@fugood/llama.node 1.4.11 → 1.4.13

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@fugood/llama.node",
   "access": "public",
-  "version": "1.4.11",
+  "version": "1.4.13",
   "description": "An another Node binding of llama.cpp",
   "main": "lib/index.js",
   "scripts": {
@@ -72,20 +72,20 @@
     "CMakeLists.txt"
   ],
   "optionalDependencies": {
-    "@fugood/node-llama-darwin-arm64": "1.4.11",
-    "@fugood/node-llama-darwin-x64": "1.4.11",
-    "@fugood/node-llama-linux-arm64": "1.4.11",
-    "@fugood/node-llama-linux-arm64-cuda": "1.4.11",
-    "@fugood/node-llama-linux-arm64-snapdragon": "1.4.11",
-    "@fugood/node-llama-linux-arm64-vulkan": "1.4.11",
-    "@fugood/node-llama-linux-x64": "1.4.11",
-    "@fugood/node-llama-linux-x64-cuda": "1.4.11",
-    "@fugood/node-llama-linux-x64-vulkan": "1.4.11",
-    "@fugood/node-llama-win32-arm64": "1.4.11",
-    "@fugood/node-llama-win32-arm64-vulkan": "1.4.11",
-    "@fugood/node-llama-win32-x64": "1.4.11",
-    "@fugood/node-llama-win32-x64-cuda": "1.4.11",
-    "@fugood/node-llama-win32-x64-vulkan": "1.4.11"
+    "@fugood/node-llama-darwin-arm64": "1.4.13",
+    "@fugood/node-llama-darwin-x64": "1.4.13",
+    "@fugood/node-llama-linux-arm64": "1.4.13",
+    "@fugood/node-llama-linux-arm64-cuda": "1.4.13",
+    "@fugood/node-llama-linux-arm64-snapdragon": "1.4.13",
+    "@fugood/node-llama-linux-arm64-vulkan": "1.4.13",
+    "@fugood/node-llama-linux-x64": "1.4.13",
+    "@fugood/node-llama-linux-x64-cuda": "1.4.13",
+    "@fugood/node-llama-linux-x64-vulkan": "1.4.13",
+    "@fugood/node-llama-win32-arm64": "1.4.13",
+    "@fugood/node-llama-win32-arm64-vulkan": "1.4.13",
+    "@fugood/node-llama-win32-x64": "1.4.13",
+    "@fugood/node-llama-win32-x64-cuda": "1.4.13",
+    "@fugood/node-llama-win32-x64-vulkan": "1.4.13"
   },
   "devDependencies": {
     "@babel/preset-env": "^7.24.4",

package/scripts/llama.cpp.patch

@@ -32,7 +32,7 @@ index 1bcba9cd8..b7cd68734 100644
  static std::string_view trim_trailing_space(std::string_view sv, int max = -1) {
      int count = 0;
 diff --git a/src/llama.cpp/common/chat.cpp b/src/llama.cpp/common/chat.cpp
-index 0a426f447..ab02be247 100644
+index 22e527bab..c3d0affca 100644
 --- a/src/llama.cpp/common/chat.cpp
 +++ b/src/llama.cpp/common/chat.cpp
 @@ -7,9 +7,6 @@
@@ -62,7 +62,7 @@ index 0a426f447..ab02be247 100644
  struct templates_params {
      json messages;
      json tools;
-@@ -751,7 +738,7 @@ static std::string apply(
+@@ -752,7 +739,7 @@ static std::string apply(
          tmpl_inputs.extra_context.merge_patch(*additional_context);
      }
      // TODO: add flag to control date/time, if only for testing purposes.
@@ -72,7 +72,7 @@ index 0a426f447..ab02be247 100644
      minja::chat_template_options tmpl_opts;
      // To avoid double BOS / EOS tokens, we're manually removing begining / trailing tokens
 diff --git a/src/llama.cpp/common/chat.h b/src/llama.cpp/common/chat.h
-index 6085510a4..263076ce2 100644
+index 8bd4a325f..333b3301f 100644
 --- a/src/llama.cpp/common/chat.h
 +++ b/src/llama.cpp/common/chat.h
 @@ -10,7 +10,18 @@
@@ -96,22 +96,22 @@ index 6085510a4..263076ce2 100644
  struct common_chat_tool_call {
      std::string name;
 diff --git a/src/llama.cpp/common/common.cpp b/src/llama.cpp/common/common.cpp
-index d4e8c7405..af3dec813 100644
+index 41b2b6833..fe9ba05aa 100644
 --- a/src/llama.cpp/common/common.cpp
 +++ b/src/llama.cpp/common/common.cpp
-@@ -1343,6 +1343,7 @@ struct llama_model_params common_model_params_to_llama(common_params & params) {
-         mparams.n_gpu_layers = params.n_gpu_layers;
+@@ -1361,6 +1361,7 @@ struct llama_model_params common_model_params_to_llama(common_params & params) {
+         mparams.devices = params.devices.data();
      }
  
