Update app.py

app.py

@@ -107,21 +107,21 @@ def respond(
     print(f"Converted input_ids dtype: {input_ids.dtype}")
     input_str= str(input_ids2)
     print('input str = ', input_str)
-## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
-    # with autocast():
-    gen_tokens = model.generate(
-    input_ids, 
-    max_new_tokens=max_tokens, 
-    # do_sample=True, 
-    temperature=temperature,
-    )
-
-    gen_text = tokenizer.decode(gen_tokens[0])
-    print(gen_text)
-    gen_text= gen_text.replace(input_str,'')
-    gen_text= gen_text.replace('<|END_OF_TURN_TOKEN|>','')
+
+    
+    # gen_tokens = model.generate(
+    # input_ids, 
+    # max_new_tokens=max_tokens, 
+    # # do_sample=True, 
+    # temperature=temperature,
+    # )
+
+    # gen_text = tokenizer.decode(gen_tokens[0])
+    # print(gen_text)
+    # gen_text= gen_text.replace(input_str,'')
+    # gen_text= gen_text.replace('<|END_OF_TURN_TOKEN|>','')
     
-    yield gen_text
+    # yield gen_text
    
   
 #     messages = [