init

app.py

@@ -1,12 +1,32 @@
+import spaces
 import gradio as gr
-from huggingface_hub import InferenceClient
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
+import os
 
-"""
-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")
+os.system('huggingface-cli download matteogeniaccio/phi-4 --local-dir ./phi-4 --include "phi-4/*"')
 
+# 加载 phi-4 模型和 tokenizer
+torch.random.manual_seed(0)
 
+model = AutoModelForCausalLM.from_pretrained(
+    "./phi-4",  # 模型路径
+    device_map="cuda",  # 使用 GPU
+    torch_dtype="auto",  # 自动选择数据类型
+    trust_remote_code=True,  # 允许远程代码加载
+)
+tokenizer = AutoTokenizer.from_pretrained("./phi-4")
+
+# 设置 pipeline
+
+pipe = pipeline(
+    "text-generation",
+    model=model,
+    tokenizer=tokenizer,
+)
+
+# 响应函数
+@spaces.GPU
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -15,34 +35,36 @@ def respond(
     temperature,
     top_p,
 ):
+    # 构造消息内容
     messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
+    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 = ""
+    # 将消息转换为字符串格式（适用于 text-generation）
+    input_text = "\n".join(
+        f"{msg['role']}: {msg['content']}" for msg in messages
+    )
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+    # 生成响应
+    generation_args = {
+        "max_new_tokens": max_tokens,
+        "temperature": temperature,
+        "top_p": top_p,
+        "do_sample": temperature > 0,
+        "return_full_text": False,
+    }
+    output = pipe(input_text, **generation_args)
+    response = output[0]["generated_text"]
 
-        response += token
-        yield response
+    # 返回流式响应
+    for token in response:
+        yield token
 
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
+# Gradio 界面
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
@@ -51,14 +73,10 @@ demo = gr.ChatInterface(
         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)",
+            maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"
         ),
     ],
 )
 
-
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()

gitignore.txt

@@ -0,0 +1 @@
+phi-4