 +    mparams.vocab_only      = params.vocab_only;
+     mparams.n_gpu_layers    = params.n_gpu_layers;
      mparams.main_gpu        = params.main_gpu;
      mparams.split_mode      = params.split_mode;
-     mparams.tensor_split    = params.tensor_split;
 diff --git a/src/llama.cpp/common/common.h b/src/llama.cpp/common/common.h
-index 334372073..e912b593a 100644
+index d6fd0d37a..477209ce5 100644
 --- a/src/llama.cpp/common/common.h
 +++ b/src/llama.cpp/common/common.h
-@@ -307,6 +307,7 @@ struct lr_opt {
+@@ -310,6 +310,7 @@ struct lr_opt {
  struct ggml_opt_optimizer_params common_opt_lr_pars(void * userdata);
  
  struct common_params {
@@ -120,7 +120,7 @@ index 334372073..e912b593a 100644
      int32_t n_ctx                 =     0; // context size, 0 == context the model was trained with
      int32_t n_batch               =  2048; // logical batch size for prompt processing (must be >=32 to use BLAS)
 diff --git a/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt b/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt
-index 28fb7612e..63f7e1ca1 100644
+index 7622d0bf4..d2edcfddb 100644
 --- a/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt
 +++ b/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt
 @@ -106,7 +106,7 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
@@ -133,14 +133,13 @@ index 28fb7612e..63f7e1ca1 100644
              check_cxx_compiler_flag(-mfp16-format=ieee GGML_COMPILER_SUPPORTS_FP16_FORMAT_I3E)
              if (NOT "${GGML_COMPILER_SUPPORTS_FP16_FORMAT_I3E}" STREQUAL "")
 diff --git a/src/llama.cpp/ggml/src/ggml-hexagon/ggml-hexagon.cpp b/src/llama.cpp/ggml/src/ggml-hexagon/ggml-hexagon.cpp
-index 6a00abacc..9e12459b6 100644
+index 365a24b49..83bf4ee62 100644
 --- a/src/llama.cpp/ggml/src/ggml-hexagon/ggml-hexagon.cpp
 +++ b/src/llama.cpp/ggml/src/ggml-hexagon/ggml-hexagon.cpp
-@@ -3226,11 +3226,26 @@ static const char * ggml_backend_hexagon_device_get_description(ggml_backend_dev
+@@ -2798,9 +2798,24 @@ static const char * ggml_backend_hexagon_device_get_description(ggml_backend_dev
      GGML_UNUSED(dev);
  }
  
