src/models/exaone-moe.cpp

#include "models.h"


llm_build_exaone_moe::llm_build_exaone_moe(const llama_model & model, const llm_graph_params & params) :
    llm_graph_context(params) {
    const int64_t n_embd_head = hparams.n_embd_head_k;

    GGML_ASSERT(n_embd_head == hparams.n_embd_head_v);
    GGML_ASSERT(n_embd_head == hparams.n_rot);

    ggml_tensor * cur;
    ggml_tensor * inpL;

    inpL = build_inp_embd(model.tok_embd);

    // inp_pos - contains the positions
    ggml_tensor * inp_pos = build_inp_pos();

    auto * inp_attn_iswa = build_attn_inp_kv_iswa();

    ggml_tensor * inp_out_ids = build_inp_out_ids();

    const int n_transformer_layers = n_layer - hparams.nextn_predict_layers;
    for (int il = 0; il < n_transformer_layers; ++il) {
        ggml_tensor * inpSA = inpL;

        // use RoPE for SWA layers
        const bool is_local_layer = hparams.is_swa(il);

        // norm
        cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
        cb(cur, "attn_norm", il);

        // self-attention
        {
            ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);

            // compute Q and K and RoPE them
            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
            cb(Qcur, "Qcur", il);

            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
            cb(Kcur, "Kcur", il);

            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
            cb(Vcur, "Vcur", il);

            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);

            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
            cb(Qcur, "Qcur_normed", il);
            cb(Kcur, "Kcur_normed", il);

            if (is_local_layer) {
                Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base,
                                     freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);

                Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base,
                                     freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
            }
            cb(Qcur, "Qcur", il);
            cb(Kcur, "Kcur", il);
            cb(Vcur, "Vcur", il);

            cur = build_attn(inp_attn_iswa,
                model.layers[il].wo, NULL,
                Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
            cb(cur, "attn_out", il);
        }
        if (il == n_transformer_layers - 1 && inp_out_ids) {
            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
        }
        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
        cb(ffn_inp, "ffn_inp", il);

        // norm
        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
        cb(cur, "ffn_norm", il);

        // feed-forward network
        if (model.layers[il].ffn_gate_inp == nullptr) {
            // dense branch
            cur = build_ffn(cur,
                    model.layers[il].ffn_up, NULL, NULL,
                    model.layers[il].ffn_gate, NULL, NULL,
                    model.layers[il].ffn_down, NULL, NULL, NULL,
                    LLM_FFN_SILU, LLM_FFN_PAR, il);
            cb(cur, "ffn_out", il);
        } else {
            // MoE branch
            ggml_tensor * moe_out = build_moe_ffn(cur,
                model.layers[il].ffn_gate_inp,
                model.layers[il].ffn_up_exps,
                model.layers[il].ffn_gate_exps,
                model.layers[il].ffn_down_exps,
                model.layers[il].ffn_exp_probs_b,
                n_expert, n_expert_used,
                LLM_FFN_SILU, hparams.expert_weights_norm,
                true, hparams.expert_weights_scale,
                (llama_expert_gating_func_type) hparams.expert_gating_func,
                il);
            cb(moe_out, "ffn_moe_out", il);

            // FFN shared expert
            {
                ggml_tensor * ffn_shexp =
                    build_ffn(cur,
                        model.layers[il].ffn_up_shexp, NULL, NULL,
                        model.layers[il].ffn_gate_shexp, NULL, NULL,
                        model.layers[il].ffn_down_shexp, NULL, NULL,
                        NULL, LLM_FFN_SILU, LLM_FFN_PAR, il);
                cb(ffn_shexp, "ffn_shexp", il);

                cur = ggml_add(ctx0, moe_out, ffn_shexp);
                cb(cur, "ffn_out", il);
            }
        }

        cur = ggml_add(ctx0, cur, ffn_inp);

        cur = build_cvec(cur, il);
        cb(cur, "l_out", il);

        // input for next layer
        inpL = cur;
    }
    cur = inpL;

    // final norm
    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);

    cb(cur, "result_norm", -1);
    res->t_embd = cur;

    // lm_head
    cur = build_lora_mm(model.output, cur);

    cb(cur, "result_output", -1);
    res->t_logits = cur;

    ggml_build_forward_expand(gf, cur);
}