diff --git a/Season2.step_into_llm/09.PEFT/PEFT_exampleWith_mrpcDataset.ipynb b/Season2.step_into_llm/09.PEFT/PEFT_exampleWith_mrpcDataset.ipynb
index 7bf5450..25b0366 100644
--- a/Season2.step_into_llm/09.PEFT/PEFT_exampleWith_mrpcDataset.ipynb
+++ b/Season2.step_into_llm/09.PEFT/PEFT_exampleWith_mrpcDataset.ipynb
@@ -4,328 +4,16 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# 环境配置\n",
-    "第一步：设置python版本为3.9.0"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 1,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "%%capture captured_output\n",
-    "!/home/ma-user/anaconda3/bin/conda create -n python-3.9.0 python=3.9.0 -y --override-channels --channel https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main\n",
-    "!/home/ma-user/anaconda3/envs/python-3.9.0/bin/pip install ipykernel"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 2,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "import json\n",
-    "import os\n",
-    "\n",
-    "data = {\n",
-    "   \"display_name\": \"python-3.9.0\",\n",
-    "   \"env\": {\n",
-    "      \"PATH\": \"/home/ma-user/anaconda3/envs/python-3.9.0/bin:/home/ma-user/anaconda3/envs/python-3.7.10/bin:/modelarts/authoring/notebook-conda/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/home/ma-user/modelarts/ma-cli/bin:/home/ma-user/modelarts/ma-cli/bin\"\n",
-    "   },\n",
-    "   \"language\": \"python\",\n",
-    "   \"argv\": [\n",
-    "      \"/home/ma-user/anaconda3/envs/python-3.9.0/bin/python\",\n",
-    "      \"-m\",\n",
-    "      \"ipykernel\",\n",
-    "      \"-f\",\n",
-    "      \"{connection_file}\"\n",
-    "   ]\n",
-    "}\n",
-    "\n",
-    "if not os.path.exists(\"/home/ma-user/anaconda3/share/jupyter/kernels/python-3.9.0/\"):\n",
-    "    os.mkdir(\"/home/ma-user/anaconda3/share/jupyter/kernels/python-3.9.0/\")\n",
-    "\n",
-    "with open('/home/ma-user/anaconda3/share/jupyter/kernels/python-3.9.0/kernel.json', 'w') as f:\n",
-    "    json.dump(data, f, indent=4)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "#### 注：以上代码运行完成后，需要重新设置kernel为python-3.9.0"
-   ]
-  },
-  {
-   "attachments": {
-    "f521c05d-5271-4aec-9ce4-bac212313cf2.png": {
-     "image/png": "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"
-    }
-   },
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "![image.png](attachment:f521c05d-5271-4aec-9ce4-bac212313cf2.png)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "\n",
+    "# MRPC数据集使用PEFT训练\n",
     "\n",
-    "第二步：安装最新版本的MindSpore框架和MindNLP套件"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 1,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple\n",
-      "Collecting mindspore==2.3.1\n",
-      "  Downloading https://ms-release.obs.cn-north-4.myhuaweicloud.com/2.3.1/MindSpore/unified/x86_64/mindspore-2.3.1-cp39-cp39-linux_x86_64.whl (946.9 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m946.9/946.9 MB\u001b[0m \u001b[31m61.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n",
-      "\u001b[?25hCollecting numpy<2.0.0,>=1.20.0 (from mindspore==2.3.1)\n",
-      "  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/54/30/c2a907b9443cf42b90c17ad10c1e8fa801975f01cb9764f3f8eb8aea638b/numpy-1.26.4-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (18.2 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m18.2/18.2 MB\u001b[0m \u001b[31m17.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n",
-      "\u001b[?25hCollecting protobuf>=3.13.0 (from mindspore==2.3.1)\n",
-      "  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/23/08/a1ce0415a115c2b703bfa798f06f0e43ca91dbe29d6180bf86a9287b15e2/protobuf-5.28.2-cp38-abi3-manylinux2014_x86_64.whl (316 kB)\n",
-      "Requirement already satisfied: asttokens>=2.0.4 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore==2.3.1) (2.4.1)\n",
-      "Collecting pillow>=6.2.0 (from mindspore==2.3.1)\n",
-      "  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/32/3f/c02268d0c6fb6b3958bdda673c17b315c821d97df29ae6969f20fb49388a/pillow-10.4.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (4.4 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m4.4/4.4 MB\u001b[0m \u001b[31m8.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n",
-      "\u001b[?25hCollecting scipy>=1.5.4 (from mindspore==2.3.1)\n",
-      "  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/35/f5/d0ad1a96f80962ba65e2ce1de6a1e59edecd1f0a7b55990ed208848012e0/scipy-1.13.1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (38.6 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m38.6/38.6 MB\u001b[0m \u001b[31m6.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m00:01\u001b[0m\n",
-      "\u001b[?25hRequirement already satisfied: packaging>=20.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore==2.3.1) (24.1)\n",
-      "Requirement already satisfied: psutil>=5.6.1 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore==2.3.1) (6.0.0)\n",
-      "Collecting astunparse>=1.6.3 (from mindspore==2.3.1)\n",
-      "  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/2b/03/13dde6512ad7b4557eb792fbcf0c653af6076b81e5941d36ec61f7ce6028/astunparse-1.6.3-py2.py3-none-any.whl (12 kB)\n",
-      "Requirement already satisfied: six>=1.12.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from asttokens>=2.0.4->mindspore==2.3.1) (1.16.0)\n",
-      "Requirement already satisfied: wheel<1.0,>=0.23.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from astunparse>=1.6.3->mindspore==2.3.1) (0.44.0)\n",
-      "Installing collected packages: protobuf, pillow, numpy, astunparse, scipy, mindspore\n",
-      "Successfully installed astunparse-1.6.3 mindspore-2.3.1 numpy-1.26.4 pillow-10.4.0 protobuf-5.28.2 scipy-1.13.1\n"
-     ]
-    }
-   ],
-   "source": [
-    "!pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/2.3.1/MindSpore/unified/x86_64/mindspore-2.3.1-cp39-cp39-linux_x86_64.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 2,
-   "metadata": {
-    "scrolled": true,
-    "tags": []
-   },
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Looking in indexes: http://repo.myhuaweicloud.com/repository/pypi/simple\n",
-      "Collecting mindnlp==0.4.0\n",
-      "  Downloading https://repo.mindspore.cn/mindspore-lab/mindnlp/newest/any/mindnlp-0.4.0-py3-none-any.whl (8.2 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m8.2/8.2 MB\u001b[0m \u001b[31m30.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n",
-      "\u001b[?25hRequirement already satisfied: mindspore>=2.2.14 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindnlp==0.4.0) (2.3.1)\n",
-      "Collecting tqdm (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/48/5d/acf5905c36149bbaec41ccf7f2b68814647347b72075ac0b1fe3022fdc73/tqdm-4.66.5-py3-none-any.whl (78 kB)\n",
-      "Collecting requests (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/f9/9b/335f9764261e915ed497fcdeb11df5dfd6f7bf257d4a6a2a686d80da4d54/requests-2.32.3-py3-none-any.whl (64 kB)\n",
-      "Collecting datasets (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/a5/52/45dab187f03d48c765b94db0464f5c10431756e47ae4cc6a8029a7d57a36/datasets-3.0.0-py3-none-any.whl (474 kB)\n",
-      "Collecting evaluate (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/a2/e7/cbca9e2d2590eb9b5aa8f7ebabe1beb1498f9462d2ecede5c9fd9735faaf/evaluate-0.4.3-py3-none-any.whl (84 kB)\n",
-      "Collecting tokenizers==0.19.1 (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/0f/cb/8fc733c8f251bac1e5c4ae52458c353b3faa98f41d734c226cad3783da03/tokenizers-0.19.1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (3.6 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m3.6/3.6 MB\u001b[0m \u001b[31m34.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
-      "\u001b[?25hCollecting safetensors (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/5d/80/81ba44fc82afbf5ca553913ac49460e325dc5cf00c317b34c14d43ebd76b/safetensors-0.4.5-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (436 kB)\n",
-      "Collecting sentencepiece (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/5f/01/c95e42eb86282b2c79305d3e0b0ca5a743f85a61262bb7130999c70b9374/sentencepiece-0.2.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.3/1.3 MB\u001b[0m \u001b[31m52.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