-+
 +// ~2GB per session for now
 +#define GGML_HEXAGON_SESSION_MEMORY_DEFAULT (2ULL * 1024 * 1024 * 1024)
 +// Max to 3.5GB
@@ -149,7 +148,6 @@ index 6a00abacc..9e12459b6 100644
  static void ggml_backend_hexagon_device_get_memory(ggml_backend_dev_t dev, size_t * free, size_t * total) {
 -    // ~2GB per session for now
 -    *free  = 2ULL * 1024 * 1024 * 1024;
--    *total = *free;
 +    const char * str_mem = getenv("GGML_HEXAGON_SESSION_MEMORY");
 +    if (str_mem) {
 +        *free = std::stoull(str_mem);
@@ -161,32 +159,34 @@ index 6a00abacc..9e12459b6 100644
 +    } else {
 +        *free = GGML_HEXAGON_SESSION_MEMORY_DEFAULT;
 +    }
++
+     *total = *free;
  
-+    *total = *free;
      GGML_UNUSED(dev);
- }
- 
-@@ -3413,10 +3428,17 @@ ggml_hexagon_registry::ggml_hexagon_registry(ggml_backend_reg_t reg) {
+@@ -3010,10 +3025,17 @@ ggml_hexagon_registry::ggml_hexagon_registry(ggml_backend_reg_t reg) {
          }
      }
  
-+#if defined(__ANDROID__)
-     if(opt_arch < 75) {
-         opt_ndev = 1;
+-    if (opt_arch < 75) {
+-        opt_ndev = 1;
 -        GGML_LOG_WARN("ggml-hex: forcing ndev to 1 for SoCs archs lower than v75.\n");
-+        GGML_LOG_WARN("ggml-hex: forcing ndev to 1 for SoCs archs lower than v75 for Android.\n");
-+    }
-+#else
-+    if(opt_arch < 73) {
-+        opt_ndev = 1;
-+        GGML_LOG_WARN("ggml-hex: forcing ndev to 1 for SoCs archs lower than v73 for Linux and Windows.\n");
-     }
-+#endif
+-    }
++    #if defined(__ANDROID__)
++        if(opt_arch < 75) {
++            opt_ndev = 1;
++            GGML_LOG_WARN("ggml-hex: forcing ndev to 1 for SoCs archs lower than v75 for Android.\n");
++        }
++    #else
++        if(opt_arch < 73) {
++            opt_ndev = 1;
++            GGML_LOG_WARN("ggml-hex: forcing ndev to 1 for SoCs archs lower than v73 for Linux and Windows.\n");
++        }
++    #endif
  
      GGML_LOG_INFO("ggml-hex: Hexagon Arch version v%d\n", opt_arch);
  
-@@ -3429,6 +3451,8 @@ ggml_hexagon_registry::ggml_hexagon_registry(ggml_backend_reg_t reg) {
-         } catch (std::exception const &exc) {
+@@ -3026,6 +3048,8 @@ ggml_hexagon_registry::ggml_hexagon_registry(ggml_backend_reg_t reg) {
+         } catch (const std::exception & exc) {
              GGML_LOG_ERROR("ggml-hex: failed to create device/session %zu\n", i);
              devices[i].context = nullptr;
 +            opt_ndev = i;

package/src/llama.cpp/common/arg.cpp

@@ -679,7 +679,6 @@ static void common_params_print_completion(common_params_context & ctx_arg) {
         "llama-quantize",
         "llama-qwen2vl-cli",
         "llama-retrieval",
-        "llama-run",
         "llama-save-load-state",
         "llama-server",
         "llama-simple",
@@ -854,6 +853,54 @@ bool common_arg_utils::is_autoy(const std::string & value) {
     return value == "auto" || value == "-1";
 }
 
