Update app.py

app.py

@@ -1,36 +1,21 @@
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
+import webbrowser # 用於開啟網頁地圖連結
 
-# --- Configuration ---
-# This model is small enough to run on a free CPU Space
+# --- 模型配置 (CPU TinyLlama) ---
 MODEL_NAME = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
-
-# --- Model Loading ---
-print("Loading model and tokenizer...")
-# Load the model directly to CPU. No need for device_map or specific dtypes.
+print("Loading TinyLlama model and tokenizer...")
 model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-print("Model and tokenizer loaded.")
+print("TinyLlama model and tokenizer loaded.")
 
-# --- Text Generation Function ---
+# --- 文字生成函數 ---
 def generate_text(prompt, max_new_tokens, temperature, top_p):
-    """
-    Generates text using the TinyLlama model with its chat template.
-    """
-    print(f"Generating response for prompt: '{prompt}'")
-
-    # Format the prompt using the model's chat template
-    # The format is <|system|>...<|user|>...<|assistant|>
-    chat = [
-        {"role": "user", "content": prompt}
-    ]
+    print(f"Generating TinyLlama response for prompt: '{prompt}'")
+    chat = [{"role": "user", "content": prompt}]
     formatted_prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
-    
-    # Tokenize the formatted prompt
     inputs = tokenizer(formatted_prompt, return_tensors="pt")
-
-    # Generate text
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
@@ -40,47 +25,56 @@ def generate_text(prompt, max_new_tokens, temperature, top_p):
             top_p=float(top_p),
             eos_token_id=tokenizer.eos_token_id
         )
+    decoded_output = tokenizer.decode(outputs[:, inputs["input_ids"].shape[-1]:][0], skip_special_tokens=True)
+    print(f"Generated response: {decoded_output}")
+    return decoded_output.strip()
 
-    # Decode the full output
-    decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    # Extract only the assistant's response
-    # The output includes the prompt, so we find the start of the assistant's part
-    assistant_marker = "<|assistant|>"
-    assistant_start_index = decoded_output.find(assistant_marker)
-    if assistant_start_index != -1:
-        response = decoded_output[assistant_start_index + len(assistant_marker):].strip()
-    else:
-        # Fallback if the marker isn't found (should be rare)
-        response = "Could not parse the model's response."
-        
-    print(f"Generated response: {response}")
-    return response
+# --- 地圖生成函數 (開啟 Google 地圖網頁) ---
+def show_map(location_query):
+    """開啟包含指定地點的 Google 地圖網頁。"""
+    base_url = "https://www.google.com/maps/search/?api=1&query="
+    map_url = base_url + location_query.replace(" ", "+")
+    webbrowser.open_new_tab(map_url)
+    return f"已在您的預設瀏覽器中開啟 {location_query} 的地圖。"
 
-# --- Gradio Interface ---
+# --- Gradio 介面 ---
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 🦙 TinyLlama Chat (CPU Version)")
-    gr.Markdown("This Space runs `TinyLlama-1.1B-Chat-v1.0`, a compact model that runs on free CPU hardware.")
-
-    with gr.Column():
-        prompt_input = gr.Textbox(
-            label="Your Prompt", 
-            placeholder="Explain the importance of bees in the ecosystem."
-        )
-        
-        with gr.Accordion("Generation Parameters", open=False):
-            max_new_tokens_slider = gr.Slider(minimum=50, maximum=512, value=256, step=1, label="Max New Tokens")
-            temperature_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.05, label="Temperature")
-            top_p_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-P (Nucleus Sampling)")
+    gr.Markdown("# 多功能應用：文字生成與地圖顯示")
+    gr.Markdown("這個應用程式可以生成文字，也可以根據您輸入的地點顯示地圖。")
 
-        generate_button = gr.Button("Generate Response ✨", variant="primary")
-        output_text = gr.Textbox(label="Model's Response", lines=10, interactive=False)
+    with gr.TabbedInterface(["文字生成", "顯示地圖"]) as tabbed:
+        with gr.TabItem("文字生成"):
+            gr.Markdown("## 使用 TinyLlama 生成文字")
+            with gr.Column():
+                prompt_input = gr.Textbox(
+                    label="您的提示詞",
+                    placeholder="請輸入您想讓模型接續的文字..."
+                )
+                with gr.Accordion("生成參數", open=False):
+                    max_new_tokens_slider = gr.Slider(minimum=50, maximum=512, value=256, step=1, label="最大生成 Token 數")
+                    temperature_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.05, label="溫度 (Temperature)")
+                    top_p_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-P (Nucleus Sampling)")
+                generate_button = gr.Button("生成文字 ✨", variant="primary")
+                output_text = gr.Textbox(label="模型的回應", lines=10, interactive=False)
+                generate_button.click(
+                    fn=generate_text,
+                    inputs=[prompt_input, max_new_tokens_slider, temperature_slider, top_p_slider],
+                    outputs=output_text
+                )
 
-    generate_button.click(
-        fn=generate_text,
-        inputs=[prompt_input, max_new_tokens_slider, temperature_slider, top_p_slider],
-        outputs=output_text
-    )
+        with gr.TabItem("顯示地圖"):
+            gr.Markdown("## 顯示指定地點的地圖")
+            location_input = gr.Textbox(
+                label="輸入地點名稱",
+                placeholder="例如：台北 101, New York Central Park, 東京鐵塔"
+            )
+            show_map_button = gr.Button("顯示地圖 🗺️", variant="primary")
+            map_output = gr.Textbox(label="地圖顯示訊息", interactive=False)
+            show_map_button.click(
+                fn=show_map,
+                inputs=location_input,
+                outputs=map_output
+            )
 
-# --- Launch the App ---
+# --- 啟動 Gradio 應用 ---
 demo.launch()