-      "\u001b[?25hCollecting regex (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/01/e6/a7256c99c312b68f01cfd4f8eae6e770906fffb3832ecb66f35ca5b86b96/regex-2024.9.11-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (781 kB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m782.0/782.0 kB\u001b[0m \u001b[31m37.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
-      "\u001b[?25hCollecting addict (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/6a/00/b08f23b7d7e1e14ce01419a467b583edbb93c6cdb8654e54a9cc579cd61f/addict-2.4.0-py3-none-any.whl (3.8 kB)\n",
-      "Collecting ml-dtypes (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/a8/6f/49effaafbc24c7665bcea42cacb22e7198bbab5b473d908c5900c6bb6a59/ml_dtypes-0.5.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (4.5 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m4.5/4.5 MB\u001b[0m \u001b[31m44.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
-      "\u001b[?25hCollecting pyctcdecode (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/a5/8a/93e2118411ae5e861d4f4ce65578c62e85d0f1d9cb389bd63bd57130604e/pyctcdecode-0.5.0-py2.py3-none-any.whl (39 kB)\n",
-      "Collecting jieba (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/c6/cb/18eeb235f833b726522d7ebed54f2278ce28ba9438e3135ab0278d9792a2/jieba-0.42.1.tar.gz (19.2 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m19.2/19.2 MB\u001b[0m \u001b[31m87.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m\n",
-      "\u001b[?25h  Preparing metadata (setup.py) ... \u001b[?25ldone\n",
-      "\u001b[?25hCollecting pytest==7.2.0 (from mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/67/68/a5eb36c3a8540594b6035e6cdae40c1ef1b6a2bfacbecc3d1a544583c078/pytest-7.2.0-py3-none-any.whl (316 kB)\n",
-      "Collecting attrs>=19.2.0 (from pytest==7.2.0->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/6a/21/5b6702a7f963e95456c0de2d495f67bf5fd62840ac655dc451586d23d39a/attrs-24.2.0-py3-none-any.whl (63 kB)\n",
-      "Collecting iniconfig (from pytest==7.2.0->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/ef/a6/62565a6e1cf69e10f5727360368e451d4b7f58beeac6173dc9db836a5b46/iniconfig-2.0.0-py3-none-any.whl (5.9 kB)\n",
-      "Requirement already satisfied: packaging in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from pytest==7.2.0->mindnlp==0.4.0) (24.1)\n",
-      "Collecting pluggy<2.0,>=0.12 (from pytest==7.2.0->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/88/5f/e351af9a41f866ac3f1fac4ca0613908d9a41741cfcf2228f4ad853b697d/pluggy-1.5.0-py3-none-any.whl (20 kB)\n",
-      "Requirement already satisfied: exceptiongroup>=1.0.0rc8 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from pytest==7.2.0->mindnlp==0.4.0) (1.2.2)\n",
-      "Collecting tomli>=1.0.0 (from pytest==7.2.0->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/97/75/10a9ebee3fd790d20926a90a2547f0bf78f371b2f13aa822c759680ca7b9/tomli-2.0.1-py3-none-any.whl (12 kB)\n",
-      "Collecting huggingface-hub<1.0,>=0.16.4 (from tokenizers==0.19.1->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/d5/ce/1f8e61cd63175cc2e79233b954b1c4e85363c788fb3a1fa23c87a25c9b81/huggingface_hub-0.25.0-py3-none-any.whl (436 kB)\n",
-      "Requirement already satisfied: numpy<2.0.0,>=1.20.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (1.26.4)\n",
-      "Requirement already satisfied: protobuf>=3.13.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (5.28.2)\n",
-      "Requirement already satisfied: asttokens>=2.0.4 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (2.4.1)\n",
-      "Requirement already satisfied: pillow>=6.2.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (10.4.0)\n",
-      "Requirement already satisfied: scipy>=1.5.4 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (1.13.1)\n",
-      "Requirement already satisfied: psutil>=5.6.1 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (6.0.0)\n",
-      "Requirement already satisfied: astunparse>=1.6.3 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from mindspore>=2.2.14->mindnlp==0.4.0) (1.6.3)\n",
-      "Collecting filelock (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/b9/f8/feced7779d755758a52d1f6635d990b8d98dc0a29fa568bbe0625f18fdf3/filelock-3.16.1-py3-none-any.whl (16 kB)\n",
-      "Collecting pyarrow>=15.0.0 (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/af/61/bcd9b58e38ead6ad42b9ed00da33a3f862bc1d445e3d3164799c25550ac2/pyarrow-17.0.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (39.9 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m39.9/39.9 MB\u001b[0m \u001b[31m101.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0ma \u001b[36m0:00:01\u001b[0m\n",
-      "\u001b[?25hCollecting dill<0.3.9,>=0.3.0 (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/c9/7a/cef76fd8438a42f96db64ddaa85280485a9c395e7df3db8158cfec1eee34/dill-0.3.8-py3-none-any.whl (116 kB)\n",
-      "Collecting pandas (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/bb/30/f6f1f1ac36250f50c421b1b6af08c35e5a8b5a84385ef928625336b93e6f/pandas-2.2.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (13.1 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m13.1/13.1 MB\u001b[0m \u001b[31m58.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m\n",
-      "\u001b[?25hCollecting xxhash (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/b4/b4/332647451ed7d2c021294b7c1e9c144dbb5586b1fb214ad4f5a404642835/xxhash-3.5.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (193 kB)\n",
-      "Collecting multiprocess (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/da/d9/f7f9379981e39b8c2511c9e0326d212accacb82f12fbfdc1aa2ce2a7b2b6/multiprocess-0.70.16-py39-none-any.whl (133 kB)\n",
-      "Collecting fsspec<=2024.6.1,>=2023.1.0 (from fsspec[http]<=2024.6.1,>=2023.1.0->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/5e/44/73bea497ac69bafde2ee4269292fa3b41f1198f4bb7bbaaabde30ad29d4a/fsspec-2024.6.1-py3-none-any.whl (177 kB)\n",
-      "Collecting aiohttp (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/23/69/200bf165b56c17854d54975f894de10dababc4d0226c07600c9abc679e7e/aiohttp-3.10.5-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.2/1.2 MB\u001b[0m \u001b[31m51.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
-      "\u001b[?25hCollecting pyyaml>=5.1 (from datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/3d/32/e7bd8535d22ea2874cef6a81021ba019474ace0d13a4819c2a4bce79bd6a/PyYAML-6.0.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (737 kB)\n",
-      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m737.4/737.4 kB\u001b[0m \u001b[31m33.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
-      "\u001b[?25hCollecting charset-normalizer<4,>=2 (from requests->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/98/69/5d8751b4b670d623aa7a47bef061d69c279e9f922f6705147983aa76c3ce/charset_normalizer-3.3.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (142 kB)\n",
-      "Collecting idna<4,>=2.5 (from requests->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/76/c6/c88e154df9c4e1a2a66ccf0005a88dfb2650c1dffb6f5ce603dfbd452ce3/idna-3.10-py3-none-any.whl (70 kB)\n",
-      "Collecting urllib3<3,>=1.21.1 (from requests->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/ce/d9/5f4c13cecde62396b0d3fe530a50ccea91e7dfc1ccf0e09c228841bb5ba8/urllib3-2.2.3-py3-none-any.whl (126 kB)\n",
-      "Collecting certifi>=2017.4.17 (from requests->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/12/90/3c9ff0512038035f59d279fddeb79f5f1eccd8859f06d6163c58798b9487/certifi-2024.8.30-py3-none-any.whl (167 kB)\n",
-      "Collecting pygtrie<3.0,>=2.1 (from pyctcdecode->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/ec/cd/bd196b2cf014afb1009de8b0f05ecd54011d881944e62763f3c1b1e8ef37/pygtrie-2.5.0-py3-none-any.whl (25 kB)\n",
-      "Collecting hypothesis<7,>=6.14 (from pyctcdecode->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/5f/25/f1fb5b3ec58ed3c6014385672d4298e2f0c7291bfcd9ffd06627a641470d/hypothesis-6.112.1-py3-none-any.whl (467 kB)\n",
-      "Requirement already satisfied: six>=1.12.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from asttokens>=2.0.4->mindspore>=2.2.14->mindnlp==0.4.0) (1.16.0)\n",
-      "Requirement already satisfied: wheel<1.0,>=0.23.0 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from astunparse>=1.6.3->mindspore>=2.2.14->mindnlp==0.4.0) (0.44.0)\n",
