Add repetition_penalty

modules/llamacpp_model.py

@@ -48,7 +48,7 @@ class LlamaCppModel:
         params.repeat_penalty = repetition_penalty
         #params.repeat_last_n = repeat_last_n
 
-        # model.params = params
+        #self.model.params = params
         self.model.add_bos()
         self.model.update_input(context)
 

modules/text_generation.py

@@ -130,7 +130,7 @@ def generate_reply(question, max_new_tokens, do_sample, temperature, top_p, typi
 
                 # RWKV has proper streaming, which is very nice.
                 # No need to generate 8 tokens at a time.
-                for reply in shared.model.generate_with_streaming(context=question, token_count=max_new_tokens, temperature=temperature, top_p=top_p, top_k=top_k):
+                for reply in shared.model.generate(context=question, token_count=max_new_tokens, temperature=temperature, top_p=top_p, top_k=top_k, repetition_penalty=repetition_penalty):
                     output = original_question+reply
                     if not (shared.args.chat or shared.args.cai_chat):
                         reply = original_question + apply_extensions(reply, "output")