echodict/llama.cpp

1#include "models.h"2 3 4llm_build_chatglm::llm_build_chatglm(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    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        cur = build_norm(inpL,25                model.layers[il].attn_norm,26                NULL,27                LLM_NORM_RMS, il);28        cb(cur, "attn_norm", il);29 30        // self-attention31        {32            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,33                    n_embd_head, n_head, n_head_kv, il);34 35            //printf("freq_base: %f freq_scale: %f ext_factor: %f attn_factor: %f\n", freq_base, freq_scale, ext_factor, attn_factor);36            Qcur = ggml_rope_ext(37                    ctx0, Qcur, inp_pos, nullptr,38                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,39                    ext_factor, attn_factor, beta_fast, beta_slow40                    );41 42            Kcur = ggml_rope_ext(43                    ctx0, Kcur, inp_pos, nullptr,44                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,45                    ext_factor, attn_factor, beta_fast, beta_slow46                    );47 48            cb(Qcur, "Qcur", il);49            cb(Kcur, "Kcur", il);50            cb(Vcur, "Vcur", il);51 52            cur = build_attn(inp_attn,53                    model.layers[il].wo, NULL, model.layers[il].wo_s,54                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);55        }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 62        // Add the input63        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);64        cb(ffn_inp, "ffn_inp", il);65 66        // FF67        {68            cur = build_norm(ffn_inp,69                    model.layers[il].ffn_norm,70                    NULL,71                    LLM_NORM_RMS, il);72            cb(cur, "ffn_norm", il);73 74            cur = build_ffn(cur,75                    model.layers[il].ffn_up,   NULL, NULL,76                    NULL,                      NULL, NULL,77                    model.layers[il].ffn_down, NULL, NULL,78                    NULL,79                    LLM_FFN_SWIGLU, LLM_FFN_SEQ, il);80            cb(cur, "ffn_out", il);81 82        }83 84        cur = ggml_add(ctx0, cur, ffn_inp);85 86        cur = build_cvec(cur, il);87        cb(cur, "l_out", il);88 89        // input for next layer90        inpL = cur;91    }92 93    cur = build_norm(inpL,94            model.output_norm,95            NULL,96            LLM_NORM_RMS, -1);97 98    cb(cur, "result_norm", -1);99    res->t_embd = cur;100 101    cur = build_lora_mm(model.output, cur);102 103    cb(cur, "result_output", -1);104    res->t_logits = cur;105 106    ggml_build_forward_expand(gf, cur);107}108