+// Simple CSV parser that handles quoted fields and escaped quotes
+// example:
+//    input:  value1,"value, with, commas","value with ""escaped"" quotes",value4
+//    output: [value1] [value, with, commas] [value with "escaped" quotes] [value4]
+static std::vector<std::string> parse_csv_row(const std::string& input) {
+    std::vector<std::string> fields;
+    std::string field;
+    bool in_quotes = false;
+
+    for (size_t i = 0; i < input.length(); ++i) {
+        char ch = input[i];
+
+        if (ch == '"') {
+            if (!in_quotes) {
+                // start of quoted field (only valid if at beginning of field)
+                if (!field.empty()) {
+                    // quote appeared in middle of unquoted field, treat as literal
+                    field += '"';
+                } else {
+                    in_quotes = true; // start
+                }
+            } else {
+                if (i + 1 < input.length() && input[i + 1] == '"') {
+                    // escaped quote: ""
+                    field += '"';
+                    ++i; // skip the next quote
+                } else {
+                    in_quotes = false; // end
+                }
+            }
+        } else if (ch == ',') {
+            if (in_quotes) {
+                field += ',';
+            } else {
+                fields.push_back(std::move(field));
+                field.clear();
+            }
+        } else {
+            field += ch;
+        }
+    }
+
+    // Add the last field
+    fields.push_back(std::move(field));
+
+    return fields;
+}
+
 common_params_context common_params_parser_init(common_params & params, llama_example ex, void(*print_usage)(int, char **)) {
     // per-example default params
     // we define here to make sure it's included in llama-gen-docs
@@ -1250,7 +1297,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         {"--in-file"}, "FNAME",
         "an input file (use comma-separated values to specify multiple files)",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 std::ifstream file(item);
                 if (!file) {
                     throw std::runtime_error(string_format("error: failed to open file '%s'\n", item.c_str()));
@@ -1397,7 +1444,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params, bool value) {
             params.warmup = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_MTMD, LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_PERPLEXITY}));
+    ).set_examples({LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_MTMD, LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_PERPLEXITY, LLAMA_EXAMPLE_DEBUG}));
     add_opt(common_arg(
         {"--spm-infill"},
         string_format(
@@ -1695,6 +1742,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.sampling.grammar = json_schema_to_grammar(json::parse(schema));
         }
     ).set_sparam());
+    add_opt(common_arg(
+        {"-bs", "--backend-sampling"},
+        "enable backend sampling (experimental) (default: disabled)",
+        [](common_params & params) {
+            params.sampling.backend_sampling = true;
+        }
+    ).set_sparam().set_env("LLAMA_ARG_BACKEND_SAMPLING"));
     add_opt(common_arg(
         {"--pooling"}, "{none,mean,cls,last,rank}",
         "pooling type for embeddings, use model default if unspecified",
@@ -1706,7 +1760,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             else if (value == "rank") { params.pooling_type = LLAMA_POOLING_TYPE_RANK; }
             else { throw std::invalid_argument("invalid value"); }
         }
-    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_POOLING"));
+    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_DEBUG}).set_env("LLAMA_ARG_POOLING"));
     add_opt(common_arg(
         {"--attention"}, "{causal,non-causal}",
         "attention type for embeddings, use model default if unspecified",
@@ -1995,7 +2049,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         {"--image", "--audio"}, "FILE",
         "path to an image or audio file. use with multimodal models, use comma-separated values for multiple files\n",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 params.image.emplace_back(item);
             }
         }
@@ -2017,7 +2071,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
     if (llama_supports_rpc()) {
         add_opt(common_arg(
             {"--rpc"}, "SERVERS",
-            "comma separated list of RPC servers",
+            "comma separated list of RPC servers (host:port)",
             [](common_params & params, const std::string & value) {
                 add_rpc_devices(value);
                 GGML_UNUSED(params);
@@ -2087,7 +2141,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         "override tensor buffer type", [](common_params & params, const std::string & value) {
             parse_tensor_buffer_overrides(value, params.tensor_buft_overrides);
         }
-    ));
+    ).set_env("LLAMA_ARG_OVERRIDE_TENSOR"));
     add_opt(common_arg(
         {"-otd", "--override-tensor-draft"}, "<tensor name pattern>=<buffer type>,...",
         "override tensor buffer type for draft model", [](common_params & params, const std::string & value) {
@@ -2137,11 +2191,18 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             }
         }
