Merge commit '467576b6cc7d2b9220f55bc635aa51469cf26fb5' into concedo_experimental

# Conflicts:
#	.gitignore
#	Makefile
#	README.md
#	common/common.h
#	docs/build.md
#	examples/infill/infill.cpp
#	examples/perplexity/perplexity.cpp
#	examples/server/README.md
#	ggml/CMakeLists.txt
#	ggml/src/CMakeLists.txt
#	ggml/src/ggml-cuda/CMakeLists.txt
#	scripts/sync-ggml-am.sh
#	scripts/sync-ggml.sh
#	tests/CMakeLists.txt
#	tests/test-backend-ops.cpp
#	tests/test-opt.cpp
#	tests/test-quantize-perf.cpp

examples/main/main.cpp

@@ -63,49 +63,6 @@ static bool file_is_empty(const std::string & path) {
     return f.tellg() == 0;
 }
 
-static void write_logfile(
-    const llama_context * ctx, const common_params & params, const llama_model * model,
-    const std::vector<llama_token> & input_tokens, const std::string & output,
-    const std::vector<llama_token> & output_tokens
-) {
-    if (params.logdir.empty()) {
-        return;
-    }
-
-    const std::string timestamp = string_get_sortable_timestamp();
-
-    const bool success = fs_create_directory_with_parents(params.logdir);
-    if (!success) {
-        LOG_ERR("%s: failed to create logdir %s, cannot write logfile\n", __func__, params.logdir.c_str());
-        return;
-    }
-
-    const std::string logfile_path = params.logdir + timestamp + ".yml";
-    FILE * logfile = fopen(logfile_path.c_str(), "w");
-
-    if (logfile == NULL) {
-        LOG_ERR("%s: failed to open logfile %s\n", __func__, logfile_path.c_str());
-        return;
-    }
-
-    fprintf(logfile, "binary: main\n");
-    char model_desc[128];
-    llama_model_desc(model, model_desc, sizeof(model_desc));
-    yaml_dump_non_result_info(logfile, params, ctx, timestamp, input_tokens, model_desc);
-
-    fprintf(logfile, "\n");
-    fprintf(logfile, "######################\n");
-    fprintf(logfile, "# Generation Results #\n");
-    fprintf(logfile, "######################\n");
-    fprintf(logfile, "\n");
-
-    yaml_dump_string_multiline(logfile, "output", output.c_str());
-    yaml_dump_vector_int(logfile, "output_tokens", output_tokens);
-
-    llama_perf_dump_yaml(logfile, ctx);
-    fclose(logfile);
-}
-
 #if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
 static void sigint_handler(int signo) {
     if (signo == SIGINT) {
@@ -116,7 +73,6 @@ static void sigint_handler(int signo) {
             console::cleanup();
             LOG("\n");
             common_perf_print(*g_ctx, *g_smpl);
-            write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
 
             // make sure all logs are flushed
             LOG("Interrupted by user\n");
@@ -927,7 +883,6 @@ int main(int argc, char ** argv) {
 
     LOG("\n\n");
     common_perf_print(ctx, smpl);
-    write_logfile(ctx, params, model, input_tokens, output_ss.str(), output_tokens);
 
     common_sampler_free(smpl);
 

