avoid to get prompt in infill mode and embedding mode

examples/server/server.cpp

@@ -901,6 +901,7 @@ struct server_context {
         slot.params.input_suffix = json_value(data, "input_suffix", default_params.input_suffix);
 
         // get prompt
+        if (!(task.infill || task.embedding)) {
             {
                 const auto & prompt = data.find("prompt");
                 if (prompt == data.end()) {
@@ -914,6 +915,8 @@ struct server_context {
                     return false;
                 }
             }
+        }
+        
 
         // penalize user-provided tokens
         {