-      "Collecting aiohappyeyeballs>=2.3.0 (from aiohttp->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/18/b6/58ea188899950d759a837f9a58b2aee1d1a380ea4d6211ce9b1823748851/aiohappyeyeballs-2.4.0-py3-none-any.whl (12 kB)\n",
-      "Collecting aiosignal>=1.1.2 (from aiohttp->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/76/ac/a7305707cb852b7e16ff80eaf5692309bde30e2b1100a1fcacdc8f731d97/aiosignal-1.3.1-py3-none-any.whl (7.6 kB)\n",
-      "Collecting frozenlist>=1.1.1 (from aiohttp->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/70/b0/6f1ebdabfb604e39a0f84428986b89ab55f246b64cddaa495f2c953e1f6b/frozenlist-1.4.1-cp39-cp39-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl (240 kB)\n",
-      "Collecting multidict<7.0,>=4.5 (from aiohttp->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/5e/41/0d0fb18c1ad574f807196f5f3d99164edf9de3e169a58c6dc2d6ed5742b9/multidict-6.1.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (124 kB)\n",
-      "Collecting yarl<2.0,>=1.0 (from aiohttp->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/ff/be/78953a3d5154b974af49ce367f1a8d4751ababdf26a66ae607b4ae625d99/yarl-1.11.1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (453 kB)\n",
-      "Collecting async-timeout<5.0,>=4.0 (from aiohttp->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/a7/fa/e01228c2938de91d47b307831c62ab9e4001e747789d0b05baf779a6488c/async_timeout-4.0.3-py3-none-any.whl (5.7 kB)\n",
-      "Requirement already satisfied: typing-extensions>=3.7.4.3 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from huggingface-hub<1.0,>=0.16.4->tokenizers==0.19.1->mindnlp==0.4.0) (4.12.2)\n",
-      "Collecting sortedcontainers<3.0.0,>=2.1.0 (from hypothesis<7,>=6.14->pyctcdecode->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/32/46/9cb0e58b2deb7f82b84065f37f3bffeb12413f947f9388e4cac22c4621ce/sortedcontainers-2.4.0-py2.py3-none-any.whl (29 kB)\n",
-      "Requirement already satisfied: python-dateutil>=2.8.2 in /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages (from pandas->datasets->mindnlp==0.4.0) (2.9.0.post0)\n",
-      "Collecting pytz>=2020.1 (from pandas->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/11/c3/005fcca25ce078d2cc29fd559379817424e94885510568bc1bc53d7d5846/pytz-2024.2-py2.py3-none-any.whl (508 kB)\n",
-      "Collecting tzdata>=2022.7 (from pandas->datasets->mindnlp==0.4.0)\n",
-      "  Downloading http://repo.myhuaweicloud.com/repository/pypi/packages/65/58/f9c9e6be752e9fcb8b6a0ee9fb87e6e7a1f6bcab2cdc73f02bb7ba91ada0/tzdata-2024.1-py2.py3-none-any.whl (345 kB)\n",
-      "Building wheels for collected packages: jieba\n",
-      "  Building wheel for jieba (setup.py) ... \u001b[?25ldone\n",
-      "\u001b[?25h  Created wheel for jieba: filename=jieba-0.42.1-py3-none-any.whl size=19314458 sha256=de190811901ea689a37a0ecc8e410ef914b6b76740894df9b47c2bdbfd51decc\n",
-      "  Stored in directory: /home/ma-user/.cache/pip/wheels/2d/22/9e/9af7e8c2773513ac75905acfb75073922bcc1aa176f730a0c9\n",
-      "Successfully built jieba\n",
-      "Installing collected packages: sortedcontainers, sentencepiece, pytz, pygtrie, jieba, addict, xxhash, urllib3, tzdata, tqdm, tomli, safetensors, regex, pyyaml, pyarrow, pluggy, multidict, ml-dtypes, iniconfig, idna, fsspec, frozenlist, filelock, dill, charset-normalizer, certifi, attrs, async-timeout, aiohappyeyeballs, yarl, requests, pytest, pandas, multiprocess, hypothesis, aiosignal, pyctcdecode, huggingface-hub, aiohttp, tokenizers, datasets, evaluate, mindnlp\n",
-      "Successfully installed addict-2.4.0 aiohappyeyeballs-2.4.0 aiohttp-3.10.5 aiosignal-1.3.1 async-timeout-4.0.3 attrs-24.2.0 certifi-2024.8.30 charset-normalizer-3.3.2 datasets-3.0.0 dill-0.3.8 evaluate-0.4.3 filelock-3.16.1 frozenlist-1.4.1 fsspec-2024.6.1 huggingface-hub-0.25.0 hypothesis-6.112.1 idna-3.10 iniconfig-2.0.0 jieba-0.42.1 mindnlp-0.4.0 ml-dtypes-0.5.0 multidict-6.1.0 multiprocess-0.70.16 pandas-2.2.2 pluggy-1.5.0 pyarrow-17.0.0 pyctcdecode-0.5.0 pygtrie-2.5.0 pytest-7.2.0 pytz-2024.2 pyyaml-6.0.2 regex-2024.9.11 requests-2.32.3 safetensors-0.4.5 sentencepiece-0.2.0 sortedcontainers-2.4.0 tokenizers-0.19.1 tomli-2.0.1 tqdm-4.66.5 tzdata-2024.1 urllib3-2.2.3 xxhash-3.5.0 yarl-1.11.1\n"
-     ]
-    }
-   ],
-   "source": [
-    "#安装mindnlp的daily包，待正式发布后可改为直接安装mindnlp包\n",
-    "!pip install https://repo.mindspore.cn/mindspore-lab/mindnlp/newest/any/mindnlp-0.4.0-py3-none-any.whl\n",
-    "# !pip install mindnlp==0.4.0"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 1,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Name: mindspore\n",
-      "Version: 2.3.1\n",
-      "Summary: MindSpore is a new open source deep learning training/inference framework that could be used for mobile, edge and cloud scenarios.\n",
-      "Home-page: https://www.mindspore.cn\n",
-      "Author: The MindSpore Authors\n",
-      "Author-email: contact@mindspore.cn\n",
-      "License: Apache 2.0\n",
-      "Location: /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages\n",
-      "Requires: asttokens, astunparse, numpy, packaging, pillow, protobuf, psutil, scipy\n",
-      "Required-by: mindnlp\n"
-     ]
-    }
-   ],
-   "source": [
-    "!pip show mindspore"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 2,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Name: mindnlp\n",
-      "Version: 0.4.0\n",
-      "Summary: An open source natural language processing research tool box. Git version: [sha1]:2fb76bf, [branch]: (HEAD, origin/master, origin/HEAD, master)\n",
-      "Home-page: https://github.com/mindlab-ai/mindnlp/tree/master/\n",
-      "Author: MindSpore Team\n",
-      "Author-email: \n",
-      "License: Apache 2.0\n",
-      "Location: /home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages\n",
-      "Requires: addict, datasets, evaluate, jieba, mindspore, ml-dtypes, pyctcdecode, pytest, regex, requests, safetensors, sentencepiece, tokenizers, tqdm\n",
-      "Required-by: \n"
-     ]
-    }
-   ],
-   "source": [
-    "!pip show mindnlp"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# mrpc_dataset\n",
-    "<pre>\n",
-    "    MRPC数据集，全称为Microsoft Research Paraphrase Corpus（微软研究院释义语料库），是一个用于NLP的句对相似性判断任务中性能评估的数据集。\n",
-    "    MRPC数据集包含了大量从新闻、网页和论坛中收集的英文句子对。每个句子对都有一个人工标注的二元标签：0表示两句话不相似，1表示它们相似。\n",
-    "</pre>"
+    "环境 python==3.9 mindnlp==0.4.1 mindspore==2.6.0"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "##### 加载mrpc数据集并拆分成训练集、验证集、测试集"
+    "## 1. 导入依赖库"
    ]
   },
   {
@@ -337,108 +25,56 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "/home/ma-user/anaconda3/envs/python-3.9.0/lib/python3.9/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
-      "  from .autonotebook import tqdm as notebook_tqdm\n",
       "Building prefix dict from the default dictionary ...\n",
       "Loading model from cache /tmp/jieba.cache\n",
-      "Loading model cost 0.753 seconds.\n",
+      "Loading model cost 0.283 seconds.\n",
       "Prefix dict has been built successfully.\n"
      ]
     }
    ],
    "source": [
-    "from mindnlp.dataset import load_dataset"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 29,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Repo card metadata block was not found. Setting CardData to empty.\n",
-      "Downloading data: 100%|██████████| 1.14M/1.14M [00:00<00:00, 1.45MB/s]\n",
-      "Downloading data: 100%|██████████| 127k/127k [00:00<00:00, 131kB/s]  \n",
-      "Downloading data: 100%|██████████| 533k/533k [00:00<00:00, 666kB/s]  \n",
-      "Generating train split: 3668 examples [00:00, 176571.87 examples/s]\n",
-      "Generating validation split: 408 examples [00:00, 48980.37 examples/s]\n",
-      "Generating test split: 1725 examples [00:00, 153982.47 examples/s]\n"
-     ]
-    }
-   ],
-   "source": [
-    "mrpc_dict = load_dataset(\"SetFit/mrpc\")   # 如果本地未下载会先下载，若已下载则会直接加载\n",
-    "mrpc_train = mrpc_dict['train']\n",
-    "mrpc_valid = mrpc_dict['validation']\n",
-    "mrpc_test = mrpc_dict['test']"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 3,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 3668 samples\n",
-      "validation: 408 samples\n",
-      "test: 1725 samples\n"
