Update utils.py

utils.py

@@ -388,9 +388,17 @@ def query(api_llm, payload):
 def llm_chain2(prompt, context):
     full_prompt = RAG_CHAIN_PROMPT.format(context=context, question=prompt)
     inputs = tokenizer_rag(full_prompt, return_tensors="pt", max_length=1024, truncation=True)
-
+    attention_mask = (inputs != tokenizer_rag.pad_token_id).long()
     #Generiere die Antwort
-    outputs = modell_rag.generate(inputs['input_ids'], max_new_tokens=1024, num_beams=2, early_stopping=True)
+    outputs = modell_rag.generate(
+        inputs, 
+        attention_mask=attention_mask,
+        max_new_tokens=1024, 
+        do_sample=True, 
+        temperature=0.9, 
+        pad_token_id=tokenizer.eos_token_id
+    )
+    #outputs = modell_rag.generate(inputs['input_ids'], max_new_tokens=1024, num_beams=2, early_stopping=True)
     answer = tokenizer_rag.decode(outputs[0], skip_special_tokens=True)
     
     return answer