Update app.py

app.py

@@ -2,19 +2,25 @@ import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 
 # ---------------- CONFIG ----------------
-MODEL_NAME = "google/gemma-2b-it-pytorch"  # instruction-tuned Gemma 2B
+MODEL_NAME = "mosaicml/mpt-7b-instruct"  # Fully open-source, no gated access
 SYSTEM_PROMPT_DEFAULT = (
     "You are a formal and polite AI assistant. "
     "Always respond appropriately depending on the selected explanation style."
 )
 
 MAX_NEW_TOKENS_DEFAULT = 256
-TEMP_DEFAULT = 0.3
+TEMP_DEFAULT = 0.7
 TOP_P_DEFAULT = 0.9
 
 # ---------------- LOAD MODEL ----------------
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
+
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_NAME,
+    trust_remote_code=True,  # required for MPT
+    torch_dtype="auto"
+)
+
 generator = pipeline(
     "text-generation",
     model=model,
@@ -61,7 +67,7 @@ def chat(user_message, chat_history, system_message, max_tokens, temperature, to
 
 # ---------------- UI ----------------
 with gr.Blocks(theme=gr.themes.Soft(primary_hue="violet", secondary_hue="pink")) as demo:
-    gr.Markdown("# 🧠 Gemma 2B Chat Assistant")
+    gr.Markdown("# 🧠 MPT-7B Instruct Chat Assistant")
 
     chatbot = gr.Chatbot(type="messages", height=500, show_copy_button=True)