try beams

ai_single_response.py

@@ -118,7 +118,7 @@ def query_gpt_model(
     batch_size=64,
     verbose=False,
     use_gpu=False,
-    beams=2,
+    beams=4,
 ):
     """
     query_gpt_model - the main function that calls the model.
@@ -156,17 +156,21 @@ def query_gpt_model(
     print("\n... generating...")
     this_result = ai.generate(
         n=1,
-        top_k=kparam,
         batch_size=batch_size,
         # the prompt input counts for text length constraints
         max_length=resp_length + pr_len,
         min_length=resp_min + pr_len,
         prompt=this_prompt,
-        temperature=temp,
-        top_p=top_p,
-        do_sample=True,
+        num_beams=beams,
+        early_stopping=True,
+        no_repeat_ngram_size=2,
+        # temperature=temp,
+        # top_k=kparam,
+        # top_p=top_p,
+        # do_sample=True,
+        # use_cache=True,
         return_as_list=True,
-        use_cache=True,
+
     )
     if verbose:
         print("\n... generated:\n")