Update app.py

app.py

@@ -3,29 +3,39 @@ from huggingface_hub import hf_hub_download
 from llama_cpp import Llama
 
 # Download the model from Hugging Face Hub
-print("Downloading model...")
+print("===== Downloading model... =====")
 model_path = hf_hub_download(
-    repo_id="unsloth/gemma-3-4b-it-GGUF",
-    filename="gemma-3-4b-it-UD-Q4_K_XL.gguf"
+    repo_id="bartowski/Dolphin3.0-Llama3.2-3B-GGUF",
+    filename="Dolphin3.0-Llama3.2-3B-Q4_K_M.gguf"
 )
-print("Model downloaded to:", model_path)
+print(f"Model downloaded to: {model_path}")
 
 # Load the model with llama-cpp-python
-print("Loading model...")
+print("===== Loading model... =====")
 llm = Llama(
     model_path=model_path,
-    n_ctx=2048  # You can adjust this if needed
+    n_ctx=2048,          # adjust as per RAM
+    n_threads=8          # adjust based on your Space CPU (8 is good default)
 )
 print("Model loaded.")
 
-# Chat function
-def chat(prompt):
-    print(f"User input: {prompt}")
+# Chat function with Dolphin 3.0 template
+def chat(user_input):
+    print(f"User input: {user_input}")
+    full_prompt = f"""### System:
+You are Dolphin 3.0, a helpful and friendly AI assistant.
+
+### User:
+{user_input}
+
+### Assistant:"""
+    
     output = llm(
-        prompt,
+        full_prompt,
         max_tokens=512,
-        stop=["</s>", "User:", "Assistant:"]
+        stop=["</s>", "### User:", "### Assistant:"]
     )
+    
     reply = output['choices'][0]['text'].strip()
     print(f"Model reply: {reply}")
     return reply
@@ -35,8 +45,8 @@ iface = gr.Interface(
     fn=chat,
     inputs="text",
     outputs="text",
-    title="Llama 2 7B Chat GGUF - Space",
-    description="Running Llama 2 7B Chat GGUF model using llama-cpp-python on Hugging Face Space"
+    title="🐬 Dolphin 3.0 - Llama 3.2 3B GGUF Chat",
+    description="Running Dolphin 3.0 Llama 3.2 3B GGUF model using llama-cpp-python on Hugging Face Space"
 )
 
 iface.launch()