Update app.py

app.py

@@ -4,36 +4,42 @@ import gradio as gr
 from llama_cpp import Llama
 import random
 
-url = 'https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q2_K.bin'
+url = 'https://huggingface.co/TheBloke/WizardLM-7B-uncensored-GGML/resolve/main/WizardLM-7B-uncensored.ggmlv3.q4_0.bin'
 filename = wget.download(url)
 llm = Llama(model_path=filename, seed=random.randint(1, 2**31))
-
-with gr.Blocks() as demo:
+theme = gr.themes.Soft(
+    primary_hue=gr.themes.Color("#ededed", "#fee2e2", "#fecaca", "#fca5a5", "#f87171", "#ef4444", "#dc2626", "#b91c1c", "#991b1b", "#7f1d1d", "#6c1e1e"),
+    neutral_hue="red",
+)
+title = """<h1 align="center">Chat with awesome WizardLM 7b model!</h1><br><h2 align="center">Llama cpp + Gradio!</h2>"""
+with gr.Blocks(theme=theme) as demo:
+    gr.HTML(title)
+    gr.HTML("This model is awesome for its size! It is only 20th the size of Chatgpt but is around 90% as good as Chatgpt. However, please don't rely on WizardLM 7b to provide 100% true information as it might be wrong sometimes. ")
     chatbot = gr.Chatbot()
     msg = gr.Textbox()
-    clear = gr.UploadButton([msg, chatbot])  # Replace gr.ClearButton with gr.UploadButton
-    # instruction = gr.Textbox(label="Instruction", placeholder="")
+    clear = gr.ClearButton([msg, chatbot])
+    #instruction = gr.Textbox(label="Instruction", placeholder=)
 
     def user(user_message, history):
         return gr.update(value="", interactive=True), history + [[user_message, None]]
 
     def bot(history):
-        # instruction = history[-1][1] or ""
+        #instruction = history[-1][1] or ""
         user_message = history[-1][0]
-        # token1 = llm.tokenize(b"### Instruction: ")
-        # token2 = llm.tokenize(instruction.encode())
-        token3 = llm.tokenize(b"### Input: ")
-        tokens3 = llm.tokenize(user_message.encode())
-        token4 = llm.tokenize(b"### Response:")
-        tokens = token3 + tokens3 + token4
+        #token1 = llm.tokenize(b"### Instruction: ")
+        #token2 = llm.tokenize(instruction.encode())
+        #token3 = llm2.tokenize(b"USER: ")
+        tokens3 = llm2.tokenize(user_message.encode())
+        token4 = llm2.tokenize(b"\n\n### Response:")
+        tokens = tokens3 + token4
         history[-1][1] = ""
         count = 0
         output = ""
-        for token in llm.generate(tokens, top_k=50, top_p=0.73, temp=0.72, repeat_penalty=1.1):
-            text = llm.detokenize([token])
+        for token in llm2.generate(tokens, top_k=50, top_p=0.73, temp=0.72, repeat_penalty=1.1):
+            text = llm2.detokenize([token])
             output += text.decode()
             count += 1
-            if count >= 500 or (token == llm.token_eos()):
+            if count >= 500 or (token == llm2.token_eos()):
                 break
             history[-1][1] += text.decode()
             yield history