llama-cpp-bin 9828.0.0__tar.gz → 9830.0.0__tar.gz

{llama_cpp_bin-9828.0.0/src/llama_cpp_bin.egg-info → llama_cpp_bin-9830.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llama-cpp-bin
-Version: 9828.0.0
+Version: 9830.0.0
 Summary: llama.cpp server binary built from source
 Author-email: Vladyslav Tkachenko <vladlearns@gmail.com>
 License: AGPL-3.0-or-later

{llama_cpp_bin-9828.0.0 → llama_cpp_bin-9830.0.0}/llama.cpp/common/arg.cpp

@@ -467,7 +467,7 @@ void common_models_handler_apply(common_models_handler & handler, common_params
         // the first part is what gets loaded, so point params.model.path at it
         if (!url_tasks.empty()) {
             std::string first_path = url_tasks.front().local_path;
-            url_tasks.front().on_done = [&]() { params.model.path = first_path; };
+            url_tasks.front().on_done = [&, first_path]() { params.model.path = first_path; };
         }
         for (auto & task : url_tasks) {
             tasks.push_back(std::move(task));
@@ -3471,7 +3471,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params) {
             params.offline = true;
         }
-    ).set_env("LLAMA_ARG_OFFLINE"));
+    ).set_examples({LLAMA_EXAMPLE_COMMON, LLAMA_EXAMPLE_DOWNLOAD}).set_env("LLAMA_ARG_OFFLINE"));
     add_opt(common_arg(
         {"-lv", "--verbosity", "--log-verbosity"}, "N",
         string_format("Set the verbosity threshold. Messages with a higher verbosity will be ignored. Values:\n"

{llama_cpp_bin-9828.0.0 → llama_cpp_bin-9830.0.0}/llama.cpp/common/common.cpp

@@ -225,7 +225,7 @@ bool set_process_priority(enum ggml_sched_priority prio) {
     }
 
     if (!SetPriorityClass(GetCurrentProcess(), p)) {
-        LOG_WRN("failed to set process priority class %d : (%d)\n", prio, (int) GetLastError());
+        COM_WRN("failed to set process priority class %d : (%d)\n", prio, (int) GetLastError());
         return false;
     }
 
@@ -251,7 +251,7 @@ bool set_process_priority(enum ggml_sched_priority prio) {
     }
 
     if (setpriority(PRIO_PROCESS, 0, p) != 0) {
-        LOG_WRN("failed to set process priority %d : %s (%d)\n", prio, strerror(errno), errno);
+        COM_WRN("failed to set process priority %d : %s (%d)\n", prio, strerror(errno), errno);
         return false;
     }
     return true;
@@ -284,14 +284,14 @@ void postprocess_cpu_params(common_cpu_params & cpuparams, const common_cpu_para
 
     if (n_set && n_set < cpuparams.n_threads) {
         // Not enough set bits, may experience performance issues.
-        LOG_WRN("Not enough set bits in CPU mask (%d) to satisfy requested thread count: %d\n", n_set, cpuparams.n_threads);
+        COM_WRN("Not enough set bits in CPU mask (%d) to satisfy requested thread count: %d\n", n_set, cpuparams.n_threads);
     }
 }
 
 bool parse_cpu_range(const std::string & range, bool (&boolmask)[GGML_MAX_N_THREADS]) {
     size_t dash_loc = range.find('-');
     if (dash_loc == std::string::npos) {
-        LOG_ERR("Format of CPU range is invalid! Expected [<start>]-[<end>].\n");
+        COM_ERR("%s", "Format of CPU range is invalid! Expected [<start>]-[<end>].\n");
         return false;
     }
 
@@ -303,7 +303,7 @@ bool parse_cpu_range(const std::string & range, bool (&boolmask)[GGML_MAX_N_THRE
     } else {
         start_i = std::stoull(range.substr(0, dash_loc));
         if (start_i >= GGML_MAX_N_THREADS) {
-            LOG_ERR("Start index out of bounds!\n");
+            COM_ERR("%s", "Start index out of bounds!\n");
             return false;
         }
     }
@@ -313,7 +313,7 @@ bool parse_cpu_range(const std::string & range, bool (&boolmask)[GGML_MAX_N_THRE
     } else {
         end_i = std::stoull(range.substr(dash_loc + 1));
         if (end_i >= GGML_MAX_N_THREADS) {
-            LOG_ERR("End index out of bounds!\n");
+            COM_ERR("%s", "End index out of bounds!\n");
             return false;
         }
     }
@@ -333,7 +333,7 @@ bool parse_cpu_mask(const std::string & mask, bool (&boolmask)[GGML_MAX_N_THREAD
     }
 
     size_t num_digits = mask.length() - start_i;
-    if (num_digits > 128) num_digits = 128;
+    num_digits = std::min<size_t>(num_digits, 128);
 
     size_t end_i = num_digits + start_i;
 
@@ -348,7 +348,7 @@ bool parse_cpu_mask(const std::string & mask, bool (&boolmask)[GGML_MAX_N_THREAD
         } else if (c >= 'A' && c <= 'F') {
             id -= 'A' - 10;
         } else {
-            LOG_ERR("Invalid hex character '%c' at position %d\n", c, int32_t(i));
+            COM_ERR("Invalid hex character '%c' at position %d\n", c, int32_t(i));
             return false;
         }
 
@@ -379,21 +379,21 @@ void common_params_print_info(const common_params & params, bool print_devices)
 #else
     const char * build_type = " (debug)";
 #endif
-    LOG_TRC("%s: build %d (%s) with %s for %s%s\n", __func__, llama_build_number(), llama_commit(), llama_compiler(), llama_build_target(), build_type);
+    COM_TRC("%s: build %d (%s) with %s for %s%s\n", __func__, llama_build_number(), llama_commit(), llama_compiler(), llama_build_target(), build_type);
 
-    LOG_INF("log_info: verbosity = %d (adjust with the `-lv N` CLI arg)\n", common_log_get_verbosity_thold());
+    COM_INF("%s: verbosity = %d (adjust with the `-lv N` CLI arg)\n", __func__, common_log_get_verbosity_thold());
 
     // device enumeration creates a primary context on CUDA backends, skip it when the caller does not own any device
     if (print_devices) {
-        LOG_INF("device_info:\n");
+        COM_TRC("%s", "device_info:\n");
         for (size_t i = 0; i < ggml_backend_dev_count(); ++i) {
             auto * dev = ggml_backend_dev_get(i);
             size_t free, total;
             ggml_backend_dev_memory(dev, &free, &total);
-            LOG_INF("  - %-8s: %s (%zu MiB, %zu MiB free)\n", ggml_backend_dev_name(dev), ggml_backend_dev_description(dev), total / 1024 / 1024, free / 1024 / 1024);
+            COM_TRC("  - %-8s: %s (%zu MiB, %zu MiB free)\n", ggml_backend_dev_name(dev), ggml_backend_dev_description(dev), total / 1024 / 1024, free / 1024 / 1024);
         }
     }
-    LOG_INF("%s\n", common_params_get_system_info(params).c_str());
+    COM_TRC("%s\n", common_params_get_system_info(params).c_str());
 }
 