     ).set_examples({LLAMA_EXAMPLE_SPECULATIVE, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_CLI}).set_env("LLAMA_ARG_N_CPU_MOE_DRAFT"));
+    GGML_ASSERT(params.n_gpu_layers < 0); // string_format would need to be extended for a default >= 0
     add_opt(common_arg(
         {"-ngl", "--gpu-layers", "--n-gpu-layers"}, "N",
-        string_format("max. number of layers to store in VRAM (default: %d)", params.n_gpu_layers),
-        [](common_params & params, int value) {
-            params.n_gpu_layers = value;
+        string_format("max. number of layers to store in VRAM, either an exact number, 'auto', or 'all' (default: %s)", params.n_gpu_layers == -1 ? "auto" : "all"),
+        [](common_params & params, const std::string & value) {
+            if (value == "auto") {
+                params.n_gpu_layers = -1;
+            } else if (value == "all") {
+                params.n_gpu_layers = -2;
+            } else {
+                params.n_gpu_layers = std::stoi(value);
+            }
             if (!llama_supports_gpu_offload()) {
                 fprintf(stderr, "warning: no usable GPU found, --gpu-layers option will be ignored\n");
                 fprintf(stderr, "warning: one possible reason is that llama.cpp was compiled without GPU support\n");
@@ -2245,37 +2306,12 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
     ));
     add_opt(common_arg(
         {"--override-kv"}, "KEY=TYPE:VALUE,...",
-        "advanced option to override model metadata by key. to specify multiple overrides, either use comma-separated or repeat this argument.\n"
+        "advanced option to override model metadata by key. to specify multiple overrides, either use comma-separated values.\n"
         "types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false,tokenizer.ggml.add_eos_token=bool:false",
         [](common_params & params, const std::string & value) {
-            std::vector<std::string> kv_overrides;
-
-            std::string current;
-            bool escaping = false;
-
-            for (const char c : value) {
-                if (escaping) {
-                    current.push_back(c);
-                    escaping = false;
-                } else if (c == '\\') {
-                    escaping = true;
-                } else if (c == ',') {
-                    kv_overrides.push_back(current);
-                    current.clear();
-                } else {
-                    current.push_back(c);
-                }
-            }
-
-            if (escaping) {
-                current.push_back('\\');
-            }
-
-            kv_overrides.push_back(current);
-
-            for (const auto & kv_override : kv_overrides) {
-                if (!string_parse_kv_override(kv_override.c_str(), params.kv_overrides)) {
-                    throw std::runtime_error(string_format("error: Invalid type for KV override: %s\n", kv_override.c_str()));
+            for (const auto & item : parse_csv_row(value)) {
+                if (!string_parse_kv_override(item.c_str(), params.kv_overrides)) {
+                    throw std::runtime_error(string_format("error: Invalid type for KV override: %s\n", item.c_str()));
                 }
             }
         }
@@ -2292,7 +2328,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         {"--lora"}, "FNAME",
         "path to LoRA adapter (use comma-separated values to load multiple adapters)",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 params.lora_adapters.push_back({ item, 1.0, "", "", nullptr });
             }
         }
@@ -2303,7 +2339,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         "path to LoRA adapter with user defined scaling (format: FNAME:SCALE,...)\n"
         "note: use comma-separated values",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 auto parts = string_split<std::string>(item, ':');
                 if (parts.size() != 2) {
                     throw std::invalid_argument("lora-scaled format: FNAME:SCALE");
@@ -2317,7 +2353,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         {"--control-vector"}, "FNAME",
         "add a control vector\nnote: use comma-separated values to add multiple control vectors",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 params.control_vectors.push_back({ 1.0f, item, });
             }
         }
@@ -2327,7 +2363,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         "add a control vector with user defined scaling SCALE\n"
         "note: use comma-separated values (format: FNAME:SCALE,...)",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 auto parts = string_split<std::string>(item, ':');
                 if (parts.size() != 2) {
                     throw std::invalid_argument("control-vector-scaled format: FNAME:SCALE");
@@ -2425,7 +2461,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         {"--context-file"}, "FNAME",
         "file to load context from (use comma-separated values to specify multiple files)",
         [](common_params & params, const std::string & value) {
-            for (const auto & item : string_split<std::string>(value, ',')) {
+            for (const auto & item : parse_csv_row(value)) {
                 std::ifstream file(item, std::ios::binary);
                 if (!file) {
                     throw std::runtime_error(string_format("error: failed to open file '%s'\n", item.c_str()));
