slm_ikllama_sys 0.1.1

ik_llama.cpp rust sys bindings
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

#include "../llama-build-context.h"
#include "../llama-model.h"
#include "../llama-context.h"

ggml_cgraph * llm_build_context::build_seedoss() {
    ggml_cgraph * gf = new_graph_custom();

    const int64_t n_embd_head = hparams.n_embd_head_v(0);
    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k(0));
    GGML_ASSERT(n_embd_head == hparams.n_rot);

    ggml_tensor * cur;
    ggml_tensor * inpL;

    inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);

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

    // KQ_mask (mask for 1 head, it will be broadcasted to all heads)
    struct ggml_tensor * KQ_mask = build_inp_KQ_mask();

    const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale;

    for (int il = 0; il < n_layer; ++il) {

        cur = build_std_attention(gf, model.layers[il].attn_norm, inpL,
                inp_pos, il == n_layer-1 && n_tokens > 1 ? inp_out_ids : nullptr, nullptr,
                KQ_mask, nullptr, nullptr, kq_scale, 0.0f, 0, il, true, false, true);

        cur = llm_build_ffn(ctx0, lctx, model.layers[il].attn_post_norm, 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, cb, il, gf, true);
        cb(cur, "ffn_out", il);

        cur = lctx.cvec.apply_to(ctx0, cur, il);
        cb(cur, "l_out", il);

        // input for next layer
        inpL = cur;
    }

    cur = build_output(lctx, ctx0, inpL, model.output, model.output_norm, cb);
    cb(cur, "result_output", -1);

    ggml_build_forward_expand(gf, cur);

    return gf;
}