 std::string common_params_get_system_info(const common_params & params) {
@@ -660,7 +660,7 @@ void string_process_escapes(std::string & input) {
 bool string_parse_kv_override(const char * data, std::vector<llama_model_kv_override> & overrides) {
     const char * sep = strchr(data, '=');
     if (sep == nullptr || sep - data >= 128) {
-        LOG_ERR("%s: malformed KV override '%s'\n", __func__, data);
+        COM_ERR("%s: malformed KV override '%s'\n", __func__, data);
         return false;
     }
     llama_model_kv_override kvo;
@@ -683,20 +683,20 @@ bool string_parse_kv_override(const char * data, std::vector<llama_model_kv_over
         } else if (std::strcmp(sep, "false") == 0) {
             kvo.val_bool = false;
         } else {
-            LOG_ERR("%s: invalid boolean value for KV override '%s'\n", __func__, data);
+            COM_ERR("%s: invalid boolean value for KV override '%s'\n", __func__, data);
             return false;
         }
     } else if (strncmp(sep, "str:", 4) == 0) {
         sep += 4;
         kvo.tag = LLAMA_KV_OVERRIDE_TYPE_STR;
         if (strlen(sep) > 127) {
-            LOG_ERR("%s: malformed KV override '%s', value cannot exceed 127 chars\n", __func__, data);
+            COM_ERR("%s: malformed KV override '%s', value cannot exceed 127 chars\n", __func__, data);
             return false;
         }
         strncpy(kvo.val_str, sep, 127);
         kvo.val_str[127] = '\0';
     } else {
-        LOG_ERR("%s: invalid type for KV override '%s'\n", __func__, data);
+        COM_ERR("%s: invalid type for KV override '%s'\n", __func__, data);
         return false;
     }
     overrides.emplace_back(std::move(kvo));
@@ -1199,8 +1199,8 @@ common_init_result::common_init_result(common_params & params, bool model_only)
     auto cparams = common_context_params_to_llama(params);
 
     if (params.fit_params) {
-        LOG_INF("%s: fitting params to device memory ...\n", __func__);
-        LOG_INF("%s: (for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on)\n", __func__);
+        COM_TRC("%s", "fitting params to device memory ...\n");
+        COM_TRC("%s", "(for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on)\n");
         common_fit_params(params.model.path.c_str(), &mparams, &cparams,
             params.tensor_split,
             params.tensor_buft_overrides.data(),
@@ -1227,7 +1227,7 @@ common_init_result::common_init_result(common_params & params, bool model_only)
         llama_adapter_lora_ptr lora;
         lora.reset(llama_adapter_lora_init(model, la.path.c_str()));
         if (lora == nullptr) {
-            LOG_ERR("%s: failed to load lora adapter '%s'\n", __func__, la.path.c_str());
+            COM_ERR("failed to load lora adapter '%s'\n", la.path.c_str());
             pimpl->model.reset(model);
             return;
         }
@@ -1246,14 +1246,14 @@ common_init_result::common_init_result(common_params & params, bool model_only)
     common_init_sampler_from_model(model, params.sampling);
 
     if (params.sampling.ignore_eos && llama_vocab_eos(vocab) == LLAMA_TOKEN_NULL) {
-        LOG_WRN("%s: warning: vocab does not have an EOS token, ignoring --ignore-eos\n", __func__);
+        COM_WRN("%s", "vocab does not have an EOS token, ignoring --ignore-eos\n");
         params.sampling.ignore_eos = false;
     }
 
     // initialize once
     for (llama_token i = 0; i < llama_vocab_n_tokens(vocab); i++) {
         if (llama_vocab_is_eog(vocab, i)) {
-            LOG_TRC("%s: added %s logit bias = %f\n", __func__, common_token_to_piece(vocab, i).c_str(), -INFINITY);
+            COM_TRC("added %s logit bias = %f\n", common_token_to_piece(vocab, i).c_str(), -INFINITY);
             params.sampling.logit_bias_eog.push_back({i, -INFINITY});
         }
     }
@@ -1291,7 +1291,7 @@ common_init_result::common_init_result(common_params & params, bool model_only)
 
     llama_context * lctx = llama_init_from_model(model, cparams);
     if (lctx == NULL) {
-        LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.path.c_str());
+        COM_ERR("failed to create context with model '%s'\n", params.model.path.c_str());
         return;
     }
 
