Update app.py

app.py

@@ -2,8 +2,10 @@ from transformers import AutoModel, AutoTokenizer
 import gradio as gr
 import mdtex2html
 
-tokenizer = AutoTokenizer.from_pretrained("fb700/chatglm-fitness-RLHF", trust_remote_code=True)
-model = AutoModel.from_pretrained("fb700/chatglm-fitness-RLHF", trust_remote_code=True).quantize(8).half().cuda()
+checkpoint = "fb700/chatglm-fitness-RLHF"
+
+tokenizer = AutoTokenizer.from_pretrained(checkpoint, trust_remote_code=True)
+model = AutoModel.from_pretrained(checkpoint, trust_remote_code=True).quantize(4).half().cuda()
 model = model.eval()
 
 """Override Chatbot.postprocess"""
@@ -74,7 +76,7 @@ def reset_state():
 
 
 with gr.Blocks() as demo:
-    gr.HTML("""<h1 align="center">帛凡 ChatGLM-6b-fitness-RLHF  演示</h1>""")
+    gr.HTML("""<h1 align="center">帛凡 ChatGLM-6b-fitness-RLHF  演示 </h1>""")
 
     chatbot = gr.Chatbot()
     with gr.Row():
@@ -86,7 +88,7 @@ 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, 409600, value=2048, step=1.0, label="Maximum length", interactive=True)
+            max_length = gr.Slider(0, 40960, value=20480, step=1.0, label="Maximum length", interactive=True)
             top_p = gr.Slider(0, 1, value=0.7, step=0.01, label="Top P", interactive=True)
             temperature = gr.Slider(0, 1, value=0.95, step=0.01, label="Temperature", interactive=True)
 
@@ -98,4 +100,4 @@ with gr.Blocks() as demo:
 
     emptyBtn.click(reset_state, outputs=[chatbot, history], show_progress=True)
 
-demo.queue().launch(share=True, inbrowser=True)
+demo.queue().launch(share=False, inbrowser=True)