examples/server/public/index.html

@@ -212,6 +212,9 @@
           <details class="collapse collapse-arrow bg-base-200 mb-2 overflow-visible">
             <summary class="collapse-title font-bold">Other sampler settings</summary>
             <div class="collapse-content">
+              <!-- Samplers queue -->
+              <settings-modal-short-input label="Samplers queue" :config-key="'samplers'" :config-default="configDefault" :config-info="configInfo" v-model="config.samplers"></settings-modal-short-input>
+              <!-- Samplers -->
               <template v-for="configKey in ['dynatemp_range', 'dynatemp_exponent', 'typical_p', 'xtc_probability', 'xtc_threshold']">
                 <settings-modal-short-input :config-key="configKey" :config-default="configDefault" :config-info="configInfo" v-model="config[configKey]" />
               </template>
@@ -231,6 +234,7 @@
             <summary class="collapse-title font-bold">Advanced config</summary>
             <div class="collapse-content">
               <label class="form-control mb-2">
+                <!-- Custom parameters input -->
                 <div class="label inline">Custom JSON config (For more info, refer to <a class="underline" href="https://github.com/ggerganov/llama.cpp/blob/master/examples/server/README.md" target="_blank" rel="noopener noreferrer">server documentation</a>)</div>
                 <textarea class="textarea textarea-bordered h-24" placeholder="Example: { &quot;mirostat&quot;: 1, &quot;min_p&quot;: 0.1 }" v-model="config.custom"></textarea>
               </label>
@@ -253,7 +257,7 @@
     <label class="input input-bordered join-item grow flex items-center gap-2 mb-2">
       <!-- Show help message on hovering on the input label -->
       <div class="dropdown dropdown-hover">
-        <div tabindex="0" role="button" class="font-bold">{{ configKey }}</div>
+        <div tabindex="0" role="button" class="font-bold">{{ label || configKey }}</div>
         <div class="dropdown-content menu bg-base-100 rounded-box z-10 w-64 p-2 shadow mt-4">
           {{ configInfo[configKey] || '(no help message available)' }}
         </div>
@@ -282,6 +286,7 @@
       apiKey: '',
       systemMessage: 'You are a helpful assistant.',
       // make sure these default values are in sync with `common.h`
+      samplers: 'dkypmxt',
       temperature: 0.8,
       dynatemp_range: 0.0,
       dynatemp_exponent: 1.0,
@@ -305,6 +310,7 @@
     const CONFIG_INFO = {
       apiKey: 'Set the API Key if you are using --api-key option for the server.',
       systemMessage: 'The starting message that defines how model should behave.',
+      samplers: 'The order at which samplers are applied, in simplified way. Default is "dkypmxt": dry->top_k->typ_p->top_p->min_p->xtc->temperature',
       temperature: 'Controls the randomness of the generated text by affecting the probability distribution of the output tokens. Higher = more random, lower = more focused.',
       dynatemp_range: 'Addon for the temperature sampler. The added value to the range of dynamic temperature, which adjusts probabilities by entropy of tokens.',
       dynatemp_exponent: 'Addon for the temperature sampler. Smoothes out the probability redistribution based on the most probable token.',
@@ -352,10 +358,16 @@
       { props: ["source"] }
     );
 
-    // inout field to be used by settings modal
+    // input field to be used by settings modal
     const SettingsModalShortInput = defineComponent({
       template: document.getElementById('settings-modal-short-input').innerHTML,
-      props: ['configKey', 'configDefault', 'configInfo', 'modelValue'],
+      props: {
+        label: { type: String, required: false },
+        configKey: String,
+        configDefault: Object,
+        configInfo: Object,
+        modelValue: [Object, String, Number],
+      },
     });
 
     // coversations is stored in localStorage
@@ -546,6 +558,7 @@
               ],
               stream: true,
               cache_prompt: true,
+              samplers: this.config.samplers,
               temperature: this.config.temperature,
               dynatemp_range: this.config.dynatemp_range,
               dynatemp_exponent: this.config.dynatemp_exponent,

examples/server/server.cpp

@@ -928,14 +928,22 @@ struct server_context {
 
         {
             const auto & samplers = data.find("samplers");
-            if (samplers != data.end() && samplers->is_array()) {
-                std::vector<std::string> sampler_names;
-                for (const auto & name : *samplers) {
-                    if (name.is_string()) {
-                        sampler_names.emplace_back(name);
+            if (samplers != data.end()) {
+                if (samplers->is_array()) {
+                    std::vector<std::string> sampler_names;
+                    for (const auto & name : *samplers) {
+                        if (name.is_string()) {
+                            sampler_names.emplace_back(name);
+                        }
                     }
+                    slot.sparams.samplers = common_sampler_types_from_names(sampler_names, false);
+                } else if (samplers->is_string()){
+                    std::string sampler_string;
+                    for (const auto & name : *samplers) {
+                        sampler_string += name;
+                    }
+                    slot.sparams.samplers = common_sampler_types_from_chars(sampler_string);
                 }
-                slot.sparams.samplers = common_sampler_types_from_names(sampler_names, false);
             } else {
                 slot.sparams.samplers = default_sparams.samplers;
             }