echodict/llama.cpp

1#include "models.h"2 3llm_build_seed_oss::llm_build_seed_oss(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    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;20 21    ggml_tensor * inp_out_ids = build_inp_out_ids();22 23    for (int il = 0; il < n_layer; ++il) {24        ggml_tensor * inpSA = inpL;25 26        // norm27        cur = build_norm(inpL,28                model.layers[il].attn_norm, NULL,29                LLM_NORM_RMS, il);30        cb(cur, "attn_norm", il);31 32        // self-attention33        {34            // compute Q and K and RoPE them35            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,36                    n_embd_head, n_head, n_head_kv, il);37 38            Qcur = ggml_rope_ext(39                    ctx0, Qcur, inp_pos, nullptr,40                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,41                    ext_factor, attn_factor, beta_fast, beta_slow42                    );43 44            Kcur = ggml_rope_ext(45                    ctx0, Kcur, inp_pos, nullptr,46                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,47                    ext_factor, attn_factor, beta_fast, beta_slow48                    );49 50            cb(Qcur, "Qcur", il);51            cb(Kcur, "Kcur", il);52            cb(Vcur, "Vcur", il);53 54            cur = build_attn(inp_attn,55                    model.layers[il].wo, model.layers[il].wo_b, model.layers[il].wo_s,56                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);57            cb(cur, "attn_out", il);58        }59        if (il == n_layer - 1 && inp_out_ids) {60            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);61            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);62        }63        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);64        cb(ffn_inp, "ffn_inp", il);65 66        // feed-forward network67        cur = build_norm(ffn_inp,68                model.layers[il].attn_post_norm, NULL,69                LLM_NORM_RMS, il);70        cb(cur, "attn_post_norm", il);71 72        cur = build_ffn(cur,73                model.layers[il].ffn_up,   NULL, NULL,74                model.layers[il].ffn_gate, NULL, NULL,75                model.layers[il].ffn_down, NULL, NULL,76                NULL,77                LLM_FFN_SILU, LLM_FFN_PAR, il);78        cb(cur, "ffn_out", il);79 80        cur = ggml_add(ctx0, cur, ffn_inp);81        cb(cur, "ffn_out", il);82 83        cur = build_cvec(cur, il);84        cb(cur, "l_out", il);85 86        // input for next layer87        inpL = cur;88    }89    cur = inpL;90 91    cur = build_norm(cur,92            model.output_norm, NULL,93            LLM_NORM_RMS, -1);94 95    cb(cur, "result_norm", -1);96    res->t_embd = cur;97 98    // lm_head99    cur = build_lora_mm(model.output, cur);100 101    cb(cur, "result_output", -1);102    res->t_logits = cur;103 104    ggml_build_forward_expand(gf, cur);105}106