Function llama_new_context_with_model

llama.cpp:9833–10034

}

struct llama_context * llama_new_context_with_model(
                 struct llama_model * model,
        struct llama_context_params   params) {

    if (!model) {
        return nullptr;
    }

    llama_context * ctx = new llama_context(*model);

    const auto & hparams = model->hparams;
    auto       & cparams = ctx->cparams;

    cparams.n_batch          = params.n_batch;
    cparams.n_threads        = params.n_threads;
    cparams.n_threads_batch  = params.n_threads_batch;
    cparams.yarn_ext_factor  = params.yarn_ext_factor;
    cparams.yarn_attn_factor = params.yarn_attn_factor;
    cparams.yarn_beta_fast   = params.yarn_beta_fast;
    cparams.yarn_beta_slow   = params.yarn_beta_slow;
    cparams.mul_mat_q        = params.mul_mat_q;

    cparams.n_ctx            = params.n_ctx           == 0    ? hparams.n_ctx_train           : params.n_ctx;
    cparams.rope_freq_base   = params.rope_freq_base  == 0.0f ? hparams.rope_freq_base_train  : params.rope_freq_base;
    cparams.rope_freq_scale  = params.rope_freq_scale == 0.0f ? hparams.rope_freq_scale_train : params.rope_freq_scale;

    cparams.n_yarn_orig_ctx  = params.yarn_orig_ctx    != 0 ? params.yarn_orig_ctx    :
                               hparams.n_yarn_orig_ctx != 0 ? hparams.n_yarn_orig_ctx :
                                                              hparams.n_ctx_train;

    auto rope_scaling_type = params.rope_scaling_type;
    if (rope_scaling_type == LLAMA_ROPE_SCALING_UNSPECIFIED) {
        rope_scaling_type = hparams.rope_scaling_type_train;
    }

    if (rope_scaling_type == LLAMA_ROPE_SCALING_NONE) {
        cparams.rope_freq_scale = 1.0f; // never scale if scaling type is none
    }

    if (cparams.yarn_ext_factor < 0.0f) { // negative indicates 'not set'
        cparams.yarn_ext_factor = rope_scaling_type == LLAMA_ROPE_SCALING_YARN ? 1.0f : 0.0f;
    }

    if (params.seed == LLAMA_DEFAULT_SEED) {
        params.seed = time(NULL);
    }

    LLAMA_LOG_INFO("%s: n_ctx      = %u\n",     __func__, cparams.n_ctx);
    LLAMA_LOG_INFO("%s: freq_base  = %.1f\n",   __func__, cparams.rope_freq_base);
    LLAMA_LOG_INFO("%s: freq_scale = %g\n",     __func__, cparams.rope_freq_scale);

    ctx->rng = std::mt19937(params.seed);
    ctx->logits_all = params.logits_all;

    ggml_type memory_type = params.f16_kv ? GGML_TYPE_F16 : GGML_TYPE_F32;

    // reserve memory for context buffers
    if (!hparams.vocab_only) {