@@ -1328,7 +1328,7 @@ common_init_result_ptr common_init_from_params(common_params & params, bool mode
 
     llama_model * model = res->model();
     if (model == NULL) {
-        LOG_ERR("%s: failed to load model '%s'\n", __func__, params.model.path.c_str());
+        COM_ERR("failed to load model '%s'\n", params.model.path.c_str());
         return res;
     }
 
@@ -1338,14 +1338,14 @@ common_init_result_ptr common_init_from_params(common_params & params, bool mode
 
     llama_context * lctx = res->context();
     if (lctx == NULL) {
-        LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.path.c_str());
+        COM_ERR("failed to create context with model '%s'\n", params.model.path.c_str());
         return res;
     }
 
     const llama_vocab * vocab = llama_model_get_vocab(model);
 
     if (params.ctx_shift && !llama_memory_can_shift(llama_get_memory(lctx))) {
-        LOG_WRN("%s: KV cache shifting is not supported for this context, disabling KV cache shifting\n", __func__);
+        COM_WRN("%s", "KV cache shifting is not supported for this context, disabling KV cache shifting\n");
         params.ctx_shift = false;
     }
 
@@ -1374,7 +1374,7 @@ common_init_result_ptr common_init_from_params(common_params & params, bool mode
         bool ok = true;
 
         if (llama_vocab_bos(vocab) == LLAMA_TOKEN_NULL) {
-            LOG_WRN("%s: warning: vocab does not have a  BOS token, reranking will not work\n", __func__);
+            COM_WRN("%s", "vocab does not have a  BOS token, reranking will not work\n");
             ok = false;
         }
 
@@ -1383,10 +1383,10 @@ common_init_result_ptr common_init_from_params(common_params & params, bool mode
         bool has_rerank_prompt = llama_model_chat_template(model, "rerank") != NULL;
 
         if (!has_eos && !has_sep && !has_rerank_prompt) {
-            LOG_WRN("%s: warning: vocab does not have an EOS token, SEP token, or rerank prompt. Reranking will not work\n", __func__);
+            COM_WRN("%s", "vocab does not have an EOS token, SEP token, or rerank prompt. Reranking will not work\n");
             ok = false;
         } else if (!has_eos) {
-            LOG_WRN("%s: warning: vocab does not have an EOS token, using SEP token as fallback\n", __func__);
+            COM_WRN("%s", "vocab does not have an EOS token, using SEP token as fallback\n");
         }
 
         if (!ok) {
@@ -1399,7 +1399,7 @@ common_init_result_ptr common_init_from_params(common_params & params, bool mode
     }
 
     if (params.warmup) {
-        LOG_INF("%s: warming up the model with an empty run - please wait ... (--no-warmup to disable)\n", __func__);
+        COM_TRC("%s", "warming up the model with an empty run - please wait ... (--no-warmup to disable)\n");
 
         std::vector<llama_token> tmp;
         llama_token bos = llama_vocab_bos(vocab);
@@ -1473,20 +1473,20 @@ common_context_seq_rm_type common_context_can_seq_rm(llama_context * ctx) {
 
     int ret = llama_decode(ctx, llama_batch_get_one(tmp.data(), tmp.size()));
     if (ret != 0) {
-        LOG_ERR("%s: llama_decode() failed: %d\n", __func__, ret);
+        COM_ERR("llama_decode() failed: %d\n", ret);
         res = COMMON_CONTEXT_SEQ_RM_TYPE_NO;
         goto done;
     }
 
     if (llama_n_rs_seq(ctx) > 0) {
-        LOG_INF("%s: the context supports bounded partial sequence removal\n", __func__);
+        COM_TRC("%s", "the context supports bounded partial sequence removal\n");
         res = COMMON_CONTEXT_SEQ_RM_TYPE_RS;
         goto done;
     }
 
