Bugfix: only eval new tokens

abetlen · Apr 15, 2023 · 89856ef · 89856ef
1 parent 887f3b7
commit 89856ef
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/llama_cpp/llama.py b/llama_cpp/llama.py
@@ -280,6 +280,7 @@ def generate(
             if self.verbose:
                 print("generate cache hit", file=sys.stderr)
             reset = False
+            tokens = tokens[len(self.tokens) :]
         ###
         if reset:
             self.reset()