, nullptr, nullptr,
|
|
Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
|
|
cb(cur, "attn_pregate", il);
|
|
|
|
gate = ggml_softplus(ctx0, gate);
|
|
cur = ggml_reshape_3d(ctx0, cur, n_embd_head, n_head_il, n_tokens);
|
|
gate = ggml_reshape_3d(ctx0, gate, 1, n_head_il, n_tokens);
|
|
cur = ggml_mul(ctx0, cur, gate);
|
|
cur = ggml_reshape_2d(ctx0, cur, n_embd_head * n_head_il, n_tokens);
|
|
cb(cur, "attn_gated", il);
|
|
|
|
cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
|
|
cb(cur, "attn_out", 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);
|
|
}
|
|
|
|
ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
|
|
cb(ffn_inp, "ffn_inp", il);
|
|
|
|
// feed-forward
|
|
cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
|
|
cb(cur, "ffn_norm", il);
|
|
|
|
if ((uint32_t) il < hparams.n_layer_dense_lead) {
|
|
cur = build_ffn(cur,
|
|
model.layers[il].ffn_up, NULL, NULL,
|
|
model.layers[il].ffn_gate, NULL, NULL,
|
|
model.layers[il].ffn_down, NULL, NULL,
|
|
NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
|
|
cb(cur, "ffn_out", il);
|
|
} else {
|
|
ggml_tensor * moe_out = build_moe_ffn(cur,
|
|
model.layers[il].ffn_gate_inp,
|
|
model.layers[il].ffn_up_exps,
|
|
model.layers[il].ffn_gate_exps,
|
|
model.layers[il].ffn_down_exps,
|
|
model.layers[il].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);
|
|
cb(moe_out, "ffn_moe_out", il);
|
|
|
|
ggml_tensor * ffn_shexp = build_ffn(cur,
|
|
model.layers[il].ffn_up_shexp, NULL, NULL,
|
|
model.layers[il].ffn_gate_shexp, NULL, NULL,
|
|
model.layers[il].ffn_down_shexp, NULL, NULL,
|
|
NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
|
|
cb(ffn_shexp, "ffn_shexp", il);
|
|
|
|
cur = ggml_add(ctx0, moe_out, ffn_shexp);
|
|
cb(cur, "ffn_out", il);
|
|
}
|
|
|
|
cur = ggml_add(ctx0, cur, ffn_inp);
|
|
|
|
cur = build_cvec(cur, il);
|
|
cb(cur, "l_out", il);
|
|
|
|
inpL = cur;
|
|
}
|
|
|
|
cur = inpL;
|
|
|
|
cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
|
|
cb(cur, "result_norm", -1);
|
|
res->t_embd = cur;
|
|
|
|
cur = build_lora_mm(model.output, cur, model.output_s);
|
|
cb(cur, "result_output", -1);
|
|
res->t_logits = cur;
|
|
|
|
ggml_build_forward_expand(gf, cur);
|
|
}
|