char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

PikaPikachuconvert : add `--fuse-qkv` flag to fuse Q/K/V into QKV during HF-to-GGUF conversion (#22780)465e49b9c2026-09-07

#include "models.h"

void llama_model_llada::load_arch_hparams(llama_model_loader & ml) {
    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);

    // LLaDA-8B has 32 layers, similar to LLaMA but for diffusion
    switch (hparams.n_layer()) {
        case 32:
            type = LLM_TYPE_8B;
            break;
        default:
            type = LLM_TYPE_UNKNOWN;
    }

    // Set non-causal attention for diffusion models
    hparams.causal_attn = false;
}

void llama_model_llada::load_arch_tensors(llama_model_loader &) {
    LLAMA_LOAD_LOCALS;

    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);

    // output
    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED);

    // if output is NULL, init from the input tok embed
    if (output == NULL) {
        output =
            create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
    }

    for (int i = 0; i < n_layer; ++i) {
        auto & layer = layers[i];

        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);

        create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
        layer.wo =
            create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
        layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), { n_embd }, TENSOR_NOT_REQUIRED);

        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), { n_embd }, 0);

        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), { n_rot / 2 },
                                         TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));

        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, 0);
        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, n_ff }, 0);

        // optional MLP bias
        layer.ffn_gate_b =
            create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), { n_ff }, TENSOR_NOT_REQUIRED);
        layer.ffn_down_b =
            create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), { n_embd }, TENSOR_NOT_REQUIRED);
        layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), { n_ff }, TENSOR_NOT_REQUIRED);
    }
}

std::unique_ptr<llm_graph_context> llama_model_llada::build_arch_graph(const llm_graph_params & params) const {
    return std::make_unique<graph>(*this, params);
}

llama_model_llada::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
    // LLaDA is similar to LLaMA but uses non-causal attention for diffusion
    const int64_t n_embd_head = hparams.n_embd_head_v();

    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
    GGML_ASSERT(n_embd_head == 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();

    // Non-causal attention for diffusion
    auto * inp_attn = build_attn_inp_no_cache();

    ggml_tensor * inp_out_ids = build_inp_out_ids();

    for (int il = 0; il < n_layer; ++il) {
        ggml_tensor * inpSA = inpL;

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

        // self-attention
        {
            // compute separate Q, K, V projections without bias, matching LLaDALlamaBlock
            auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur,
                    n_embd_head, n_head, n_head_kv, il);

            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, 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, nullptr, 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,
                    model.layers[il].wo, NULL, model.layers[il].wo_s,
                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il);
        }
        if (il == n_layer - 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);

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

        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);

        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;

    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, model.output_s);

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

    ggml_build_forward_expand(gf, cur);
}