Merge pull request #114 from arnej27959/arnej/remove-unused-code

remove unused and duplicated code

Author: kherud

src/main/cpp/jllama.cpp

@@ -452,16 +452,6 @@ JNIEXPORT void JNICALL Java_de_kherud_llama_LlamaModel_loadModel(JNIEnv *env, jo
         llama_init_dft.context.reset();
     }
 
-    ctx_server->chat_templates = common_chat_templates_init(ctx_server->model, params.chat_template);
-    try {
-        common_chat_format_example(ctx_server->chat_templates.get(), params.use_jinja);
-    } catch (const std::exception &e) {
-        SRV_WRN("%s: The chat template that comes with this model is not yet supported, falling back to chatml. This "
-                "may cause the model to output suboptimal responses\n",
-                __func__);
-        ctx_server->chat_templates = common_chat_templates_init(ctx_server->model, "chatml");
-    }
-
     // print sample chat example to make it clear which template is used
     LOG_INF("%s: chat template, chat_template: %s, example_format: '%s'\n", __func__,
             common_chat_templates_source(ctx_server->chat_templates.get()),
@@ -860,4 +850,4 @@ JNIEXPORT jbyteArray JNICALL Java_de_kherud_llama_LlamaModel_jsonSchemaToGrammar
     nlohmann::ordered_json c_schema_json = nlohmann::ordered_json::parse(c_schema);
     const std::string c_grammar = json_schema_to_grammar(c_schema_json);
     return parse_jbytes(env, c_grammar);
-}
+}

