Function ggml_flash_ff

ggml.c:5616–5644

// ggml_flash_ff

struct ggml_tensor * ggml_flash_ff(
        struct ggml_context * ctx,
        struct ggml_tensor  * a,
        struct ggml_tensor  * b0,
        struct ggml_tensor  * b1,
        struct ggml_tensor  * c0,
        struct ggml_tensor  * c1) {
    GGML_ASSERT(ggml_can_mul_mat(b0, a));
    // TODO: more checks

    bool is_node = false;

    if (a->grad || b0->grad || b1->grad || c0->grad || c1->grad) {
        is_node = true;
    }

    //struct ggml_tensor * result = ggml_dup_tensor(ctx, a);
    struct ggml_tensor * result = ggml_new_tensor(ctx, GGML_TYPE_F32, a->n_dims, a->ne);

    result->op   = GGML_OP_FLASH_FF;
    result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
    result->src[0] = a;
    result->src[1] = b0;
    result->src[2] = b1;
    result->src[3] = c0;
    result->src[4] = c1;

    return result;
}

// ggml_flash_attn_back