Initial commit with app code only

README.md

@@ -1,5 +1,5 @@
 ---
-title: Teacher Basic Qwen3-0.6b
+title: Teacher Model Api
 emoji: 💬
 colorFrom: yellow
 colorTo: purple
@@ -7,6 +7,7 @@ sdk: gradio
 sdk_version: 5.0.1
 app_file: app.py
 pinned: false
+short_description: 一个基于qwen3-0.6b训练的的文本分类模型(调用API）
 ---
 
 An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).

app.py

@@ -1,64 +1,134 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+import torch
+from threading import Thread
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# --- 配置您的本地模型路径 ---
+MODEL_PATH = "badanwang/teacher_basic_qwen3-0.6b" # 指向包含模型文件的文件夹
 
+# --- 加载模型和分词器 ---
+# 确保您的模型和分词器与您的任务兼容 (例如，文本生成/对话)
+# 如果您的Space有GPU，模型会自动尝试使用GPU (如果transformers和torch配置正确)
+try:
+    print(f"Loading tokenizer from: {MODEL_PATH}")
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
+    print(f"Loading model from: {MODEL_PATH}")
+    # 如果内存有限或希望在CPU上运行，可以添加 device_map="auto" 或 device_map="cpu"
+    # 对于较大的模型，device_map="auto" 可能需要 accelerate库
+    model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, device_map="cpu") # "auto" 会尝试使用GPU
+    print("Model and tokenizer loaded successfully.")
+except Exception as e:
+    print(f"Error loading model or tokenizer: {e}")
+    # 可以在这里引发异常或设置一个标志，以便在Gradio界面中显示错误
+    tokenizer = None
+    model = None
 
 def respond(
-    message,
+    message: str,
     history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
+    system_message: str,
+    max_tokens: int,
+    temperature: float,
+    top_p: float,
 ):
-    messages = [{"role": "system", "content": system_message}]
+    if model is None or tokenizer is None:
+        yield "Error: Model or tokenizer not loaded. Please check the logs."
+        return
+
+    # 确保tokenizer有pad_token_id，如果和eos_token_id一样，这是常见的设置
+    if tokenizer.pad_token_id is None:
+        print("tokenizer.pad_token_id is None. Setting it to tokenizer.eos_token_id.")
+        tokenizer.pad_token_id = tokenizer.eos_token_id
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
 
+    messages = [{"role": "system", "content": system_message}]
+    for user_msg, assistant_msg in history:
+        if user_msg:
+            messages.append({"role": "user", "content": user_msg})
+        if assistant_msg:
+            messages.append({"role": "assistant", "content": assistant_msg})
     messages.append({"role": "user", "content": message})
 
-    response = ""
+    try:
+        # 步骤 1: 使用聊天模板获取完整的提示字符串
+        # tokenize=False 表示我们先获取格式化后的字符串
+        prompt_string = tokenizer.apply_chat_template(
+            messages,
+            tokenize=False,
+            add_generation_prompt=True
+        )
+
+        # 步骤 2: 对格式化后的字符串进行分词，以获取 input_ids 和 attention_mask
+        # padding=True 和 truncation=True 是好的实践，虽然对于单个序列生成可能不是严格必须，
+        # 但如果模型期望固定长度或进行批处理，则非常重要。
+        # 对于单个序列，attention_mask 通常是全1（直到截断）。
+        inputs_dict = tokenizer(
+            prompt_string,
+            return_tensors="pt",
+            padding=False,  # 对于单个序列生成，通常不需要填充，除非模型特别要求
+            truncation=True,
+            max_length=tokenizer.model_max_length if hasattr(tokenizer, 'model_max_length') else 2048 # 使用一个合理的最大长度
+        )
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
+        input_ids = inputs_dict.input_ids.to(model.device)
+        attention_mask = inputs_dict.attention_mask.to(model.device) # <--- 获取 attention_mask
+
+    except Exception as e:
+        print(f"Error tokenizing chat or applying template: {e}")
+        yield f"Error during input processing: {str(e)}"
+        return
+
+    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+
+    generation_kwargs = dict(
+        input_ids=input_ids,           # <--- 使用 input_ids
+        attention_mask=attention_mask, # <--- 传递 attention_mask
+        streamer=streamer,
+        max_new_tokens=max_tokens,
+        do_sample=True,
         temperature=temperature,
         top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+        pad_token_id=tokenizer.pad_token_id # 明确传递 pad_token_id
+    )
+    # 确保 temperature > 0 for sampling
+    if temperature <= 0:
+        generation_kwargs["do_sample"] = False # 如果温度为0或负数，则不进行采样
 
-        response += token
-        yield response
+    thread = Thread(target=model.generate, kwargs=generation_kwargs)
+    thread.start()
+
+    response = ""
+    try:
+        for new_text in streamer:
+            if new_text is not None: # 确保 new_text 不是 None
+                response += new_text
+                yield response
+    except Exception as e:
+        print(f"Error during streaming response: {e}")
+        yield "Error: Could not generate response stream."
 
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", 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.Textbox(value="You are a helpful AI assistant.", label="System message"), # 默认系统消息
+        gr.Slider(minimum=1, maximum=4096, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"), # 调整了范围和默认值
         gr.Slider(
-            minimum=0.1,
+            minimum=0.01, # top_p 通常不会是0
             maximum=1.0,
             value=0.95,
-            step=0.05,
+            step=0.01, # 更细的调整粒度
             label="Top-p (nucleus sampling)",
         ),
     ],
+    title="My Local Chatbot",
+    description="Chat with a model running locally in this Hugging Face Space.",
 )
 
 
 if __name__ == "__main__":
-    demo.launch()
+    if model is None or tokenizer is None:
+        print("Cannot launch Gradio app because model or tokenizer failed to load.")
+    else:
+        demo.launch() # 在Hugging Face Spaces中，不需要 share=True

requirements.txt

@@ -1 +1,5 @@
-huggingface_hub==0.25.2
+huggingface_hub==0.25.2
+torch
+transformers
+gradio
+accelerate