     // try to remove the last tokens
     if (!llama_memory_seq_rm(mem, 0, 1, -1)) {
-        LOG_TRC("%s: the context does not support partial sequence removal\n", __func__);
+        COM_TRC("%s", "the context does not support partial sequence removal\n");
         res = COMMON_CONTEXT_SEQ_RM_TYPE_FULL;
         goto done;
     }
@@ -1803,13 +1803,13 @@ static common_control_vector_data common_control_vector_load_one(const common_co
     };
     struct gguf_context * ctx_gguf = gguf_init_from_file(load_info.fname.c_str(), meta_gguf_params);
     if (!ctx_gguf) {
-        LOG_ERR("%s: failed to load control vector file from %s\n", __func__, load_info.fname.c_str());
+        COM_ERR("failed to load control vector file from %s\n", load_info.fname.c_str());
         return result;
     }
 
     int32_t n_tensors = gguf_get_n_tensors(ctx_gguf);
     if (n_tensors == 0) {
-        LOG_WRN("%s: no direction tensors found in %s\n", __func__, load_info.fname.c_str());
+        COM_WRN("no direction tensors found in %s\n", load_info.fname.c_str());
     }
 
     for (int i = 0; i < n_tensors; i++) {
@@ -1827,23 +1827,23 @@ static common_control_vector_data common_control_vector_load_one(const common_co
             }
         }
         if (layer_idx < 0) {
-            LOG_ERR("%s: invalid/unparsable direction tensor layer index in %s\n", __func__, load_info.fname.c_str());
+            COM_ERR("invalid/unparsable direction tensor layer index in %s\n", load_info.fname.c_str());
             result.n_embd = -1;
             break;
         } else if (layer_idx == 0) {
-            LOG_ERR("%s: invalid (zero) direction tensor layer index in %s\n", __func__, load_info.fname.c_str());
+            COM_ERR("invalid (zero) direction tensor layer index in %s\n", load_info.fname.c_str());
             result.n_embd = -1;
             break;
         }
 
         struct ggml_tensor * tensor = ggml_get_tensor(ctx, name.c_str());
         if (tensor->type != GGML_TYPE_F32) {
-            LOG_ERR("%s: invalid (non-F32) direction tensor type in %s\n", __func__, load_info.fname.c_str());
+            COM_ERR("invalid (non-F32) direction tensor type in %s\n", load_info.fname.c_str());
             result.n_embd = -1;
             break;
         }
         if (ggml_n_dims(tensor) != 1) {
-            LOG_ERR("%s: invalid (non-1D) direction tensor shape in %s\n", __func__, load_info.fname.c_str());
+            COM_ERR("invalid (non-1D) direction tensor shape in %s\n", load_info.fname.c_str());
             result.n_embd = -1;
             break;
         }
@@ -1851,7 +1851,7 @@ static common_control_vector_data common_control_vector_load_one(const common_co
         if (result.n_embd == -1) {
             result.n_embd = ggml_nelements(tensor);
         } else if (ggml_nelements(tensor) != result.n_embd) {
-            LOG_ERR("%s: direction tensor in %s does not match previous dimensions\n", __func__, load_info.fname.c_str());
+            COM_ERR("direction tensor in %s does not match previous dimensions\n", load_info.fname.c_str());
             result.n_embd = -1;
             break;
         }
@@ -1868,7 +1868,7 @@ static common_control_vector_data common_control_vector_load_one(const common_co
     }
 
     if (result.n_embd == -1) {
-        LOG_WRN("%s: skipping %s due to invalid direction tensors\n", __func__, load_info.fname.c_str());
+        COM_WRN("skipping %s due to invalid direction tensors\n", load_info.fname.c_str());
         result.data.clear();
     }
 
@@ -1889,7 +1889,7 @@ common_control_vector_data common_control_vector_load(const std::vector<common_c
             break;
         }
         if (result.n_embd != -1 && result.n_embd != cur.n_embd) {
-            LOG_ERR("%s: control vectors in %s does not match previous dimensions\n", __func__, info.fname.c_str());
+            COM_ERR("control vectors in %s does not match previous dimensions\n", info.fname.c_str());
             result.n_embd = -1;
             break;
         }
@@ -1905,7 +1905,7 @@ common_control_vector_data common_control_vector_load(const std::vector<common_c
     }
 
     if (result.n_embd == -1) {
-        LOG_ERR("%s: no valid control vector files passed\n", __func__);
+        COM_ERR("%s", "no valid control vector files passed\n");
         result.data.clear();
     }
 