src/main/cpp/server.hpp

@@ -3269,151 +3269,3 @@ struct server_context {
         };
     }
 };
-
-static void common_params_handle_model_default(std::string &model, const std::string &model_url, std::string &hf_repo,
-                                               std::string &hf_file, const std::string &hf_token) {
-    if (!hf_repo.empty()) {
-        // short-hand to avoid specifying --hf-file -> default it to --model
-        if (hf_file.empty()) {
-            if (model.empty()) {
-                auto auto_detected = common_get_hf_file(hf_repo, hf_token);
-                if (auto_detected.first.empty() || auto_detected.second.empty()) {
-                    exit(1); // built without CURL, error message already printed
-                }
-                hf_repo = auto_detected.first;
-                hf_file = auto_detected.second;
-            } else {
-                hf_file = model;
-            }
-        }
-        // make sure model path is present (for caching purposes)
-        if (model.empty()) {
-            // this is to avoid different repo having same file name, or same file name in different subdirs
-            std::string filename = hf_repo + "_" + hf_file;
-            // to make sure we don't have any slashes in the filename
-            string_replace_all(filename, "/", "_");
-            model = fs_get_cache_file(filename);
-        }
-    } else if (!model_url.empty()) {
-        if (model.empty()) {
-            auto f = string_split<std::string>(model_url, '#').front();
-            f = string_split<std::string>(f, '?').front();
-            model = fs_get_cache_file(string_split<std::string>(f, '/').back());
-        }
-    } else if (model.empty()) {
-        model = DEFAULT_MODEL_PATH;
-    }
-}
-
-// parse the given jparams (see de.kherud.llama.args.ModelParameters#toString()) from JSON to the required C++ struct.
-static void server_params_parse(json jparams, common_params &params) {
-    common_params default_params;
-
-    params.sampling.seed = json_value(jparams, "seed", default_params.sampling.seed);
-    params.cpuparams.n_threads = json_value(jparams, "n_threads", default_params.cpuparams.n_threads);
-    params.speculative.cpuparams.n_threads =
-        json_value(jparams, "n_threads_draft", default_params.speculative.cpuparams.n_threads);
-    params.cpuparams_batch.n_threads = json_value(jparams, "n_threads_batch", default_params.cpuparams_batch.n_threads);
-    params.speculative.cpuparams_batch.n_threads =
-        json_value(jparams, "n_threads_batch_draft", default_params.speculative.cpuparams_batch.n_threads);
-    params.n_predict = json_value(jparams, "n_predict", default_params.n_predict);
-    params.n_ctx = json_value(jparams, "n_ctx", default_params.n_ctx);
-    params.n_batch = json_value(jparams, "n_batch", default_params.n_batch);
-    params.n_ubatch = json_value(jparams, "n_ubatch", default_params.n_ubatch);
-    params.n_keep = json_value(jparams, "n_keep", default_params.n_keep);
-
-    params.speculative.n_max = json_value(jparams, "n_draft", default_params.speculative.n_max);
-    params.speculative.n_min = json_value(jparams, "n_draft_min", default_params.speculative.n_min);
-
-    params.n_chunks = json_value(jparams, "n_chunks", default_params.n_chunks);
-    params.n_parallel = json_value(jparams, "n_parallel", default_params.n_parallel);
-    params.n_sequences = json_value(jparams, "n_sequences", default_params.n_sequences);
-    params.speculative.p_split = json_value(jparams, "p_split", default_params.speculative.p_split);
-    params.grp_attn_n = json_value(jparams, "grp_attn_n", default_params.grp_attn_n);
-    params.grp_attn_w = json_value(jparams, "grp_attn_w", default_params.grp_attn_w);
-    params.n_print = json_value(jparams, "n_print", default_params.n_print);
-    params.rope_freq_base = json_value(jparams, "rope_freq_base", default_params.rope_freq_base);
-    params.rope_freq_scale = json_value(jparams, "rope_freq_scale", default_params.rope_freq_scale);
-    params.yarn_ext_factor = json_value(jparams, "yarn_ext_factor", default_params.yarn_ext_factor);
-    params.yarn_attn_factor = json_value(jparams, "yarn_attn_factor", default_params.yarn_attn_factor);
-    params.yarn_beta_fast = json_value(jparams, "yarn_beta_fast", default_params.yarn_beta_fast);
-    params.yarn_beta_slow = json_value(jparams, "yarn_beta_slow", default_params.yarn_beta_slow);
-    params.yarn_orig_ctx = json_value(jparams, "yarn_orig_ctx", default_params.yarn_orig_ctx);
-    params.defrag_thold = json_value(jparams, "defrag_thold", default_params.defrag_thold);
-    params.numa = json_value(jparams, "numa", default_params.numa);
-    params.rope_scaling_type = json_value(jparams, "rope_scaling_type", default_params.rope_scaling_type);
-    params.pooling_type = json_value(jparams, "pooling_type", default_params.pooling_type);
-    params.model = json_value(jparams, "model", default_params.model);
-    params.speculative.model = json_value(jparams, "model_draft", default_params.speculative.model);
-    params.model_alias = json_value(jparams, "model_alias", default_params.model_alias);
-    params.model_url = json_value(jparams, "model_url", default_params.model_url);
-    params.hf_repo = json_value(jparams, "hf_repo", default_params.hf_repo);
-    params.hf_file = json_value(jparams, "hf_file", default_params.hf_file);
-    params.prompt = json_value(jparams, "prompt", default_params.prompt);
-    params.prompt_file = json_value(jparams, "prompt_file", default_params.prompt_file);
-    params.path_prompt_cache = json_value(jparams, "path_prompt_cache", default_params.path_prompt_cache);
-    params.input_prefix = json_value(jparams, "input_prefix", default_params.input_prefix);
-    params.input_suffix = json_value(jparams, "input_suffix", default_params.input_suffix);
-    params.antiprompt = json_value(jparams, "antiprompt", default_params.antiprompt);
-    params.lookup_cache_static = json_value(jparams, "lookup_cache_static", default_params.lookup_cache_static);
-    params.lookup_cache_dynamic = json_value(jparams, "lookup_cache_dynamic", default_params.lookup_cache_dynamic);
-    params.logits_file = json_value(jparams, "logits_file", default_params.logits_file);
-    // params.lora_adapters = json_value(jparams, "lora_adapter", default_params.lora_adapters);
-    params.embedding = json_value(jparams, "embedding", default_params.embedding);
-    params.escape = json_value(jparams, "escape", default_params.escape);
-    params.cont_batching = json_value(jparams, "cont_batching", default_params.cont_batching);
-    params.flash_attn = json_value(jparams, "flash_attn", default_params.flash_attn);
-    params.input_prefix_bos = json_value(jparams, "input_prefix_bos", default_params.input_prefix_bos);
-    params.sampling.ignore_eos = json_value(jparams, "ignore_eos", default_params.sampling.ignore_eos);
-    params.use_mmap = json_value(jparams, "use_mmap", default_params.use_mmap);
-    params.use_mlock = json_value(jparams, "use_mlock", default_params.use_mlock);
-    params.no_kv_offload = json_value(jparams, "no_kv_offload", default_params.no_kv_offload);
-    params.chat_template = json_value(jparams, "chat_template", default_params.chat_template);
-
-    if (jparams.contains("n_gpu_layers")) {
-        if (llama_supports_gpu_offload()) {
-            params.n_gpu_layers = json_value(jparams, "n_gpu_layers", default_params.n_gpu_layers);
-            params.speculative.n_gpu_layers =
-                json_value(jparams, "n_gpu_layers_draft", default_params.speculative.n_gpu_layers);
-        } else {
-            SRV_WRN("Not compiled with GPU offload support, --n-gpu-layers option will be ignored. "
-                    "See main README.md for information on enabling GPU BLAS support: %s = %d",
-                    "n_gpu_layers", params.n_gpu_layers);
-        }
-    }
-
-    if (jparams.contains("split_mode")) {
-        params.split_mode = json_value(jparams, "split_mode", default_params.split_mode);
-// todo: the definition checks here currently don't work due to cmake visibility reasons
-#ifndef GGML_USE_CUDA
-        fprintf(stderr, "warning: llama.cpp was compiled without CUDA. Setting the split mode has no effect.\n");
-#endif
-    }
-
-    if (jparams.contains("tensor_split")) {
-#if defined(GGML_USE_CUDA) || defined(GGML_USE_SYCL)
-        std::vector<float> tensor_split = jparams["tensor_split"].get<std::vector<float>>();
-        GGML_ASSERT(tensor_split.size() <= llama_max_devices());
-
-        for (size_t i_device = 0; i_device < llama_max_devices(); ++i_device) {
-            if (i_device < tensor_split.size()) {
-                params.tensor_split[i_device] = tensor_split.at(i_device);
-            } else {
-                params.tensor_split[i_device] = 0.0f;
-            }
-        }
-#else
-        SRV_WRN("%s", "llama.cpp was compiled without CUDA. It is not possible to set a tensor split.\n");
-#endif // GGML_USE_CUDA
-    }
-
-    if (jparams.contains("main_gpu")) {
-#if defined(GGML_USE_CUDA) || defined(GGML_USE_SYCL)
-        params.main_gpu = json_value(jparams, "main_gpu", default_params.main_gpu);
-#else
-        SRV_WRN("%s", "llama.cpp was compiled without CUDA. It is not possible to set a main GPU.");
-#endif
-    }
-
-    common_params_handle_model_default(params.model, params.model_url, params.hf_repo, params.hf_file, params.hf_token);
-}