Update app.py

app.py

@@ -93,7 +93,7 @@ def embedding_shift(original_embedding,shift_embeddings,prefix_embedding,suffix_
     return input_embeddings
 def engine(input_embeds):
     output_text = []
-    batch_size = 20
+    batch_size = 10
     with torch.no_grad():
         for start in range(0,len(input_embeds),batch_size):
             batch_input_embeds = input_embeds[start:start+batch_size]
@@ -160,7 +160,7 @@ def gradient_cuff_reject(message,with_defense, sample_times,perturb_times,thresh
     return False
 
 def chat(message, history, with_defense,perturb_times):
-    sample_times=20
+    sample_times=10
     threshold=thresholds[perturb_times-1]
     if gradient_cuff_reject(message,with_defense, sample_times, perturb_times, threshold):
         answer="[Gradient Cuff Rejection] I cannot fulfill your request".split(" ")
@@ -184,7 +184,7 @@ def chat(message, history, with_defense,perturb_times):
     generate_kwargs = dict(
         model_inputs,
         streamer=streamer,
-        max_new_tokens=256,
+        max_new_tokens=10,
         do_sample=True,
         top_p=0.90,
         temperature=0.6,