Update app.py

app.py

@@ -1,40 +1,47 @@
 import gradio as gr
-import spaces
 from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
 CSS = """
 #qwen-md .katex-display { display: inline; }
-#qwen-md .katex-display>.katex { display: inline; }
-#qwen-md .katex-display>.katex>.katex-html { display: inline; }
 """
 
+# 下载GGUF模型文件（如果本地已下载，此代码会跳过）
 hf_hub_download(
     repo_id="bartowski/Qwen2.5-Math-7B-Instruct-GGUF",
     filename="Qwen2.5-Math-7B-Instruct-Q6_K_L.gguf",
     local_dir="./models",
 )
 
-# 模型加载函数明确包装GPU环境
-@spaces.GPU
-def load_llm():
-    llm = Llama(
-        model_path="models/Qwen2.5-Math-7B-Instruct-Q6_K_L.gguf",
-        flash_attn=True,
-        n_ctx=4096,
-        n_batch=512,
-        chat_format="chatml",
-        n_gpu_layers=-1,
-    )
-    return llm
+# 仅CPU模式加载模型，去除flash attention和显存优化参数
+llm = Llama(
+    model_path="models/Qwen2.5-Math-7B-Instruct-Q6_K_L.gguf",
+    n_ctx=2048,   # 更短的上下文提高速度
+    n_batch=256,  # 小批量减少CPU负载
+    n_threads=8,  # 明确线程数 (可根据你的CPU核心数调整)
+    chat_format="chatml",
+    verbose=False
+)
 
-# 全局加载模型（仅加载一次，提升速度）
-llm = load_llm()
+# Gradio 组件定义（简化配置）
+input_text = gr.Textbox(label="Ask math questions here")
+output_md = gr.Markdown(label="Answer", elem_id="qwen-md", show_copy_button=True)
+target_lang = gr.Dropdown(choices=["Chinese", "English"], value="Chinese", label="Output Language")
+new_tokens = gr.Slider(minimum=1, maximum=512, value=256, step=1, label="Max new tokens")
+temperature = gr.Slider(minimum=0, maximum=1.0, value=0.2, step=0.05, label="Temperature")
+top_p = gr.Slider(minimum=0.0, maximum=1.0, value=0.9, step=0.05, label="Top P")
+submit_btn = gr.Button("Ask")
+banner = gr.Markdown(value="### 📖 **Qwen2.5-Math 7B GGUF** - Optimized for CPU")
+
+# 响应函数（精简版，更快速）
+def respond(input_text, lang, max_tokens, temperature, top_p):
+    sys_msg = "你是一个乐于助人的数学助手，请使用中文回答。" if lang == "Chinese" else "You are a helpful math assistant. Please answer in English."
+    messages = [
+        {"role": "system", "content": sys_msg},
+        {"role": "user", "content": input_text},
+    ]
 
-# 模型推理函数也使用GPU环境
-@spaces.GPU
-def generate_response(messages, max_tokens, temperature, top_p):
-    response = llm.create_chat_completion(
+    stream_response = llm.create_chat_completion(
         messages=messages,
         stream=True,
         max_tokens=max_tokens,
@@ -42,69 +49,11 @@ def generate_response(messages, max_tokens, temperature, top_p):
         top_p=top_p,
     )
 
-    message_repl = ""
-    for chunk in response:
-        delta = chunk['choices'][0]["delta"]
-        if "content" in delta:
-            message_repl += delta["content"]
-            yield message_repl
-
-# Gradio组件设置
-output_md = gr.Markdown(
-    label="Answer",
-    value="Answer will be presented here",
-    latex_delimiters=[
-        {"left": "\\(", "right": "\\)", "display": True},
-        {"left": "\\[", "right": "\\]", "display": True},
-    ],
-    elem_id="qwen-md",
-    show_copy_button=True,
-    container=True,
-    render=False,
-)
-
-target_lang = gr.Dropdown(
-    choices=["Chinese", "English"],
-    value="Chinese",
-    label="Output Language",
-    interactive=True,
-    render=False,
-)
-
-new_tokens = gr.Slider(
-    minimum=1, maximum=1024, value=128, step=1, label="Max new tokens", render=False
-)
-
-temperature = gr.Slider(
-    minimum=0, maximum=1.0, value=0.2, step=0.1, label="Temperature", render=False
-)
-
-top_p = gr.Slider(
-    minimum=0.0, maximum=1.0, value=0.8, step=0.05, label="Top P", render=False
-)
-
-input_text = gr.Textbox(label="Ask math questions here", render=False)
-submit_btn = gr.Button(value="Ask", render=False)
-
-banner = gr.Markdown(value="""
-# 📖 Qwen2.5-Math GGUF
-Fast mathematical reasoning with ZeroGPU optimized.
-""")
-
-# Gradio响应函数，调用GPU推理
-def respond(input_text, lang, max_tokens, temperature, top_p):
-    sys_msg = (
-        "你是一个乐于助人的数学助手. 你使用中文回答问题"
-        if lang == "Chinese" else
-        "You are a helpful math assistant. You should always provide your answer in English."
-    )
-    messages = [
-        {"role": "system", "content": sys_msg},
-        {"role": "user", "content": input_text},
-    ]
-
-    # GPU生成器直接返回
-    yield from generate_response(messages, max_tokens, temperature, top_p)
+    result = ""
+    for chunk in stream_response:
+        content = chunk['choices'][0]["delta"].get("content", "")
+        result += content
+        yield result.strip()
 
 with gr.Blocks(css=CSS, theme="NoCrypt/miku") as demo:
     submit_btn.click(