echodict/llama.cpp

1#include "models.h"2 3template<bool iswa>4llm_build_phi3<iswa>::llm_build_phi3(const llama_model & model, const llm_graph_params & params) : 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    using inp_attn_type = std::conditional_t<iswa, llm_graph_input_attn_kv_iswa, llm_graph_input_attn_kv>;18    inp_attn_type * inp_attn = nullptr;19 20    if constexpr (iswa) {21        inp_attn = build_attn_inp_kv_iswa();22    } else {23        inp_attn = build_attn_inp_kv();24    }25    ggml_tensor * inp_out_ids = build_inp_out_ids();26 27    for (int il = 0; il < n_layer; ++il) {28        auto * residual = inpL;29 30        // self-attention31        {32            // rope freq factors for 128k context33            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);34 35            ggml_tensor* attn_norm_output = build_norm(inpL,36                    model.layers[il].attn_norm,37                    model.layers[il].attn_norm_b,38                    LLM_NORM_RMS, il);39            cb(attn_norm_output, "attn_norm", il);40 41            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], attn_norm_output,42                    n_embd_head, n_head, n_head_kv, il);43            Qcur = ggml_rope_ext(44                    ctx0, Qcur, inp_pos, rope_factors,45                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,46                    ext_factor, attn_factor, beta_fast, beta_slow47                    );48 49            Kcur = ggml_rope_ext(50                    ctx0, Kcur, inp_pos, rope_factors,51                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,52                    ext_factor, attn_factor, beta_fast, beta_slow53                    );54 55            cb(Qcur, "Qcur", il);56            cb(Kcur, "Kcur", il);57            cb(Vcur, "Vcur", il);58 59            Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd_head)));60            cb(Qcur, "Qcur", il);61 62            cur = build_attn(inp_attn,63                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,64                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f, il);65        }66        if (il == n_layer - 1 && inp_out_ids) {67            cur      = ggml_get_rows(ctx0, cur,      inp_out_ids);68            residual = ggml_get_rows(ctx0, residual, inp_out_ids);69        }70        cur = ggml_add(ctx0, cur, residual);71        residual = cur;72 73        cur = build_norm(cur,74                model.layers[il].ffn_norm, model.layers[il].ffn_norm_b,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            cur = build_ffn(cur,81                    model.layers[il].ffn_up,   NULL, NULL,82                    NULL,                      NULL, NULL,83                    model.layers[il].ffn_down, NULL, NULL,84                    NULL,85                    LLM_FFN_SWIGLU, LLM_FFN_SEQ, il);86            cb(cur, "ffn_out", il);87        } else {88            // MoE branch89            cur = build_moe_ffn(cur,90                    model.layers[il].ffn_gate_inp,91                    model.layers[il].ffn_up_exps,92                    model.layers[il].ffn_gate_exps,93                    model.layers[il].ffn_down_exps,94                    nullptr,95                    n_expert, n_expert_used,96                    LLM_FFN_SILU, true,97                    hparams.expert_weights_scale,98                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,99                    il);100            cb(cur, "ffn_moe_out", il);101        }102        cur = ggml_add(ctx0, residual, cur);103 104        cur = build_cvec(cur, il);105        cb(cur, "l_out", il);106 107        // input for next layer108        inpL = cur;109    }110    cur = build_norm(inpL,111            model.output_norm,112            model.output_norm_b,113            LLM_NORM_RMS, -1);114 115    cb(cur, "result_norm", -1);116    res->t_embd = cur;117 118    cur = build_lora_mm(model.output, cur);119 120    if (model.output_b != nullptr) {121        cb(cur, "result_output_no_bias", -1);122        cur = ggml_add(ctx0, cur, model.output_b);123    }124    cb(cur, "result_output", -1);125    res->t_logits = cur;126 127    ggml_build_forward_expand(gf, cur);128}129 130// Explicit template instantiations131template struct llm_build_phi3<false>;132template struct llm_build_phi3<true>;133