info string changes

app.py

@@ -29,21 +29,21 @@ from models import (
 model_labels_list = list(model_labels)
 
 # load and warm up (compile) all the models
-models = []
-for preset in model_presets:
-    model = load_model(preset)
-    chat_template = get_appropriate_chat_template(preset)
-    chat_state = ChatState(model, "", chat_template)
-    prompt, response = chat_state.send_message("Hello")
-    print("model " + preset + " loaded and initialized.")
-    print("The model responded: " + response)
-    models.append(model)
+# models = []
+# for preset in model_presets:
+#     model = load_model(preset)
+#     chat_template = get_appropriate_chat_template(preset)
+#     chat_state = ChatState(model, "", chat_template)
+#     prompt, response = chat_state.send_message("Hello")
+#     print("model " + preset + " loaded and initialized.")
+#     print("The model responded: " + response)
+#     models.append(model)
 
 # For local debugging
-# model = keras_hub.models.Llama3CausalLM.from_preset(
-#     "hf://meta-llama/Llama-3.2-1B-Instruct", dtype="bfloat16"
-# )
-# models = [model, model, model, model, model]
+model = keras_hub.models.Llama3CausalLM.from_preset(
+    "hf://meta-llama/Llama-3.2-1B-Instruct", dtype="bfloat16"
+)
+models = [model, model, model, model, model]
 
 
 def chat_turn_assistant_1(
@@ -184,10 +184,9 @@ with gr.Blocks(fill_width=True, title="Keras demo") as demo:
         gr.HTML(
             "<H2> Battle of the Keras chatbots on TPU</H2>"
             + "All the models are loaded into the TPU memory. "
-            + "You can call any of them and compare their answers. <br/>"
-            + "The entire chat history is fed to the models at every submission. "
-            + "This demo is runnig on a Google TPU v5e 2x4 (8 cores). <br/>",
-            + "All models run on `bfloat16` precision.",            
+            + "You can call any of them and compare their answers. "
+            + "The entire chat<br/>history is fed to the models at every submission. "
+            + "This demo is runnig on a Google TPU v5e 2x4 (8 cores) in bfloat16 precision."
         )
     with gr.Row():
         sel1, sel2 = instantiate_select_boxes(0, 1, model_labels_list)