Fix time_elapsed issue

app.py

@@ -223,12 +223,19 @@ def diffusion_chat(question, max_it, pause_length, sharpness, clustering, noise_
     last_tokens = []
     prev_decoded_tokens = []
 
+    generation_start = time.time()
+
     for i in range(max_it):
         print('Generating output')
 
         # Model step
         generated_tokens, confidences = generate_diffusion_text(current_tokens)
 
+        elapsed = time.time() - generation_start
+        remaining = pause_length - elapsed
+        if remaining > 0:
+            time.sleep(remaining)
+
         # Save full output for noising step
         current_tokens = ori_input_tokens[:answer_start] + generated_tokens[answer_start:]
 
@@ -289,7 +296,7 @@ def diffusion_chat(question, max_it, pause_length, sharpness, clustering, noise_
         current_tokens = ori_input_tokens[:answer_start] + noised_answer[answer_start:]
         
         yield f"<b>Iteration {i+1}/{max_it} (before noising):</b><br>" + "".join(highlighted).replace('\n', '<br>')
-        # time.sleep(pause_length)
+        generation_start = time.time()
 
 
     answer_ids = current_tokens[answer_start:]