echodict/llama.cpp

1#include "models.h"2 3llm_build_nemotron::llm_build_nemotron(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {4    const int64_t n_embd_head = hparams.n_embd_head_v();5 6    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());7    //GGML_ASSERT(n_embd_head == n_rot);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    for (int il = 0; il < n_layer; ++il) {22        ggml_tensor * inpSA = inpL;23 24        // norm25        cur = build_norm(inpL,26                model.layers[il].attn_norm,27                model.layers[il].attn_norm_b,28                LLM_NORM, il);29        cb(cur, "attn_norm", il);30 31        // self-attention32        {33            // compute Q and K and RoPE them34            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,35                    n_embd_head, n_head, n_head_kv, il);36 37            Qcur = ggml_rope_ext(38                    ctx0, Qcur, inp_pos, nullptr,39                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,40                    ext_factor, attn_factor, beta_fast, beta_slow41                    );42 43            Kcur = ggml_rope_ext(44                    ctx0, Kcur, inp_pos, nullptr,45                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,46                    ext_factor, attn_factor, beta_fast, beta_slow47                    );48 49            cb(Qcur, "Qcur", il);50            cb(Kcur, "Kcur", il);51            cb(Vcur, "Vcur", il);52 53            cur = build_attn(inp_attn,54                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,55                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);56        }57        if (il == n_layer - 1 && inp_out_ids) {58            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);59            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);60        }61        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);62        cb(ffn_inp, "ffn_inp", il);63 64        // feed-forward network65        cur = build_norm(ffn_inp,66                model.layers[il].ffn_norm,67                model.layers[il].ffn_norm_b,68                LLM_NORM, il);69        cb(cur, "ffn_norm", il);70 71        cur = build_ffn(cur,72                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,73                NULL,                      NULL,                        NULL,74                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,75                NULL,76                LLM_FFN_RELU_SQR, LLM_FFN_SEQ, il);77 78        cur = ggml_add(ctx0, cur, ffn_inp);79        cb(cur, "ffn_out", il);80 81        cur = build_cvec(cur, il);82        cb(cur, "l_out", il);83 84        // input for next layer85        inpL = cur;86    }87    cur = inpL;88 89    cur = build_norm(cur,90            model.output_norm, model.output_norm_b,91            LLM_NORM, -1);92 93    cb(cur, "result_norm", -1);94    res->t_embd = cur;95 96    // lm_head97    cur = build_lora_mm(model.output, cur);98 99    cb(cur, "result_output", -1);100    res->t_logits = cur;101 102    ggml_build_forward_expand(gf, cur);103}104