-     ]
-    }
-   ],
-   "source": [
-    "# 打印每个数据集的样本数量\n",
-    "for k,v in mrpc_dict.items():\n",
-    "    print(f\"{k}: {len(v)} samples\")"
+    "import os\n",
+    "import json\n",
+    "import copy\n",
+    "import mindspore\n",
+    "from mindspore import context, Tensor, ops\n",
+    "from mindspore.dataset import NumpySlicesDataset, SequentialSampler\n",
+    "from mindspore.common.parameter import Parameter\n",
+    "from mindspore.nn import AdamWeightDecay\n",
+    "from mindnlp.engine import Evaluator\n",
+    "from mindnlp.metrics import Accuracy\n",
+    "from mindnlp.common.grad import value_and_grad\n",
+    "from mindnlp.dataset import load_dataset\n",
+    "from mindnlp.transformers import GPT2Tokenizer, GPT2ForSequenceClassification\n",
+    "from mindnlp.peft import LoraConfig, get_peft_model, TaskType\n",
+    "from tqdm.auto import tqdm\n",
+    "\n",
+    "# 导入辅助函数\n",
+    "from train_llama_lora.fix_mrpc_training import (\n",
+    "    print_dataset_keys,\n",
+    "    improved_forward_fn,\n",
+    "    improved_train_step,\n",
+    "    examine_batch,\n",
+    "    prepare_mrpc_batch\n",
+    ")\n"
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": 6,
+   "cell_type": "markdown",
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "text1: Amrozi accused his brother , whom he called \" the witness \" , of deliberately distorting his evidence .\n",
-      "text2: Referring to him as only \" the witness \" , Amrozi accused his brother of deliberately distorting his evidence .\n",
-      "label: 1\n",
-      "idx: 0\n",
-      "label_text: equivalent\n"
-     ]
-    }
-   ],
    "source": [
-    "# 打印原数据集的样本格式及其内容\n",
-    "for dataDict in mrpc_train.create_dict_iterator():\n",
-    "    for col_name, data in dataDict.items():\n",
-    "        print(f\"{col_name}: {data}\")\n",
-    "    break"
+    "## 2. 定义数据处理类和函数"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
-    "import json\n",
-    "import copy\n",
-    "\n",
-    "from mindspore.dataset import NumpySlicesDataset, SequentialSampler\n",
-    "\n",
     "class InputExample(object):\n",
+    "    \"\"\"单个输入示例，包含一个全局唯一标识符、文本A、可选的文本B和标签\"\"\"\n",
     "    def __init__(self, guid, text_a, text_b=None, label=None):\n",
-    "        \"\"\"\n",
-    "            InputExample表示单个输入示例\n",
-    "            包含一个全局唯一标识符（guid）、文本 A（text_a）、可选的文本 B（text_b）和标签（label）\n",
-    "        \"\"\"\n",
     "        self.guid = guid\n",
     "        self.text_a = text_a\n",
     "        self.text_b = text_b\n",
@@ -448,20 +84,24 @@
     "        return str(self.to_json_string())\n",
     "\n",
     "    def to_dict(self):\n",
-    "        \"\"\"Serializes this instance to a Python dictionary.\"\"\"\n",
+    "        \"\"\"将实例序列化为Python字典\"\"\"\n",
     "        output = copy.deepcopy(self.__dict__)\n",
     "        return output\n",
     "\n",
     "    def to_json_string(self):\n",
-    "        \"\"\"Serializes this instance to a JSON string.\"\"\"\n",
-    "        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + \"\\n\"    \n",
-    "\n",
+    "        \"\"\"将实例序列化为JSON字符串\"\"\"\n",
+    "        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + \"\\n\"    \n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
     "class InputFeatures(object):\n",
+    "    \"\"\"表示模型输入特征，包含输入ID、注意力掩码、标记类型ID、标签和输入长度\"\"\"\n",
     "    def __init__(self, input_ids, attention_mask, token_type_ids, label, input_len):\n",
-    "        \"\"\"\n",
-    "            InputFeatures 表示模型输入特征\n",
-    "            包含输入 ID（input_ids）、注意力掩码（attention_mask）、标记类型 ID（token_type_ids）、标签（label）和输入长度（input_len）\n",
-    "        \"\"\"\n",
     "        self.input_ids = input_ids\n",
     "        self.attention_mask = attention_mask\n",
     "        self.token_type_ids = token_type_ids\n",
@@ -472,43 +112,41 @@
     "        return str(self.to_json_string())\n",
     "\n",
     "    def to_dict(self):\n",
-    "        \"\"\"Serializes this instance to a Python dictionary.\"\"\"\n",
+    "        \"\"\"将实例序列化为Python字典\"\"\"\n",
     "        output = copy.deepcopy(self.__dict__)\n",
     "        return output\n",
     "\n",
     "    def to_json_string(self):\n",
-    "        \"\"\"Serializes this instance to a JSON string.\"\"\"\n",
-    "        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + \"\\n\""
+    "        \"\"\"将实例序列化为JSON字符串\"\"\"\n",
+    "        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + \"\\n\"\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 4,
    "metadata": {},
    "outputs": [],
    "source": [
     "def convert_dataset_to_examples(ds):\n",
-    "    \"\"\"\n",
-    "        Convert dataset to examples.\n",
-    "        将数据集ds转换为 InputExample 实例列表examples\n",
-    "        使用 mindspore.dataset 的迭代器遍历数据集，并将每个样本转换为 InputExample 对象。\n",
-    "    \"\"\"\n",
+    "    \"\"\"将数据集转换为示例列表\"\"\"\n",
     "    examples = []\n",
     "    iter0 = ds.create_tuple_iterator()\n",
-    "    #for i, (label, text_a, text_b) in enumerate(iter0):\n",
     "    for i, (text_a, text_b, label, idx, label_text) in enumerate(iter0):\n",
-    "        # print(str(text_a.asnumpy()), str(text_b.asnumpy()))\n",
     "        examples.append(\n",
     "            InputExample(guid=i, text_a=str(text_a.asnumpy()), text_b=str(text_b.asnumpy()), label=int(label))\n",
     "        )\n",
     "    \n",
-    "    return examples\n",
-    "\n",
+    "    return examples\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [],
+   "source": [
     "def _truncate_seq_pair(tokens_a, tokens_b, max_length):\n",
-    "    \"\"\"\n",
-    "        Truncates a sequence pair in place to the maximum length.\n",
-    "        即保持文本对的意义，同时截断文本对，使其总长度不超过指定的最大长度max_length\n",
-    "    \"\"\"\n",
+    "    \"\"\"截断文本对，使其总长度不超过指定的最大长度\"\"\"\n",
     "    while True:\n",
     "        total_length = len(tokens_a) + len(tokens_b)\n",
     "        if total_length <= max_length:\n",
@@ -517,7 +155,7 @@
     "        if len(tokens_a) > len(tokens_b):\n",
     "            tokens_a.pop()\n",
     "        else:\n",
-    "            tokens_b.pop()"
+    "            tokens_b.pop()\n"
    ]
   },
   {
@@ -527,10 +165,7 @@
    "outputs": [],
    "source": [
     "def convert_examples_to_features(examples, tokenizer, max_seq_length=512):\n",
-    "    \"\"\"\n",
-    "        将 InputExample 实例列表转换为 InputFeatures 实例列表。\n",
-    "        使用 tokenizer 对文本进行编码，生成模型输入所需的特征。\n",
-    "    \"\"\"\n",
+    "    \"\"\"将示例列表转换为特征列表\"\"\"\n",
     "    features = []\n",
     "\n",
     "    for ex_index, example in enumerate(examples):\n",
@@ -559,14 +194,10 @@
     "            for token in tokens_b[1:]:\n",
     "                tokens.append(token)\n",
     "                token_type_ids.append(1)\n",
-    "            # tokens.append(\"[SEP]\")\n",
-    "            # token_type_ids.append(1)\n",
     "\n",
     "        tokenizer.return_token=False\n",
-    "        # input_ids = tokenizer.execute_py(example.text_a).tolist() + tokenizer.execute_py(example.text_b).tolist()\n",
     "        input_ids = tokenizer.convert_tokens_to_ids(tokens)\n",
     "\n",
-    "        # print(tokenizer.execute_py(np.array(tokens)).tolist())\n",
     "        # The mask has 1 for real tokens and 0 for padding tokens. Only real\n",
     "        # tokens are attended to.\n",
     "        attention_mask = [1] * len(input_ids)\n",
@@ -584,16 +215,6 @@
     "        \n",
     "        label_id = example.label\n",
     "\n",
-    "        # if ex_index < 5:\n",
-    "        #     print(\"*** Example ***\")\n",
-    "        #     print(\"guid: %s\" % (example.guid))\n",
-    "        #     print(\"tokens: %s\"%\" \".join([str(x) for x in tokens]))\n",
-    "        #     print(\"input_ids: %s\" % \" \".join([str(x) for x in input_ids]))\n",
-    "        #     print(\"attention_mask: %s\" % \" \".join([str(x) for x in attention_mask]))\n",
-    "        #     print(\"token_type_ids: %s\" % \" \".join([str(x) for x in token_type_ids]))\n",
-    "        #     print(\"label: %s (id = %d)\" % (example.label, label_id))\n",
-    "        #     print(\"input length: %d\" % (input_len))\n",
-    "\n",
     "        features.append(\n",
     "            InputFeatures(input_ids=input_ids,\n",