@@ -2016,13 +2016,13 @@ bool common_prompt_batch_decode(
         // memory, so we can't just remove the last token from the memory and replay the last token which
         // is the reason for this logic.
         if (llama_decode(ctx, llama_batch_get_one(const_cast<llama_token*>(all_tokens.data() + offset), n_tokens_before_last))) {
-            LOG_ERR("%s : failed to eval\n", __func__);
+            COM_ERR("%s", "failed to eval\n");
             return false;
         }
         n_past += n_tokens_before_last;
 
         llama_state_save_file(ctx, state_path.data(), all_tokens.data(), all_tokens.size());
-        LOG_INF("saved session before last token to %s, n_new = %zu\n", state_path.data(), all_tokens.size());
+        COM_INF("saved session before last token to %s, n_new = %zu\n", state_path.data(), all_tokens.size());
 
         llama_token last_token = all_tokens.back();
         llama_batch batch = llama_batch_get_one(&last_token, 1);
@@ -2030,13 +2030,13 @@ bool common_prompt_batch_decode(
         batch.pos = &pos;
 
         if (llama_decode(ctx, batch)) {
-            LOG_ERR("%s : failed to eval last token\n", __func__);
+            COM_ERR("%s", "failed to eval last token\n");
             return false;
         }
         n_past++;
     } else {
         if (llama_decode(ctx, llama_batch_get_one(const_cast<llama_token*>(all_tokens.data() + offset), n_new))) {
-            LOG_ERR("%s : failed to eval\n", __func__);
+            COM_ERR("%s", "failed to eval\n");
             return false;
         }
         n_past += n_new;

{llama_cpp_bin-9828.0.0 → llama_cpp_bin-9830.0.0}/llama.cpp/common/common.h

@@ -25,6 +25,13 @@
 #define DIRECTORY_SEPARATOR '/'
 #endif // _WIN32
 
+#define COM_DBG(fmt, ...) LOG_DBG("cmn  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
+#define COM_TRC(fmt, ...) LOG_TRC("cmn  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
+#define COM_INF(fmt, ...) LOG_INF("cmn  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
+#define COM_WRN(fmt, ...) LOG_WRN("cmn  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
+#define COM_ERR(fmt, ...) LOG_ERR("cmn  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
+#define COM_CNT(fmt, ...) LOG_CNT(""              fmt,               __VA_ARGS__)
+
 #define die(msg)          do { fputs("error: " msg "\n", stderr);                exit(1); } while (0)
 #define die_fmt(fmt, ...) do { fprintf(stderr, "error: " fmt "\n", __VA_ARGS__); exit(1); } while (0)
 

{llama_cpp_bin-9828.0.0 → llama_cpp_bin-9830.0.0}/llama.cpp/common/fit.cpp

@@ -233,7 +233,7 @@ static void common_params_fit_impl(
         sum_projected_used = dmds_full.back().mb.total();
         sum_free           = dmds_full.back().total;
         sum_projected_free = sum_free - sum_projected_used;
-        LOG_INF("%s: projected to use %" PRId64 " MiB of host memory vs. %" PRId64 " MiB of total host memory\n",
+        LOG_TRC("%s: projected to use %" PRId64 " MiB of host memory vs. %" PRId64 " MiB of total host memory\n",
             __func__, sum_projected_used/MiB, sum_free/MiB);
         if (sum_projected_free >= margins[0]) {
             LOG_TRC("%s: will leave %" PRId64 " >= %" PRId64 " MiB of system memory, no changes needed\n",

