llama-cpp-sys-4 0.10.0

Low Level Bindings to llama.cpp
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149

#include "models.h"

void llama_model_maion_coder::load_arch_hparams(llama_model_loader & ml) {
    ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
    ml.get_key_or_arr(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp_arr, hparams.n_layer_all);
    ml.get_key(LLM_KV_INTERLEAVE_MOE_LAYER_STEP,         hparams.n_moe_layer_step);
    ml.get_key(LLM_KV_MOE_LATENT_SIZE,                   hparams.moe_latent_size);

    if (hparams.n_moe_layer_step == 0 || hparams.moe_latent_size == 0 || hparams.n_expert == 0) {
        throw std::runtime_error("maion-coder requires a positive MoE layer step, latent size and expert count");
    }

    ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa);
    uint32_t n_swa_layers = 0;
    ml.get_arr_n(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, n_swa_layers);
    if (hparams.n_swa == 0 || n_swa_layers != hparams.n_layer_all) {
        throw std::runtime_error("maion-coder requires a positive sliding window and a sliding window pattern for each layer");
    }
    ml.get_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl);
    hparams.rope_pattern = hparams.is_swa_impl;
    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;

    type = LLM_TYPE_UNKNOWN;
}

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

    const int64_t moe_n_embd = hparams.moe_latent_size;

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

    for (int i = 0; i < n_layer; ++i) {
        const bool is_moe_layer = (i + 1) % hparams.n_moe_layer_step == 0;
        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.attn_q_norm    = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), {n_embd_head_k}, 0);
        layer.attn_k_norm    = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM,    "weight", i), {n_embd_head_k}, 0);
        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);

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

        if (is_moe_layer) {
            const int64_t n_ff_exp = hparams.n_ff_exp(i);
            layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,    "weight", i), {n_embd, n_expert}, 0);
            layer.ffn_latent_down = create_tensor(tn(LLM_TENSOR_FFN_LATENT_DOWN, "weight", i), {n_embd, moe_n_embd}, 0);
            layer.ffn_gate_exps   = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS,   "weight", i), {moe_n_embd, n_ff_exp, n_expert}, 0);
            layer.ffn_up_exps     = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,     "weight", i), {moe_n_embd, n_ff_exp, n_expert}, 0);
            layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS,   "weight", i), {n_ff_exp, moe_n_embd, n_expert}, 0);
            layer.ffn_latent_up   = create_tensor(tn(LLM_TENSOR_FFN_LATENT_UP,   "weight", i), {moe_n_embd, n_embd}, 0);
        } else {
            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
        }
    }
}

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

llama_model_maion_coder::graph::graph(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 == n_rot);

    ggml_tensor * inpL = build_inp_embd(model.tok_embd);
    ggml_tensor * inp_pos = build_inp_pos();
    auto * inp_attn = build_attn_inp_kv_iswa();
    ggml_tensor * inp_out_ids = build_inp_out_ids();

    const float kq_scale = 1.0f / sqrtf(float(n_embd_head));

    for (int il = 0; il < n_layer; ++il) {
        // export per-layer input hidden state so speculative drafts (e.g. dflash/eagle3) can read it
        res->t_layer_inp[il] = inpL;

        ggml_tensor * inpSA = inpL;
        ggml_tensor * cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
        cb(cur, "attn_norm", il);

        auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur, n_embd_head, n_head, n_head_kv, il);
        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);

        // Full-attention layers do not use RoPE.
        if (hparams.is_swa(il)) {
            const float freq_base_l  = model.get_rope_freq_base(cparams, il);
            const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
            Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l, 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_l, freq_scale_l, ext_factor, attn_factor, beta_fast, beta_slow);
        }

        cur = build_attn(inp_attn, model.layers[il].wo, NULL, model.layers[il].wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, 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);
        }

        cur = build_norm(cur, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
        cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);

        if (model.layers[il].ffn_gate_inp == nullptr) {
            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);
        } else {
            ggml_tensor * router_logits = build_lora_mm(model.layers[il].ffn_gate_inp, cur);
            ggml_tensor * inp_latent = build_lora_mm(model.layers[il].ffn_latent_down, cur);
            cur = build_moe_ffn(inp_latent,
                    model.layers[il].ffn_gate_inp,
                    model.layers[il].ffn_up_exps,
                    model.layers[il].ffn_gate_exps,
                    model.layers[il].ffn_down_exps,
                    nullptr,
                    n_expert, hparams.n_expert_used(il),
                    LLM_FFN_SILU, true, 0.0f,
                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
                    il, router_logits);
            cur = build_lora_mm(model.layers[il].ffn_latent_up, cur);
        }

        cur = build_norm(cur, model.layers[il].ffn_post_norm, NULL, LLM_NORM_RMS, il);
        cur = ggml_add(ctx0, cur, ffn_inp);
        cur = build_cvec(cur, il);
        inpL = cur;
    }

    ggml_tensor * cur = build_norm(inpL, model.output_norm, NULL, LLM_NORM_RMS, -1);
    cb(cur, "result_norm", -1);
    res->t_embd = cur;
    cur = build_lora_mm(model.output, cur, model.output_s);
    cb(cur, "result_output", -1);
    res->t_logits = cur;
}