echodict/llama.cpp

1#include "models.h"2 3llm_build_mimo2_iswa::llm_build_mimo2_iswa(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {4    ggml_tensor * cur;5    ggml_tensor * inpL;6 7    inpL = build_inp_embd(model.tok_embd);8 9    ggml_tensor * inp_pos = build_inp_pos();10    auto * inp_attn = build_attn_inp_kv_iswa();11    ggml_tensor * inp_out_ids = build_inp_out_ids();12 13    for (int il = 0; il < n_layer; ++il) {14        ggml_tensor * inpSA = inpL;15 16        uint32_t n_head_l    = hparams.n_head(il);17        uint32_t n_head_kv_l = hparams.n_head_kv(il);18        const float freq_base_l  = model.get_rope_freq_base(cparams, il);19        const float freq_scale_l = model.get_rope_freq_scale(cparams, il);20 21        cur = inpL;22 23        // self_attention24        {25            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);26            cb(cur, "attn_norm", il);27 28            // compute Q and K and RoPE them29            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);30            cb(Qcur, "Qcur", il);31 32            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);33            cb(Kcur, "Kcur", il);34 35            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);36            cb(Vcur, "Vcur", il);37 38            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head_k, n_head_l,    n_tokens);39            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head_k, n_head_kv_l, n_tokens);40            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head_v, n_head_kv_l, n_tokens);41 42            Qcur = ggml_rope_ext(43                ctx0, Qcur, inp_pos, nullptr,44                n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,45                ext_factor, attn_factor, beta_fast, beta_slow46                );47 48            Kcur = ggml_rope_ext(49                ctx0, Kcur, inp_pos, nullptr,50                n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,51                ext_factor, attn_factor, beta_fast, beta_slow52                );53 54            cb(Qcur, "Qcur", il);55            cb(Kcur, "Kcur", il);56            cb(Vcur, "Vcur", il);57 58            ggml_tensor * sinks = model.layers[il].attn_sinks;59 60            cur = build_attn(inp_attn,61                    model.layers[il].wo, NULL, model.layers[il].wo_s,62                    Qcur, Kcur, Vcur, nullptr, sinks, nullptr, 1.0f/sqrtf(float(n_embd_head_k)), il);63        }64 65        if (il == n_layer - 1 && inp_out_ids) {66            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);67            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);68        }69 70        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);71        cb(ffn_inp, "ffn_inp", il);72 73        cur = build_norm(ffn_inp,74                model.layers[il].ffn_norm, NULL,75                LLM_NORM_RMS, il);76        cb(cur, "ffn_norm", il);77 78        // feed-forward network79        if (model.layers[il].ffn_gate_inp == nullptr) {80            // dense branch81            cur = build_ffn(cur,82                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,83                    model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,84                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,85                    NULL,86                    LLM_FFN_SILU, LLM_FFN_PAR, il);87            cb(cur, "ffn_out", il);88        } else {89            // MoE branch90            cur = build_moe_ffn(cur,91                    model.layers[il].ffn_gate_inp,92                    model.layers[il].ffn_up_exps,93                    model.layers[il].ffn_gate_exps,94                    model.layers[il].ffn_down_exps,95                    model.layers[il].ffn_exp_probs_b,96                    n_expert, n_expert_used,97                    LLM_FFN_SILU, true,98                    hparams.expert_weights_scale,99                    LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID,100                    il);101            cb(cur, "ffn_moe_out", il);102        }103 104        cur = ggml_add(ctx0, cur, ffn_inp);105 106        cur = build_cvec(cur, il);107        cb(cur, "l_out", il);108 109        // input for next layer110        inpL = cur;111    }112 113    cur = inpL;114 115    cur = build_norm(cur,116            model.output_norm, NULL,117            LLM_NORM_RMS, -1);118 119    cb(cur, "result_norm", -1);120    res->t_embd = cur;121 122    // lm_head123    cur = build_lora_mm(model.output, cur);124 125    cb(cur, "result_output", -1);126    res->t_logits = cur;127 128    ggml_build_forward_expand(gf, cur);129}130