Fix import

app.py

@@ -70,16 +70,31 @@ def parse_text(text):
     text = "".join(lines)
     return text
 
+stream = True
 
 def predict(input, chatbot, max_length, top_p, temperature, history, past_key_values):
+    model.generation_config.temperature = temperature
+    model.generation_config.top_p = top_p
+    model.generation_config.max_new_tokens = max_length
     chatbot.append((parse_text(input), ""))
-    for response, history, past_key_values in model.stream_chat(tokenizer, input, history, past_key_values=past_key_values,
-                                                                return_past_key_values=True,
-                                                                max_length=max_length, top_p=top_p,
-                                                                temperature=temperature):
+
+    history.append({"role": "user", "content": parse_text(input)})
+    if stream:
+        position = 0
+        for response in model.chat(tokenizer, history, stream=True):
+            chatbot[-1] = (parse_text(input), parse_text(response))
+            if torch.backends.mps.is_available():
+                torch.mps.empty_cache()
+            yield chatbot, history, past_key_values
+        print(response)
+        history.append({"role": "assistant", "content": response})
+    else:
+        response = model.chat(tokenizer, history)
+        print(response)
         chatbot[-1] = (parse_text(input), parse_text(response))
 
-        yield chatbot, history, past_key_values
+    yield chatbot, history, past_key_values
+
 
 
 def reset_user_input():
@@ -103,9 +118,9 @@ with gr.Blocks() as demo:
                 submitBtn = gr.Button("Submit", variant="primary")
         with gr.Column(scale=1):
             emptyBtn = gr.Button("Clear History")
-            max_length = gr.Slider(0, 32768, value=8192, step=1.0, label="Maximum length", interactive=True)
-            top_p = gr.Slider(0, 1, value=0.8, step=0.01, label="Top P", interactive=True)
-            temperature = gr.Slider(0, 1, value=0.95, step=0.01, label="Temperature", interactive=True)
+            max_length = gr.Slider(0, 2048, value=1024, step=1.0, label="Maximum length", interactive=True)
+            top_p = gr.Slider(0, 1, value=0.85, step=0.01, label="Top P", interactive=True)
+            temperature = gr.Slider(0, 1, value=0.9, step=0.01, label="Temperature", interactive=True)
 
     history = gr.State([])
     past_key_values = gr.State(None)