Update app.py

app.py

@@ -1,21 +1,34 @@
 import os
-import gradio as gr
 import copy
+import gradio as gr
 from llama_cpp import Llama
-from huggingface_hub import hf_hub_download  
+from huggingface_hub import hf_hub_download
 
+# Fix for Python 3.13: audioop was removed from the standard library.
+# This try/except block handles the missing dependency if audioop-lts is installed.
+try:
+    import audioop
+except ImportError:
+    try:
+        import audioop_lts as audioop
+    except ImportError:
+        print("Warning: audioop not found. If Gradio fails to load, install 'audioop-lts'.")
 
-llm = Llama(
-    model_path=hf_hub_download(
-        repo_id=os.environ.get("REPO_ID", "unsloth/NVIDIA-Nemotron-3-Nano-4B-GGUF:Q4_K_M"),# Run inference directly in the terminal:
+# 1. Download the model correctly
+# Repo: unsloth/NVIDIA-Nemotron-3-Nano-4B-GGUF
+# File: NVIDIA-Nemotron-3-Nano-4B-Q4_K_M.gguf
+model_path = hf_hub_download(
+    repo_id=os.environ.get("REPO_ID", "unsloth/NVIDIA-Nemotron-3-Nano-4B-GGUF"),
+    filename=os.environ.get("MODEL_FILE", "NVIDIA-Nemotron-3-Nano-4B-Q4_K_M.gguf"),
+)
 
-        filename=os.environ.get("MODEL_FILE", "NVIDIA-Nemotron-3-Nano-4B-GGUF:Q4_K_M"),
-    ),
+# 2. Initialize the Llama model
+llm = Llama(
+    model_path=model_path,
     n_ctx=2048,
-    n_gpu_layers=50, # change n_gpu_layers if you have more or less VRAM 
+    n_gpu_layers=-1, # -1 uses all available GPU layers, change to 0 for CPU only
 ) 
 
-
 def generate_text(
     message,
     history: list[tuple[str, str]],
@@ -25,11 +38,12 @@ def generate_text(
     top_p,
 ):
     temp = ""
+    # Standard ChatML / Llama format logic
     input_prompt = f"[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n "
     for interaction in history:
-        input_prompt = input_prompt + str(interaction[0]) + " [/INST] " + str(interaction[1]) + " </s><s> [INST] "
-
-    input_prompt = input_prompt + str(message) + " [/INST] "
+        input_prompt += f"{interaction[0]} [/INST] {interaction[1]} </s><s> [INST] "
+    
+    input_prompt += f"{message} [/INST] "
 
     output = llm(
         input_prompt,
@@ -39,51 +53,38 @@ def generate_text(
         repeat_penalty=1.1,
         max_tokens=max_tokens,
         stop=[
-            "<|prompter|>",
+            "[/INST]",
+            "</s>",
             "<|endoftext|>",
-            "<|endoftext|> \n",
-            "ASSISTANT:",
             "USER:",
-            "SYSTEM:",
+            "ASSISTANT:",
         ],
         stream=True,
     )
+    
     for out in output:
         stream = copy.deepcopy(out)
         temp += stream["choices"][0]["text"]
         yield temp
 
-
+# 3. Define the Gradio Interface
 demo = gr.ChatInterface(
     generate_text,
-    title="llama-cpp-python on GPU",
-    description="Running LLM with https://github.com/abetlen/llama-cpp-python",
+    title="NVIDIA Nemotron-3 Nano (Llama-cpp)",
+    description="Running NVIDIA Nemotron-3-Nano-4B via llama-cpp-python",
     examples=[
         ['How to setup a human base on Mars? Give short answer.'],
         ['Explain theory of relativity to me like I’m 8 years old.'],
-        ['What is 9,000 * 9,000?'],
-        ['Write a pun-filled happy birthday message to my friend Alex.'],
-        ['Justify why a penguin might make a good king of the jungle.']
+        ['What is 9,000 * 9,000?']
     ],
     cache_examples=False,
-    retry_btn=None,
-    undo_btn="Delete Previous",
-    clear_btn="Clear",
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Textbox(value="You are a helpful and friendly AI assistant.", label="System message"),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p"),
     ],
 )
 
-
 if __name__ == "__main__":
-    demo.launch()
-
+    demo.launch(server_name="0.0.0.0", server_port=7860)