     "                          attention_mask=attention_mask,\n",
@@ -611,11 +232,7 @@
    "outputs": [],
    "source": [
     "def load_examples(tokenizer, max_seq_length, mrpc_datas):\n",
-    "    \"\"\"load_examples using load_dataset\n",
-    "        加载数据集并转换为模型训练所需的特征：\n",
-    "            首先加载 MRPC 数据集的指定部分（训练或测试）\n",
-    "            然后调用 convert_examples_to_features 函数转换为模型输入所需的特征\n",
-    "    \"\"\"\n",
+    "    \"\"\"加载数据集并转换为模型训练所需的特征\"\"\"\n",
     "    \n",
     "    train_examples = convert_dataset_to_examples(mrpc_datas)\n",
     "\n",
@@ -629,22 +246,30 @@
     "    all_labels = [f.label for f in features]\n",
     "    dataset = ((all_input_ids, all_attention_mask, all_token_type_ids, all_lens, all_labels))\n",
     "\n",
-    "    return dataset\n",
-    "\n",
+    "    return dataset\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
     "def get_dataloader_from_ds(ds, batch_size):\n",
+    "    \"\"\"从数据集创建数据加载器\"\"\"\n",
     "    train_sampler = SequentialSampler()  # 应用 SequentialSampler 以顺序方式采样数据\n",
     "    col_names = ['input_ids', 'attention_mask', 'token_type_ids', 'lens', 'labels']\n",
     "    train_dataloader = NumpySlicesDataset(ds, sampler=train_sampler, column_names=col_names)  # 使用 NumpySlicesDataset 包装数据集\n",
     "    train_dataloader = train_dataloader.batch(batch_size)  # 根据指定批次大小 进行 批处理\n",
     "\n",
-    "    return train_dataloader"
+    "    return train_dataloader\n"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "<b>此次将使用GPT2的词汇表对数据集的样本特征进行token转换</b>"
+    "## 3. 设置训练参数"
    ]
   },
   {
@@ -656,76 +281,124 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 26.0/26.0 [00:00<00:00, 134kB/s]\n",
-      "0.99MB [00:00, 3.22MB/s]\n",
-      "446kB [00:00, 1.77MB/s]\n",
-      "1.29MB [00:00, 4.27MB/s]\n"
+      "[WARNING] ME(10862:123128869447488,MainProcess):2025-05-21-18:18:23.362.307 [mindspore/context.py:1401] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead.\n"
+     ]
+    }
+   ],
+   "source": [
+    "# 定义训练参数\n",
+    "class Args:\n",
+    "    def __init__(self):\n",
+    "        self.save_dir = \"./saved_models\"  # 模型保存目录\n",
+    "        self.lr = 1e-4                  # 学习率\n",
+    "        self.num_epochs = 1            # 训练轮数\n",
+    "        self.debug = False              # 是否启用调试模式\n",
+    "        self.batch_size = 16            # 批次大小\n",
+    "        self.max_seq_len = 256          # 最大序列长度\n",
+    "        self.model_name = \"gpt2\"        # 基础模型名称\n",
+    "        self.use_lora = True            # 是否使用LoRA进行微调\n",
+    "\n",
+    "args = Args()\n",
+    "\n",
+    "# 设置运行模式\n",
+    "context.set_context(mode=context.PYNATIVE_MODE, device_target=\"GPU\")\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 4. 加载数据集"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "加载MRPC数据集...\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Repo card metadata block was not found. Setting CardData to empty.\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "3\n"
+      "加载tokenizer...\n",
+      "添加了 3 个特殊token\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/home/yyy/桌面/mindnlp/mindnlp/transformers/tokenization_utils_base.py:1526: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted, and will be then set to `False` by default. \n",
+      "  warnings.warn(\n"
      ]
     }
    ],
    "source": [
-    "from mindnlp.transformers import GPT2Tokenizer\n",
+    "print(\"加载MRPC数据集...\")\n",
+    "# 加载MRPC数据集\n",
+    "mrpc_dict = load_dataset(\"SetFit/mrpc\")\n",
+    "mrpc_train = mrpc_dict['train']\n",
+    "mrpc_valid = mrpc_dict['validation']\n",
+    "mrpc_test = mrpc_dict['test']\n",
     "\n",
-    "tokenizer = GPT2Tokenizer.from_pretrained(\"gpt2\")\n",
-    "# add sepcial token: <PAD>\n",
+    "print(\"加载tokenizer...\")\n",
+    "# 加载tokenizer\n",
+    "tokenizer = GPT2Tokenizer.from_pretrained(args.model_name)\n",
+    "# 添加特殊token\n",
     "special_tokens_dict = {\n",
     "    \"bos_token\": \"<bos>\",\n",
     "    \"eos_token\": \"<eos>\",\n",
     "    \"pad_token\": \"<pad>\",\n",
     "}\n",
     "num_added_toks = tokenizer.add_special_tokens(special_tokens_dict)\n",
-    "print(num_added_toks)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 3,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# 此次实验的指定的一些超参数\n",
-    "class Args:\n",
-    "    save_dir=\".mindnlp/peft_model/mrpc_IA3\"   # The output directory where the model checkpoints will be written.\n",
-    "    batch_size = 8  # Batch size per GPU/CPU for training\n",
-    "    model_name_or_path = \"gpt2\"   \n",
-    "    num_epochs = 5\n",
-    "    lr = 2e-5    # learning rate\n",
-    "    max_seq_len = 256\n",
-    "    debug = True\n",
-    "    is_lora = True\n",
-    "\n",
-    "args = Args()"
+    "print(f\"添加了 {num_added_toks} 个特殊token\")\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 11,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "处理数据集...\n"
+     ]
+    }
+   ],
    "source": [
-    "# load dataset, 将每个样本转换为GPT2模型输入的正确格式\n",
+    "print(\"处理数据集...\")\n",
+    "# 处理数据集\n",
     "train_ds = load_examples(tokenizer, args.max_seq_len, mrpc_train)\n",
     "valid_ds = load_examples(tokenizer, args.max_seq_len, mrpc_valid)\n",
-    "test_ds = load_examples(tokenizer, args.max_seq_len, mrpc_test)"
+    "test_ds = load_examples(tokenizer, args.max_seq_len, mrpc_test)\n",
+    "\n",
+    "# 转换为dataloader\n",
+    "train_dataloader = get_dataloader_from_ds(train_ds, args.batch_size)\n",
+    "valid_dataloader = get_dataloader_from_ds(valid_ds, args.batch_size)\n",
+    "test_dataloader = get_dataloader_from_ds(test_ds, args.batch_size)\n"
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": 11,
+   "cell_type": "markdown",
    "metadata": {},
-   "outputs": [],
    "source": [
-    "# 转换为mindspore框架下数据加载器（包含批量处理）\n",
-    "train_dataloader = get_dataloader_from_ds(train_ds, args.batch_size)\n",
-    "valid_dataloader = get_dataloader_from_ds(valid_ds, args.batch_size)\n",
-    "test_dataloader = get_dataloader_from_ds(test_ds, args.batch_size)"
+    "## 5. 加载模型和配置PEFT"
    ]
   },
   {
@@ -734,296 +407,472 @@
    "metadata": {},
    "outputs": [
     {
-     "data": {
-      "text/plain": [
-       "['input_ids', 'attention_mask', 'token_type_ids', 'lens', 'labels']"
-      ]
-     },
-     "execution_count": 12,
-     "metadata": {},
-     "output_type": "execute_result"
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "加载模型...\n",
+      "[MS_ALLOC_CONF]Runtime config:  enable_vmm:True  vmm_align_size:2MB\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Some weights of GPT2ForSequenceClassification were not initialized from the model checkpoint at gpt2 and are newly initialized: ['score.weight']\n",
+      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "应用LoRA配置...\n",
+      "可训练参数信息:\n",
+      "trainable params: 296,448 || all params: 124,740,096 || trainable%: 0.237652534755144\n"
+     ]
     }
    ],
    "source": [
-    "train_dataloader.get_col_names()   # 数据集样本的列名"
+    "print(\"加载模型...\")\n",
+    "# 加载模型\n",
+    "model = GPT2ForSequenceClassification.from_pretrained(args.model_name, num_labels=2, force_download=False, use_safetensors=False)\n",
+    "model.config.pad_token_id = tokenizer.pad_token_id\n",
+    "model.resize_token_embeddings(model.config.vocab_size + num_added_toks)\n",
+    "\n",
+    "# 配置PEFT (LoRA)\n",
+    "if args.use_lora:\n",
+    "    print(\"应用LoRA配置...\")\n",
+    "    peft_config = LoraConfig(task_type=TaskType.SEQ_CLS, inference_mode=False, r=8, lora_alpha=32, lora_dropout=0.1, fan_in_fan_out=True)\n",
+    "    model = get_peft_model(model, peft_config)\n",
+    "    print(\"可训练参数信息:\")\n",
+    "    model.print_trainable_parameters()\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": []
-  },
-  {
-   "cell_type": "markdown",
+   "execution_count": 13,
    "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "找到 25 个可训练参数\n",
+      "转换后的参数数量: 25\n"
+     ]
+    }
+   ],
    "source": [
-    "# 加载预训练模型并进行微调"
+    "# 获取可训练参数\n",
+    "params = []\n",
+    "if hasattr(model, \"get_trainable_parameters\"):\n",
+    "    params = model.get_trainable_parameters()\n",
+    "elif hasattr(model, \"trainable_params\"):\n",
+    "    params = model.trainable_params()\n",
+    "\n",
+    "print(f\"找到 {len(params)} 个可训练参数\")\n",