@@ -2572,7 +2608,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params, int value) {
             params.embd_normalize = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_EMBEDDING}));
+    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_DEBUG}));
     add_opt(common_arg(
         {"--embd-output-format"}, "FORMAT",
         "empty = default, \"array\" = [[],[]...], \"json\" = openai style, \"json+\" = same \"json\" + cosine similarity matrix, \"raw\" = plain whitespace-delimited output (one embedding per line)",
@@ -2650,7 +2686,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params) {
             params.embedding = true;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_EMBEDDINGS"));
+    ).set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_DEBUG}).set_env("LLAMA_ARG_EMBEDDINGS"));
     add_opt(common_arg(
         {"--rerank", "--reranking"},
         string_format("enable reranking endpoint on server (default: %s)", "disabled"),
@@ -2661,9 +2697,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
     ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_RERANKING"));
     add_opt(common_arg(
         {"--api-key"}, "KEY",
-        "API key to use for authentication (default: none)",
+        "API key to use for authentication, multiple keys can be provided as a comma-separated list (default: none)",
         [](common_params & params, const std::string & value) {
-            params.api_keys.push_back(value);
+            for (const auto & key : parse_csv_row(value)) {
+                if (!key.empty()) {
+                    params.api_keys.push_back(key);
+                }
+            }
         }
     ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_API_KEY"));
     add_opt(common_arg(
@@ -2677,7 +2717,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             std::string key;
             while (std::getline(key_file, key)) {
                 if (!key.empty()) {
-                        params.api_keys.push_back(key);
+                    params.api_keys.push_back(key);
                 }
             }
             key_file.close();
@@ -2699,7 +2739,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
     ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_SSL_CERT_FILE"));
     add_opt(common_arg(
         {"--chat-template-kwargs"}, "STRING",
-        string_format("sets additional params for the json template parser"),
+        "sets additional params for the json template parser, must be a valid json object string, e.g. '{\"key1\":\"value1\",\"key2\":\"value2\"}'",
         [](common_params & params, const std::string & value) {
             auto parsed = json::parse(value);
             for (const auto & item : parsed.items()) {
@@ -3175,11 +3215,19 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.speculative.devices = parse_device_list(value);
         }
     ).set_examples({LLAMA_EXAMPLE_SPECULATIVE, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_CLI}));
+    GGML_ASSERT(params.speculative.n_gpu_layers < 0); // string_format would need to be extended for a default >= 0
     add_opt(common_arg(
         {"-ngld", "--gpu-layers-draft", "--n-gpu-layers-draft"}, "N",
-        "number of layers to store in VRAM for the draft model",
-        [](common_params & params, int value) {
-            params.speculative.n_gpu_layers = value;
+        string_format("max. number of draft model layers to store in VRAM, either an exact number, 'auto', or 'all' (default: %s)",
+            params.speculative.n_gpu_layers == -1 ? "auto" : "all"),
+        [](common_params & params, const std::string & value) {
+            if (value == "auto") {
+                params.speculative.n_gpu_layers = -1;
+            } else if (value == "all") {
+                params.speculative.n_gpu_layers = -2;
+            } else {
+                params.speculative.n_gpu_layers = std::stoi(value);
+            }
             if (!llama_supports_gpu_offload()) {
                 fprintf(stderr, "warning: no usable GPU found, --gpu-layers-draft option will be ignored\n");
                 fprintf(stderr, "warning: one possible reason is that llama.cpp was compiled without GPU support\n");
@@ -3329,6 +3377,27 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             }
         }
     ).set_examples({ LLAMA_EXAMPLE_FINETUNE }));
+    add_opt(common_arg(
+        {"--save-logits"},
+        string_format("save final logits to files for verification (default: %s)", params.save_logits ? "true" : "false"),
+        [](common_params & params) {
+            params.save_logits = true;
+        }
+    ).set_examples({LLAMA_EXAMPLE_DEBUG}));
+    add_opt(common_arg(
+        {"--logits-output-dir"}, "PATH",
+        string_format("directory for saving logits output files (default: %s)", params.logits_output_dir.c_str()),
+        [](common_params & params, const std::string & value) {
+            params.logits_output_dir = value;
+        }
+    ).set_examples({LLAMA_EXAMPLE_DEBUG}));
+    add_opt(common_arg(
+        {"--tensor-filter"}, "REGEX",
+        "filter tensor names for debug output (regex pattern, can be specified multiple times)",
+        [](common_params & params, const std::string & value) {
+            params.tensor_filter.push_back(value);
+        }
+    ).set_examples({LLAMA_EXAMPLE_DEBUG}));
 
