Update app.py

app.py

@@ -3,6 +3,50 @@ from transformers import pipeline
 import logging
 import torch
 import numpy as np
+from huggingface_hub import InferenceClient
+import gradio as gr
+
+client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
+
+def format_prompt(message, history):
+  prompt = "[Instuction]:\n"
+  for user_prompt, bot_response in history:
+    prompt += f"\n{user_prompt}: {bot_response}"
+  
+  prompt += f"\n{message}<|endoftext|>\n"
+  return prompt
+
+def generate(input_text, history, system_prompt, temperature, max_length, top_p, repetition_penalty):
+    input_text = input_text.strip()
+    history = history[:-1] if input_text == '[continue conversation]' else history
+    history = list(filter(lambda x: x[0] != '', history))
+
+    system_prompt = system_prompt.replace("[INST]", "[Instruction]").strip() + "\n"
+    prompt = format_prompt(input_text, history)
+    
+    result = client.generate(
+      prompt,
+      max_length=int(max_length),
+      temperature=float(temperature),
+      top_p=float(top_p),
+      repetition_penalty=float(repetition_penalty),
+      num_return_sequences=1,
+      do_sample=True
+    )[0]['generated_text'].strip()
+
+    return {"history": history + [(input_text, result)], "result": result}
+
+iface = gr.Interface(fn=generate,
+                     inputs=gr.Inputs(text="input_text",
+                                       textarea="system_prompt",
+                                       sliders={"temperature": (0.0, 1.0, 0.1),
+                                                "max_length": (20, 256, 1),
+                                                "top_p": (0.1, 1.0, 0.1),
+                                                "repetition_penalty": (1.0, 2.0, 0.1)}),
+                     outputs="markdown",
+                     interpretation="notext",
+                     examples={"My first question": ["input_text": "How old are you?", "system_prompt": "", "temperature": 0.5, "max_length": 50, "top_p": 0.5, "repetition_penalty": 1.1}},
+                     allow_flagging="never")
 
 # Logging Setup
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')