Update app.py

app.py

@@ -4,22 +4,22 @@ import gradio as gr
 client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 
 def format_prompt(message, history, system_prompt=None):
-prompt = "<s>"
-for user_prompt, bot_response in history:
-prompt += f"[INST] {user_prompt} [/INST]"
-prompt += f" {bot_response}</s> "
-if system_prompt:
-prompt += f"[SYS] {system_prompt} [/SYS]"
-prompt += f"[INST] {message} [/INST]"
-return prompt
+    prompt = "<s>"
+    for user_prompt, bot_response in history:
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
+    if system_prompt:
+        prompt += f"[SYS] {system_prompt} [/SYS]"
+    prompt += f"[INST] {message} [/INST]"
+    return prompt
 
 def generate(
-prompt, history, system_prompt=None, temperature=0.3, max_new_tokens=8192, top_p=0.9, repetition_penalty=1.0,
+    prompt, history, system_prompt=None, temperature=0.2, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0,
 ):
-temperature = float(temperature)
-if temperature < 1e-2:
-temperature = 1e-2
-top_p = float(top_p)
+    temperature = float(temperature)
+    if temperature < 1e-2:
+        temperature = 1e-2
+    top_p = float(top_p)
 
     generate_kwargs = dict(
         temperature=temperature,