Remove debug-associated and other misceallanous changes

Author: pwilkin

common/arg.cpp

@@ -1951,13 +1951,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.kv_unified = true;
         }
     ).set_env("LLAMA_ARG_KV_SPLIT"));
-    add_opt(common_arg(
-        {"--dump-cache"},
-        "dump cache statistics after each token generation",
-        [](common_params & params) {
-            params.dump_cache = true;
-        }
-    ).set_examples({LLAMA_EXAMPLE_MAIN}));
     add_opt(common_arg(
         {"--no-context-shift"},
         string_format("disables context shift on infinite text generation (default: %s)", params.ctx_shift ? "disabled" : "enabled"),

common/common.h

@@ -399,8 +399,6 @@ struct common_params {
     ggml_type cache_type_k = GGML_TYPE_F16; // KV cache data type for the K
     ggml_type cache_type_v = GGML_TYPE_F16; // KV cache data type for the V
 
-    bool dump_cache = false; // dump cache statistics after each token
-
     common_conversation_mode conversation_mode = COMMON_CONVERSATION_MODE_AUTO;
 
     // multimodal models (see tools/mtmd)

examples/eval-callback/eval-callback.cpp

@@ -154,7 +154,7 @@ static bool ggml_debug(struct ggml_tensor * t, bool ask, void * user_data) {
 
     if (!ggml_is_quantized(t->type)) {
         uint8_t * data = is_host ? (uint8_t *) t->data : cb_data->data.data();
-        ggml_print_tensor(data, t->type, t->ne, t->nb, 8);
+        ggml_print_tensor(data, t->type, t->ne, t->nb, 3);
     }
 
     return true;

examples/model-conversion/qwen3stories.sh

@@ -6,7 +6,6 @@
   "reportDuplicateImport": "error",
   "reportDeprecated": "warning",
   "reportUnnecessaryTypeIgnoreComment": "information",
-  "reportAttributeAccessIssue": "warning",
   "disableBytesTypePromotions": false, // TODO: change once Python 3.12 is the minimum
   "executionEnvironments": [
     {

pyrightconfig.json

@@ -1362,7 +1362,7 @@ void llama_context::output_reorder() {
 //
 
 uint32_t llama_context::graph_max_nodes() const {
-    return std::max<uint32_t>(16384, 512u*model.n_tensors());
+    return std::max<uint32_t>(8192, 128u*model.n_tensors());
 }
 
 llm_graph_result * llama_context::get_gf_res_reserve() const {