+    "\n",
+    "# 转换参数\n",
+    "converted_params = []\n",
+    "for param in params:\n",
+    "    if hasattr(param, \"data\") and hasattr(param, \"name\"):\n",
+    "        converted_param = Parameter(param.data, name=param.name)\n",
+    "    else:\n",
+    "        converted_param = param\n",
+    "    converted_params.append(converted_param)\n",
+    "\n",
+    "print(f\"转换后的参数数量: {len(converted_params)}\")\n",
+    "\n",
+    "# 使用转换后的参数\n",
+    "optimizer = AdamWeightDecay(params=converted_params, learning_rate=args.lr)\n",
+    "metric = Accuracy()\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": 14,
    "metadata": {},
    "outputs": [],
    "source": [
-    "import os\n",
-    "import mindspore\n",
-    "import numpy as np\n",
-    "\n",
-    "from tqdm import tqdm \n",
-    "from mindspore.nn import AdamWeightDecay"
+    "# 创建保存目录\n",
+    "os.makedirs(args.save_dir, exist_ok=True)\n"
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": null,
+   "cell_type": "markdown",
    "metadata": {},
-   "outputs": [],
    "source": [
-    "from mindnlp.transformers import  GPT2ForSequenceClassification"
+    "## 6. 定义训练和评估函数"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [
     {
-     "name": "stderr",
+     "name": "stdout",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 523M/523M [00:32<00:00, 16.7MB/s] \n",
-      "100%|██████████| 124/124 [00:00<00:00, 365kB/s]\n",
-      "The following parameters in models are missing parameter:\n",
-      "['score.weight']\n"
+      "训练数据集列名: ['input_ids', 'attention_mask', 'token_type_ids', 'lens', 'labels']\n",
+      "处理后训练数据集列名: ['input_ids', 'attention_mask', 'token_type_ids', 'labels']\n"
      ]
     }
    ],
    "source": [
-    "model = GPT2ForSequenceClassification.from_pretrained(\"gpt2\", num_labels = 2)"
+    "# 预处理数据集，移除不需要的'lens'列\n",
+    "def remove_lens_column(dataset):\n",
+    "    \"\"\"移除数据集中的'lens'列\"\"\"\n",
+    "    columns_to_keep = [col for col in dataset.get_col_names() if col != 'lens']\n",
+    "    return dataset.project(columns=columns_to_keep)\n",
+    "\n",
+    "# 打印数据集列名\n",
+    "print(\"训练数据集列名:\", train_dataloader.get_col_names())\n",
+    "\n",
+    "# 处理数据集\n",
+    "train_dataset = remove_lens_column(train_dataloader)\n",
+    "valid_dataset = remove_lens_column(valid_dataloader)\n",
+    "test_dataset = remove_lens_column(test_dataloader)\n",
+    "\n",
+    "print(\"处理后训练数据集列名:\", train_dataset.get_col_names())\n"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": 16,
    "metadata": {},
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "Embedding<vocab_size=50260, embedding_size=768, use_one_hot=False, weight=Parameter (Tensor(shape=[50260, 768], dtype=Float32, value=[...], name=wte.weight), requires_grad=True), dtype=Float32, padding_idx=None>"
-      ]
-     },
-     "execution_count": 16,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
+   "outputs": [],
    "source": [
-    "model.config.pad_token_id = tokenizer.pad_token_id\n",
-    "model.resize_token_embeddings(model.config.vocab_size + num_added_toks)"
+    "# 定义前向函数(使用改进的版本)\n",
+    "def forward_fn(model, data):\n",
+    "    \"\"\"前向计算函数\"\"\"\n",
+    "    # 只在第一个批次时启用详细输出\n",
+    "    verbose = getattr(forward_fn, 'first_batch', False)\n",
+    "    if verbose:\n",
+    "        forward_fn.first_batch = False\n",
+    "    \n",
+    "    # 使用改进版本\n",
+    "    return improved_forward_fn(model, data, verbose=verbose)\n",
+    "\n",
+    "# 设置第一个批次标志\n",
+    "forward_fn.first_batch = True\n",
+    "\n",
+    "# 使用value_and_grad包装前向计算函数\n",
+    "grad_fn = value_and_grad(forward_fn, None, optimizer.parameters)\n"
    ]
   },
   {
-   "cell_type": "markdown",
+   "cell_type": "code",
+   "execution_count": 17,
    "metadata": {},
+   "outputs": [],
    "source": [
-    "给出的例子中微调算法采用IA3, 该算法通过学习向量来对激活层加权进行缩放，从而获得更强的性能，同时仅引入相对少量的新参数。"
+    "# 定义训练步骤(使用改进的版本)\n",
+    "def train_step(model, optimizer, data):\n",
+    "    \"\"\"执行一个训练步骤\"\"\"\n",
+    "    # 首先处理批次数据,确保包含所需字段\n",
+    "    processed_data = prepare_mrpc_batch(data)\n",
+    "    # 使用改进的训练步骤\n",
+    "    return improved_train_step(model, optimizer, processed_data, grad_fn)\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [],
    "source": [
-    "from mindnlp.peft import LoraConfig, get_peft_model\n",
-    "from mindnlp.peft import TaskType"
+    "# 定义评估函数\n",
+    "def evaluate(model, dataset, metric):\n",
+    "    \"\"\"评估模型\"\"\"\n",
+    "    metric.clear()\n",
+    "    model.set_train(False)\n",
+    "    \n",
+    "    for data in dataset.create_dict_iterator():\n",
+    "        # 处理批次数据,确保包含所需字段\n",
+    "        processed_data = prepare_mrpc_batch(data)\n",
+    "        \n",
+    "        # 确保标签单独传递\n",
+    "        if 'labels' in processed_data:\n",
+    "            labels = processed_data.pop('labels')\n",
+    "        elif 'label' in processed_data:\n",
+    "            labels = processed_data.pop('label')\n",
+    "        else:\n",
+    "            print(\"警告: 评估数据中没有标签!\")\n",
+    "            continue\n",
+    "        \n",
+    "        try:\n",
+    "            # 前向计算\n",
+    "            outputs = model(**processed_data)\n",
+    "            \n",
+    "            # 更新指标\n",
+    "            if isinstance(outputs, tuple):\n",
+    "                logits = outputs[0]\n",
+    "            elif hasattr(outputs, \"logits\"):\n",
+    "                # 处理SequenceClassifierOutputWithPast对象\n",
+    "                logits = outputs.logits\n",
+    "            else:\n",
+    "                logits = outputs\n",
+    "                \n",
+    "            metric.update(logits, labels)\n",
+    "        except Exception as e:\n",
+    "            print(f\"评估时出错: {e}\")\n",
+    "            continue\n",
+    "    \n",
+    "    # 计算指标\n",
+    "    result = metric.eval()\n",
+    "    model.set_train(True)\n",
+    "    \n",
+    "    return result\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 7. 模型训练"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 19,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "trainable params: 296,448 || all params: 124,737,792 || trainable%: 0.23765692437461133\n"
+      "检查数据格式...\n",
+      "\n",
+      "批次数据检查:\n",
+      "批次包含 4 个键:\n",
+      "  input_ids: 形状=(16, 256), 类型=mindspore.int64\n",
+      "    前5个值: [50256 50256 50256 50256     0]\n",
+      "  attention_mask: 形状=(16, 256), 类型=mindspore.int64\n",
+      "    前5个值: [1 1 1 1 0]\n",
+      "  token_type_ids: 形状=(16, 256), 类型=mindspore.int64\n",
+      "    前5个值: [0 0 1 1 0]\n",
+      "  labels: 形状=(16,), 类型=mindspore.int64\n",
+      "    前5个值: [1 0 1 0 1]\n",
+      "    唯一标签值: [0 1]\n"
      ]
     }
    ],
    "source": [
-    "if args.is_lora:\n",
-    "    # build peft model\n",
-    "    peft_config = LoraConfig(task_type=TaskType.SEQ_CLS, inference_mode=False, r=8, lora_alpha=32, lora_dropout=0.1, fan_in_fan_out=True)\n",
-    "    model = get_peft_model(model, peft_config)\n",
-    "    # print(model)\n",
-    "    model.print_trainable_parameters()"
+    "# 在开始训练前检查一个批次数据\n",
+    "print(\"检查数据格式...\")\n",
+    "first_batch = next(train_dataset.create_dict_iterator())\n",
+    "examine_batch(first_batch)\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "from mindnlp._legacy.engine import Trainer, Evaluator\n",
-    "from mindnlp._legacy.metrics import Accuracy\n",
-    "from mindnlp._legacy.engine.callbacks import CheckpointCallback, BestModelCallback"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 21,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# optimizer\n",
-    "optimizer = AdamWeightDecay(params=model.trainable_params(), learning_rate=args.lr)\n",
-    "\n",
-    "metric = Accuracy()\n",
-    "ckpoint_cb = CheckpointCallback(save_path=args.save_dir, ckpt_name='gpt2_mrpc_finetune', epochs=1, keep_checkpoint_max=2)\n",
-    "best_model_cb = BestModelCallback(save_path=args.save_dir, ckpt_name='gpt2_mrpc_finetune_best', auto_load=True)\n",
-    "\n",
-    "if args.debug:\n",
-    "    args.num_epochs = 2\n",
-    "\n",
-    "trainer = Trainer(network=model, train_dataset=train_dataloader,\n",
-    "                  eval_dataset=valid_dataloader, metrics=metric,\n",
-    "                  epochs=args.num_epochs, optimizer=optimizer, \n",