     // presets
     add_opt(common_arg(
@@ -3518,15 +3587,15 @@ void common_params_add_preset_options(std::vector<common_arg> & args) {
         [](common_params &, const std::string &) { /* unused */ }
     ).set_env(COMMON_ARG_PRESET_LOAD_ON_STARTUP).set_preset_only());
 
+    args.push_back(common_arg(
+        {"stop-timeout"}, "SECONDS",
+        "in server router mode, force-kill model instance after this many seconds of graceful shutdown",
+        [](common_params &, int) { /* unused */ }
+    ).set_env(COMMON_ARG_PRESET_STOP_TIMEOUT).set_preset_only());
+
     // args.push_back(common_arg(
     //     {"pin"},
     //     "in server router mode, do not unload this model if models_max is exceeded",
     //     [](common_params &) { /* unused */ }
     // ).set_preset_only());
-
-    // args.push_back(common_arg(
-    //     {"unload-idle-seconds"}, "SECONDS",
-    //     "in server router mode, unload models idle for more than this many seconds",
-    //     [](common_params &, int) { /* unused */ }
-    // ).set_preset_only());
 }

package/src/llama.cpp/common/arg.h

@@ -10,6 +10,7 @@
 
 // pseudo-env variable to identify preset-only arguments
 #define COMMON_ARG_PRESET_LOAD_ON_STARTUP "__PRESET_LOAD_ON_STARTUP"
+#define COMMON_ARG_PRESET_STOP_TIMEOUT    "__PRESET_STOP_TIMEOUT"
 
 //
 // CLI argument parsing

package/src/llama.cpp/common/chat-parser.cpp

@@ -1395,6 +1395,14 @@ static void common_chat_parse_seed_oss(common_chat_msg_parser & builder) {
     builder.consume_reasoning_with_xml_tool_calls(form, "<seed:think>", "</seed:think>");
 }
 
+static void common_chat_parse_solar_open(common_chat_msg_parser & builder) {
+    builder.try_parse_reasoning("<|think|>", "<|end|><|begin|>assistant<|content|>");
+
+    // TODO: Tool calling
+
+    builder.add_content(builder.consume_rest());
+}
+
 static void common_chat_parse_content_only(common_chat_msg_parser & builder) {
     builder.try_parse_reasoning("<think>", "</think>");
     builder.add_content(builder.consume_rest());
@@ -1479,6 +1487,9 @@ static void common_chat_parse(common_chat_msg_parser & builder) {
         case COMMON_CHAT_FORMAT_XIAOMI_MIMO:
             common_chat_parse_xiaomi_mimo(builder);
             break;
+        case COMMON_CHAT_FORMAT_SOLAR_OPEN:
+            common_chat_parse_solar_open(builder);
+            break;
         default:
             throw std::runtime_error(std::string("Unsupported format: ") + common_chat_format_name(builder.syntax().format));
     }