Update app.py

app.py

@@ -1,23 +1,59 @@
-# Load model directly
 import streamlit as st
-import transformers
+from huggingface_hub import InferenceClient
 
-st.title("A Simple Interface for a Language Model")
+# Initialize the client
+client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 
-st.subheader("Input Text")
-input_text = st.text_area("Enter your text here", "Type something here...")
+# Function to format the prompt
+def format_prompt(message, history):
+    prompt = "<s>"
+    for user_prompt, bot_response in history:
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
+    prompt += f"[INST] {message} [/INST]"
+    return prompt
 
-if st.button("Generate Response"):
-    # Initialize tokenizer and model
-    tokenizer = transformers.AutoTokenizer.from_pretrained("microsoft/phi-2")
-    model = transformers.AutoModelForCausalLM.from_pretrained("microsoft/phi-2")
+# Function to generate response
+def generate(prompt, history, temperature=0.2, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0):
+    temperature = max(float(temperature), 1e-2)
+    top_p = float(top_p)
 
-    # Encode input text
-    inputs = tokenizer(input_text, return_tensors="pt")
+    generate_kwargs = dict(
+        temperature=temperature,
+        max_new_tokens=max_new_tokens,
+        top_p=top_p,
+        repetition_penalty=repetition_penalty,
+        do_sample=True,
+        seed=42,
+    )
 
-    # Generate response
-    response = model.generate(**inputs, max_length=100, do_sample=True)
+    formatted_prompt = format_prompt(prompt, history)
 
-    # Decode response
-    st.subheader("Generated Response")
-    st.write(tokenizer.decode(response[0]))
+    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
+    output = ""
+    for response in stream:
+        output += response.token.text
+    return output
+
+# Streamlit interface
+st.title("Mistral 8x7b Chat")
+
+# Chat history
+if 'history' not in st.session_state:
+    st.session_state.history = []
+
+# User input
+user_input = st.text_input("Your message:", key="user_input")
+
+# Generate response and update history
+if st.button("Send"):
+    if user_input:
+        bot_response = generate(user_input, st.session_state.history)
+        st.session_state.history.append((user_input, bot_response))
+        st.session_state.user_input = ""
+
+# Display conversation
+chat_text = ""
+for user_msg, bot_msg in st.session_state.history:
+    chat_text += f"You: {user_msg}\nBot: {bot_msg}\n\n"
+st.text_area("Chat", value=chat_text, height=300, disabled=True)