-    "                  callbacks=[ckpoint_cb, best_model_cb],\n",
-    "                  jit=False)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": 20,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "The train will start from the checkpoint saved in '.mindnlp/peft_model/mrpc_IA3'.\n"
-     ]
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Epoch 0: 100%|██████████| 459/459 [03:05<00:00,  2.47it/s, loss=0.6872016] \n"
-     ]
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Checkpoint: 'gpt2_mrpc_finetune_epoch_0.ckpt' has been saved in epoch: 0.\n"
+      "开始训练: 1个epoch, 共230步\n"
      ]
     },
     {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Evaluate: 100%|██████████| 51/51 [00:07<00:00,  7.18it/s]\n"
-     ]
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Evaluate Score: {'Accuracy': 0.6838235294117647}\n",
-      "---------------Best Model: 'gpt2_mrpc_finetune_best.ckpt' has been saved in epoch: 0.---------------\n"
-     ]
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Epoch 1: 100%|██████████| 459/459 [03:04<00:00,  2.49it/s, loss=0.6677042] \n"
-     ]
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "49bd450da0d5491cb4e37870386dffdc",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "  0%|          | 0/230 [00:00<?, ?it/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Checkpoint: 'gpt2_mrpc_finetune_epoch_1.ckpt' has been saved in epoch: 1.\n"
+      "数据包含以下键: ['input_ids', 'attention_mask', 'token_type_ids', 'labels']\n",
+      "使用'labels'作为标签\n",
+      "Epoch 1/1, 平均损失: 2.0934\n",
+      "模型已保存至 ./saved_models/epoch_1\n",
+      "验证集准确率: 0.5760\n",
+      "找到更好的模型! 已保存至 ./saved_models/best_model\n"
      ]
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Evaluate: 100%|██████████| 51/51 [00:07<00:00,  7.21it/s]\n"
-     ]
-    },
+    }
+   ],
+   "source": [
+    "# 训练循环\n",
+    "model.set_train(True)\n",
+    "best_acc = 0\n",
+    "total_steps = args.num_epochs * train_dataset.get_dataset_size()\n",
+    "\n",
+    "print(f\"开始训练: {args.num_epochs}个epoch, 共{total_steps}步\")\n",
+    "\n",
+    "for epoch in range(args.num_epochs):\n",
+    "    # 训练一个epoch\n",
+    "    model.set_train(True)\n",
+    "    train_loss = 0\n",
+    "    train_steps = 0\n",
+    "    \n",
+    "    # 重置第一个批次标志（每个epoch只在第一个批次详细输出）\n",
+    "    forward_fn.first_batch = epoch == 0\n",
+    "    \n",
+    "    progress_bar = tqdm(train_dataset.create_dict_iterator(), total=train_dataset.get_dataset_size())\n",
+    "    for batch in progress_bar:\n",
+    "        loss = train_step(model, optimizer, batch)\n",
+    "        \n",
+    "        # 错误处理\n",
+    "        if isinstance(loss, mindspore.Tensor):\n",
+    "            try:\n",
+    "                loss_value = loss.asnumpy()\n",
+    "                train_loss += loss_value\n",
+    "            except:\n",
+    "                print(f\"警告: 无法转换损失值 {loss}\")\n",
+    "                train_loss += 1000.0  # 使用一个大的默认值\n",
+    "        else:\n",
+    "            # 处理SequenceClassifierOutputWithPast类型的返回值\n",
+    "            try:\n",
+    "                if hasattr(loss, \"loss\"):\n",
+    "                    loss_value = loss.loss.asnumpy()\n",
+    "                    train_loss += loss_value\n",
+    "                else:\n",
+    "                    print(f\"警告: 返回对象没有loss属性 {type(loss)}\")\n",
+    "                    train_loss += 1000.0  # 使用一个大的默认值\n",
+    "            except Exception as e:\n",
+    "                print(f\"警告: 处理损失值时出错 {e}\")\n",
+    "                train_loss += 1000.0  # 使用一个大的默认值\n",
+    "            \n",
+    "        train_steps += 1\n",
+    "        \n",
+    "        # 更新进度条\n",
+    "        progress_bar.set_description(f\"Epoch {epoch+1}/{args.num_epochs}\")\n",
+    "        progress_bar.set_postfix(loss=train_loss/train_steps)\n",
+    "        \n",
+    "        # 在debug模式下,训练几个批次后停止\n",
+    "        if args.debug and train_steps >= 5:\n",
+    "            print(\"Debug模式: 提前停止训练\")\n",
+    "            break\n",
+    "    \n",
+    "    # 计算该epoch的平均损失\n",
+    "    avg_loss = train_loss / train_steps\n",
+    "    print(f\"Epoch {epoch+1}/{args.num_epochs}, 平均损失: {avg_loss:.4f}\")\n",
+    "    \n",
+    "    # 保存模型检查点\n",
+    "    try:\n",
+    "        # 使用模型自带的save_pretrained方法替代mindspore.save_checkpoint\n",
+    "        save_dir = os.path.join(args.save_dir, f\"epoch_{epoch+1}\")\n",
+    "        os.makedirs(save_dir, exist_ok=True)\n",
+    "        model.save_pretrained(save_dir)\n",
+    "        print(f\"模型已保存至 {save_dir}\")\n",
+    "    except Exception as e:\n",
+    "        print(f\"保存模型时出错: {e}\")\n",
+    "    \n",
+    "    # 评估\n",
+    "    acc = evaluate(model, valid_dataset, metric)\n",
+    "    print(f\"验证集准确率: {acc:.4f}\")\n",
+    "    \n",
+    "    # 保存最佳模型\n",
+    "    if acc > best_acc:\n",
+    "        best_acc = acc\n",
+    "        try:\n",
+    "            # 使用模型自带的save_pretrained方法替代mindspore.save_checkpoint\n",
+    "            best_save_dir = os.path.join(args.save_dir, \"best_model\")\n",
+    "            os.makedirs(best_save_dir, exist_ok=True)\n",
+    "            model.save_pretrained(best_save_dir)\n",
+    "            print(f\"找到更好的模型! 已保存至 {best_save_dir}\")\n",
+    "        except Exception as e:\n",
+    "            print(f\"保存最佳模型时出错: {e}\")\n",
+    "    \n",
+    "    # 在debug模式下,只运行一个epoch\n",
+    "    if args.debug:\n",
+    "        print(\"Debug模式: 提前停止训练\")\n",
+    "        break\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 8. 在测试集上评估并保存最终模型"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "metadata": {},
+   "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Evaluate Score: {'Accuracy': 0.6838235294117647}\n",
-      "Loading best model from '.mindnlp/peft_model/mrpc_IA3' with '['Accuracy']': [0.6838235294117647]...\n",
-      "---------------The model is already load the best model from 'gpt2_mrpc_finetune_best.ckpt'.---------------\n"
+      "在测试集上评估...\n",
+      "测试集准确率: 0.5832\n"
      ]
     }
    ],
    "source": [
-    "trainer.run(tgt_columns=\"labels\")"
+    "# 在测试集上评估\n",
+    "print(\"在测试集上评估...\")\n",
+    "test_acc = evaluate(model, test_dataset, metric)\n",
+    "print(f\"测试集准确率: {test_acc:.4f}\")\n"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 23,
+   "execution_count": 22,
    "metadata": {},
    "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Evaluate: 100%|██████████| 216/216 [00:30<00:00,  7.13it/s]"
-     ]
-    },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Evaluate Score: {'Accuracy': 0.664927536231884}\n"
-     ]
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "\n"
+      "最终模型和tokenizer已保存至 ./saved_models/final_model\n",
+      "训练信息已记录到README.md\n"
      ]
     }
    ],
    "source": [
-    "evaluator = Evaluator(network=model, eval_dataset=test_dataloader, metrics=metric)\n",
-    "evaluator.run(tgt_columns=\"labels\")"
+    "# 保存最终模型和tokenizer\n",
+    "try:\n",
+    "    final_save_dir = os.path.join(args.save_dir, \"final_model\")\n",
+    "    os.makedirs(final_save_dir, exist_ok=True)\n",
+    "    \n",
+    "    # 保存模型\n",
+    "    model.save_pretrained(final_save_dir)\n",
+    "    \n",
+    "    # 保存tokenizer\n",
+    "    tokenizer.save_pretrained(final_save_dir)\n",
+    "    \n",
+    "    print(f\"最终模型和tokenizer已保存至 {final_save_dir}\")\n",
+    "    \n",
+    "    # 创建README文件，记录训练信息\n",
+    "    readme_path = os.path.join(final_save_dir, \"README.md\")\n",
+    "    with open(readme_path, \"w\", encoding=\"utf-8\") as f:\n",
+    "        f.write(f\"# MRPC数据集上的GPT2模型\\n\\n\")\n",
+    "        f.write(f\"## 训练信息\\n\")\n",
+    "        f.write(f\"- 模型: {args.model_name}\\n\")\n",
+    "        f.write(f\"- 数据集: MRPC\\n\")\n",
+    "        f.write(f\"- 训练轮数: {args.num_epochs}\\n\")\n",
+    "        f.write(f\"- 批次大小: {args.batch_size}\\n\")\n",
+    "        f.write(f\"- 学习率: {args.lr}\\n\")\n",
+    "        f.write(f\"- 验证集准确率: {best_acc:.4f}\\n\")\n",
+    "        f.write(f\"- 测试集准确率: {test_acc:.4f}\\n\")\n",
+    "        \n",
+    "    print(\"训练信息已记录到README.md\")\n",
+    "except Exception as e:\n",
+    "    print(f\"保存最终模型时出错: {e}\")\n"
    ]
   }
  ],
  "metadata": {
-  "AIGalleryInfo": {