{llama_cpp_bin-9828.0.0 → llama_cpp_bin-9830.0.0}/llama.cpp/common/reasoning-budget.cpp

@@ -65,12 +65,12 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
             if (ctx->start_matcher.advance(token)) {
                 ctx->state = REASONING_BUDGET_COUNTING;
                 ctx->remaining = ctx->budget;
-                LOG_INF("reasoning-budget: activated, budget=%d tokens\n", ctx->budget);
+                COM_TRC("activated, budget=%d tokens\n", ctx->budget);
 
                 if (ctx->remaining <= 0) {
                     ctx->state = REASONING_BUDGET_FORCING;
                     ctx->force_pos = 0;
-                    LOG_INF("reasoning-budget: budget=0, forcing immediately\n");
+                    COM_TRC("%s", "budget=0, forcing immediately\n");
                 }
             }
             break;
@@ -80,7 +80,7 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
         {
             if (ctx->end_matcher.advance(token)) {
                 ctx->state = REASONING_BUDGET_DONE;
-                LOG_INF("reasoning-budget: deactivated (natural end)\n");
+                COM_TRC("%s", "deactivated (natural end)\n");
                 break;
             }
 
@@ -95,7 +95,7 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
                     ctx->state = REASONING_BUDGET_FORCING;
                     ctx->force_pos = 0;
                     ctx->end_matcher.reset();
-                    LOG_INF("reasoning-budget: UTF-8 complete, now forcing end sequence\n");
+                    COM_TRC("%s", "UTF-8 complete, now forcing end sequence\n");
                 }
             } else if (ctx->state == REASONING_BUDGET_COUNTING) {
                 ctx->remaining--;
@@ -104,11 +104,11 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
                         ctx->state = REASONING_BUDGET_FORCING;
                         ctx->force_pos = 0;
                         ctx->end_matcher.reset();
-                        LOG_INF("reasoning-budget: budget exhausted, forcing end sequence\n");
+                        COM_TRC("%s", "budget exhausted, forcing end sequence\n");
                     } else {
                         ctx->state = REASONING_BUDGET_WAITING_UTF8;
                         ctx->end_matcher.reset();
-                        LOG_INF("reasoning-budget: budget exhausted, waiting for UTF-8 completion\n");
+                        COM_TRC("%s", "budget exhausted, waiting for UTF-8 completion\n");
                     }
                 }
             }
@@ -118,7 +118,7 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
             ctx->force_pos++;
             if (ctx->force_pos >= ctx->forced_tokens.size()) {
                 ctx->state = REASONING_BUDGET_DONE;
-                LOG_INF("reasoning-budget: forced sequence complete, done\n");
+                COM_TRC("%s", "forced sequence complete, done\n");
             }
             break;
         case REASONING_BUDGET_DONE:
@@ -128,12 +128,12 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
                 ctx->state = REASONING_BUDGET_COUNTING;
                 ctx->remaining = ctx->budget;
                 ctx->end_matcher.reset();
-                LOG_INF("reasoning-budget: re-activated on new start tag, budget=%d tokens\n", ctx->budget);
+                COM_TRC("re-activated on new start tag, budget=%d tokens\n", ctx->budget);
 
                 if (ctx->remaining <= 0) {
                     ctx->state = REASONING_BUDGET_FORCING;
                     ctx->force_pos = 0;
-                    LOG_INF("reasoning-budget: budget=0, forcing immediately\n");
+                    COM_TRC("%s", "budget=0, forcing immediately\n");
                 }
             }
             break;
@@ -264,7 +264,7 @@ bool common_reasoning_budget_force(struct llama_sampler * smpl) {
     ctx->state = REASONING_BUDGET_FORCING;
     ctx->force_pos = 0;
     ctx->end_matcher.reset();
-    LOG_INF("reasoning-budget: forced into forcing state (manual transition)\n");
+    COM_TRC("%s", "forced into forcing state (manual transition)\n");
 
     return true;
 }