Quick fixes by comments

common/arg.cpp

@@ -948,14 +948,14 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         }
     ).set_sparam());
     add_opt(common_arg(
-        {"-xtc-p", "--xtc-probability"}, "N",
+        {"--xtc-probability"}, "N",
         string_format("xtc probability (default: %.1f, 0.0 = disabled)", (double)params.sparams.xtc_probability),
         [](common_params & params, const std::string & value) {
             params.sparams.xtc_probability = std::stof(value);
         }
     ).set_sparam());
     add_opt(common_arg(
-        {"-xtc-t", "--xtc-threshold"}, "N",
+        {"--xtc-threshold"}, "N",
         string_format("xtc threshold (default: %.1f, 1.0 = disabled)", (double)params.sparams.xtc_threshold),
         [](common_params & params, const std::string & value) {
             params.sparams.xtc_threshold = std::stof(value);

common/common.cpp

@@ -2104,7 +2104,7 @@ void yaml_dump_non_result_info(FILE * stream, const common_params & params, cons
     fprintf(stream, "top_k: %d # default: 40\n", sparams.top_k);
     fprintf(stream, "top_p: %f # default: 0.95\n", sparams.top_p);
     fprintf(stream, "min_p: %f # default: 0.0\n", sparams.min_p);
-    fprintf(stream, "xtc_probability: %f # default: 0.5\n", sparams.xtc_probability);
+    fprintf(stream, "xtc_probability: %f # default: 0.0\n", sparams.xtc_probability);
     fprintf(stream, "xtc_threshold: %f # default: 0.1\n", sparams.xtc_threshold);
     fprintf(stream, "typ_p: %f # default: 1.0\n", sparams.typ_p);
     fprintf(stream, "verbose_prompt: %s # default: false\n", params.verbose_prompt ? "true" : "false");

common/common.h

@@ -111,7 +111,7 @@ struct common_sampler_params {
     float   top_p             = 0.95f; // 1.0 = disabled
     float   min_p             = 0.05f; // 0.0 = disabled
     float   xtc_probability   = 0.00f; // 0.0 = disabled
-    float   xtc_threshold     = 0.10f; // 0.5 = disabled
+    float   xtc_threshold     = 0.10f; // > 0.5 disables XTC
     float   tfs_z             = 1.00f; // 1.0 = disabled
     float   typ_p             = 1.00f; // typical_p, 1.0 = disabled
     float   temp              = 0.80f; // <= 0.0 to sample greedily, 0.0 to not output probabilities

examples/server/public/index-new.html

@@ -44,7 +44,7 @@
       top_p: 1.0, // 1.0 = disabled
       min_p: 0.05, // 0 = disabled; recommended for non-english: ~ 0.4
       xtc_probability: 0.0, // 0 = disabled;
-      xtc_threshold: 0.1, // 0.5 = disabled;
+      xtc_threshold: 0.1, // > 0.5 disables XTC;
       tfs_z: 1.0, // 1.0 = disabled
       typical_p: 1.0, // 1.0 = disabled
       presence_penalty: 0.0, // 0.0 = disabled

examples/server/public/index.html

@@ -308,7 +308,7 @@
       top_p: 0.95, // 1.0 = disabled
       min_p: 0.05, // 0 = disabled
       xtc_probability: 0.0, // 0 = disabled;
-      xtc_threshold: 0.1, // 0.5 = disabled;
+      xtc_threshold: 0.1, // > 0.5 disables XTC;
       tfs_z: 1.0, // 1.0 = disabled
       typical_p: 1.0, // 1.0 = disabled
       presence_penalty: 0.0, // 0.0 = disabled

src/llama-sampling.cpp

@@ -1081,7 +1081,7 @@ static void llama_sample_xtc_apply(struct llama_sampler * smpl, llama_token_data
 
     if (ctx->probability <= 0.0f
         || ctx->threshold > 0.5f
-        || cur_p->size <= 2) {
+        || cur_p->size < 2) {
         return;
     }