Fix streaming hang on last token when cache is on.

Author: abetlen

llama_cpp/llama.py

@@ -848,11 +848,6 @@ def _create_completion(
                 finish_reason = "length"
                 break
 
-        if self.cache:
-            if self.verbose:
-                print("Llama._create_completion: cache save", file=sys.stderr)
-            self.cache[prompt_tokens + completion_tokens] = self.save_state()
-
         if self.verbose:
             llama_cpp.llama_print_timings(self.ctx)
 
@@ -941,8 +936,17 @@ def _create_completion(
                         }
                     ],
                 }
+            if self.cache:
+                if self.verbose:
+                    print("Llama._create_completion: cache save", file=sys.stderr)
+                self.cache[prompt_tokens + completion_tokens] = self.save_state()
             return
 
+        if self.cache:
+            if self.verbose:
+                print("Llama._create_completion: cache save", file=sys.stderr)
+            self.cache[prompt_tokens + completion_tokens] = self.save_state()
+
         text_str = text.decode("utf-8", errors="ignore")
 
         if echo: