echodict/llama.cpp

1#include "models.h"2 3llm_build_bailingmoe2::llm_build_bailingmoe2(const llama_model & model, const llm_graph_params & params) :4    llm_graph_context(params) {5    const int64_t n_embd_head = hparams.n_embd_head_v();6 7    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());8 9    ggml_tensor * cur;10    ggml_tensor * inpL;11 12    inpL = build_inp_embd(model.tok_embd);13 14    // inp_pos - contains the positions15    ggml_tensor * inp_pos = build_inp_pos();16 17    auto * inp_attn = build_attn_inp_kv();18 19    ggml_tensor * inp_out_ids = build_inp_out_ids();20 21    const int n_transformer_layers = n_layer - hparams.nextn_predict_layers;22    for (int il = 0; il < n_transformer_layers; ++il) {23        ggml_tensor * inpSA = inpL;24 25        // norm26        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);27        cb(cur, "attn_norm", il);28 29        // self_attention30        {31            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,32                    n_embd_head, n_head, n_head_kv, il);33 34            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);35            cb(Qcur, "Qcur_normed", il);36 37            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,38                                 ext_factor, attn_factor, beta_fast, beta_slow);39 40            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);41            cb(Kcur, "Kcur_normed", il);42 43            Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,44                                 ext_factor, attn_factor, beta_fast, beta_slow);45 46            cb(Qcur, "Qcur", il);47            cb(Kcur, "Kcur", il);48            cb(Vcur, "Vcur", il);49 50            cur = build_attn(inp_attn,51                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,52                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);53        }54 55        if (il == n_transformer_layers - 1 && inp_out_ids) {56            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);57            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);58        }59 60        ggml_tensor * sa_out = ggml_add(ctx0, cur, inpSA);61        cb(sa_out, "sa_out", il);62 63        // MoE branch64        cur = build_norm(sa_out, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);65        cb(cur, "ffn_norm", il);66 67        if (static_cast<uint32_t>(il) < hparams.n_layer_dense_lead) {68            cur = build_ffn(cur,69                    model.layers[il].ffn_up, NULL, NULL,70                    model.layers[il].ffn_gate, NULL, NULL,71                    model.layers[il].ffn_down, NULL, NULL,72                    NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);73            cb(cur, "ffn_out", il);74        } else {75            ggml_tensor * moe_out = build_moe_ffn(cur,76                model.layers[il].ffn_gate_inp,77                model.layers[il].ffn_up_exps,78                model.layers[il].ffn_gate_exps,79                model.layers[il].ffn_down_exps,80                model.layers[il].ffn_exp_probs_b,81                n_expert, n_expert_used,82                LLM_FFN_SILU, hparams.expert_weights_norm,83                hparams.expert_weights_scale,84                (llama_expert_gating_func_type) hparams.expert_gating_func,85                il);86            cb(moe_out, "ffn_moe_out", il);87 88            {89                ggml_tensor * ffn_shexp =90                    build_ffn(cur,91                        model.layers[il].ffn_up_shexp, NULL, NULL,92                        model.layers[il].ffn_gate_shexp, NULL, NULL,93                        model.layers[il].ffn_down_shexp, NULL, NULL,94                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);95                cb(ffn_shexp, "ffn_shexp", il);96 97                cur = ggml_add(ctx0, moe_out, ffn_shexp);98                cb(cur, "ffn_out", il);99            }100        }101 102        cur = ggml_add(ctx0, cur, sa_out);103 104        cur = build_cvec(cur, il);105        cb(cur, "l_out", il);106 107        // input for next layer108        inpL = cur;109    }110 111    cur = inpL;112 113    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);114 115    cb(cur, "result_norm", -1);116    res->t_embd = cur;117 118    // lm_head119    cur = build_lora_mm(model.output, cur);120 121    cb(cur, "result_output", -1);122    res->t_logits = cur;123 124    ggml_build_forward_expand(gf, cur);125}126