Enable direct prompting

app.py

@@ -203,13 +203,13 @@ def diffusion_chat(question, eot_weight, mask_weight, max_it, pause_length, shar
 
     print('started generation')
     prompt = f"User: {question}\nAssistant:"
-    prompt = ""
+    prompt = question
     input_ids = tokenizer.encode(prompt, add_special_tokens=False)
     answer_start = find_answer_start(input_ids, assistant_marker_ids)
     # if answer_start is None:
     #     yield "Error: Could not find Assistant marker in input."
     #     return
-    answer_start = 0
+    answer_start = len(input_ids)
     
     if len(input_ids) < 256:
         input_ids += [pad_token] * (256 - len(input_ids))