-   "item_id": "1325b719-fc78-46c4-8f47-9f3623e9b0f4"
-  },
-  "flavorInfo": {
-   "architecture": "X86_64",
-   "category": "GPU"
-  },
-  "imageInfo": {
-   "id": "e1a07296-22a8-4f05-8bc8-e936c8e54202",
-   "name": "mindspore1.7.0-cuda10.1-py3.7-ubuntu18.04"
-  },
   "kernelspec": {
-   "display_name": "python-3.9.0",
+   "display_name": "mindspore_py39",
    "language": "python",
-   "name": "python-3.9.0"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -1035,7 +884,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.0"
+   "version": "3.9.19"
   }
  },
  "nbformat": 4,
diff --git a/Season2.step_into_llm/09.PEFT/train_llama_lora/fix_mrpc_training.py b/Season2.step_into_llm/09.PEFT/train_llama_lora/fix_mrpc_training.py
new file mode 100644
index 0000000..c547381
--- /dev/null
+++ b/Season2.step_into_llm/09.PEFT/train_llama_lora/fix_mrpc_training.py
@@ -0,0 +1,141 @@
+"""
+用于修复MRPC数据集训练的辅助函数
+"""
+
+import mindspore
+from mindspore import Tensor, ops
+
+def print_dataset_keys(dataset):
+    """打印数据集的列名"""
+    print(f"数据集列名: {dataset.get_col_names()}")
+    # 尝试获取一个样本并打印其键
+    iterator = dataset.create_dict_iterator()
+    try:
+        sample = next(iterator)
+        print(f"样本中的键: {list(sample.keys())}")
+        for key, value in sample.items():
+            print(f"  {key}: 形状={value.shape}, 类型={value.dtype}")
+    except:
+        print("无法获取样本数据")
+
+def improved_forward_fn(model, data, verbose=False):
+    """改进的前向计算函数,有更好的错误处理"""
+    # 移除不支持的字段
+    data_copy = dict(data)  # 创建数据的副本以避免修改原始数据
+    
+    if 'lens' in data_copy:
+        del data_copy['lens']
+    
+    # 检查数据键
+    if verbose:
+        print(f"数据包含以下键: {list(data_copy.keys())}")
+    
+    # 检查'labels'是否存在,如果不存在查找替代键
+    label_keys = ['labels', 'label']
+    label_found = False
+    
+    for key in label_keys:
+        if key in data_copy:
+            labels = data_copy.pop(key)
+            label_found = True
+            if verbose:
+                print(f"使用'{key}'作为标签")
+            break
+    
+    if not label_found:
+        # 如果找不到标签,创建一个虚拟标签
+        if verbose:
+            print("警告: 找不到标签字段,使用虚拟标签")
+        # 如果有input_ids,使用其批次大小
+        if 'input_ids' in data_copy:
+            batch_size = data_copy['input_ids'].shape[0]
+        else:
+            batch_size = 1
+        labels = ops.zeros(batch_size, dtype=mindspore.int32)
+    
+    # 使用模型进行前向计算
+    try:
+        outputs = model(**data_copy, labels=labels)
+        
+        # 获取损失
+        if isinstance(outputs, tuple):
+            loss = outputs[0]
+        elif hasattr(outputs, "loss"):
+            # 处理SequenceClassifierOutputWithPast对象
+            loss = outputs.loss
+        else:
+            loss = outputs
+            
+        return loss
+    except Exception as e:
+        print(f"前向计算出错: {e}")
+        # 返回一个默认的损失
+        return Tensor(1000.0)
+
+def improved_train_step(model, optimizer, data, grad_fn):
+    """改进的训练步骤函数,有更好的错误处理"""
+    try:
+        # 计算损失和梯度
+        loss, grads = grad_fn(model, data)
+        
+        # 应用梯度
+        optimizer(grads)
+        
+        return loss
+    except Exception as e:
+        print(f"训练步骤出错: {e}")
+        # 返回一个默认的损失
+        return Tensor(1000.0)
+
+def examine_batch(batch):
+    """检查批次数据结构"""
+    print("\n批次数据检查:")
+    
+    if not isinstance(batch, dict):
+        print(f"批次不是字典类型,而是 {type(batch)}")
+        return
+    
+    print(f"批次包含 {len(batch)} 个键:")
+    
+    for key, value in batch.items():
+        if isinstance(value, Tensor):
+            print(f"  {key}: 形状={value.shape}, 类型={value.dtype}")
+            # 打印值的摘要
+            if value.size < 10:
+                print(f"    值: {value}")
+            else:
+                flat_value = value.reshape(-1)
+                print(f"    前5个值: {flat_value[:5]}")
+                
+                # 检查是否是标签
+                if key.lower() in ['label', 'labels']:
+                    unique_labels = ops.unique(flat_value)[0]
+                    print(f"    唯一标签值: {unique_labels}")
+        else:
+            print(f"  {key}: 类型={type(value)}")
+
+def prepare_mrpc_batch(batch):
+    """准备MRPC批次以进行训练"""
+    # 移除不支持的字段
+    result = {}
+    
+    # 复制所有支持的字段
+    for key in ['input_ids', 'attention_mask', 'token_type_ids']:
+        if key in batch:
+            result[key] = batch[key]
+    
+    # 处理标签
+    if 'labels' in batch:
+        result['labels'] = batch['labels']
+    elif 'label' in batch:
+        # 重命名label为labels
+        result['labels'] = batch['label']
+    else:
+        # 如果没有标签,创建一个虚拟标签
+        if 'input_ids' in batch:
+            batch_size = batch['input_ids'].shape[0]
+        else:
+            batch_size = 1
+        result['labels'] = ops.zeros(batch_size, dtype=mindspore.int32)
+    
+    return result 
\ No newline at end of file
diff --git a/Season2.step_into_llm/09.PEFT/train_llama_lora/mrpc_dataset.py b/Season2.step_into_llm/09.PEFT/train_llama_lora/mrpc_dataset.py
deleted file mode 100644
index ea3a444..0000000
--- a/Season2.step_into_llm/09.PEFT/train_llama_lora/mrpc_dataset.py
+++ /dev/null
@@ -1,169 +0,0 @@
-
-import json
-import copy
-
-from mindspore.dataset import RandomSampler, DistributedSampler, NumpySlicesDataset, SequentialSampler
-from mindnlp.dataset import MRPC
-
-class InputExample(object):
-    def __init__(self, guid, text_a, text_b=None, label=None):
-        self.guid = guid
-        self.text_a = text_a
-        self.text_b = text_b
-        self.label = label
-
-    def __repr__(self):
-        return str(self.to_json_string())
-
-    def to_dict(self):
-        """Serializes this instance to a Python dictionary."""
-        output = copy.deepcopy(self.__dict__)
-        return output
-
-    def to_json_string(self):
-        """Serializes this instance to a JSON string."""
-        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n"    
-
-class InputFeatures(object):
-    def __init__(self, input_ids, attention_mask, token_type_ids, label,input_len):
-        self.input_ids = input_ids
-        self.attention_mask = attention_mask
-        self.token_type_ids = token_type_ids
-        self.input_len = input_len
-        self.label = label
-
-    def __repr__(self):
-        return str(self.to_json_string())
-
-    def to_dict(self):
-        """Serializes this instance to a Python dictionary."""
-        output = copy.deepcopy(self.__dict__)
-        return output
-
-    def to_json_string(self):
-        """Serializes this instance to a JSON string."""
-        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n"
-    
-
-def convert_dataset_to_examples(ds):
-    """Convert dataset to examples."""
-    examples = []
-    iter = ds.create_tuple_iterator()
-    for i, (label, text_a, text_b) in enumerate(iter):
-        # print(str(text_a.asnumpy()), str(text_b.asnumpy()))
-        examples.append(
-            InputExample(guid=i, text_a=str(text_a.asnumpy()), text_b=str(text_b.asnumpy()), label=int(label))
-        )
-    
-    return examples
-
-def _truncate_seq_pair(tokens_a, tokens_b, max_length):
-    """Truncates a sequence pair in place to the maximum length."""
-    while True:
-        total_length = len(tokens_a) + len(tokens_b)
-        if total_length <= max_length:
-            break
-        if len(tokens_a) > len(tokens_b):
-            tokens_a.pop()
-        else:
-            tokens_b.pop()
-
-
-def convert_examples_to_features(examples, tokenizer, max_seq_length=512):
-    features = []
-
-    for ex_index, example in enumerate(examples):
-        tokenizer.return_token = True
-        tokens_a = tokenizer.execute_py(example.text_a)
-        tokens_b  = None
-        if example.text_b:
-            tokens_b = tokenizer.execute_py(example.text_b)
-        if tokens_b is not None:
-            # Modifies `tokens_a` and `tokens_b` in place so that the total
-            # length is less than the specified length.
-            # Account for [CLS], [SEP], [SEP] with "- 3"
-            _truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3)
-        else:
-            # Account for [CLS] and [SEP] with "- 2"
-            if len(tokens_a) > max_seq_length - 2:
-                tokens_a = tokens_a[0:(max_seq_length - 2)]
-
-        tokens = []
-        token_type_ids = []
-        for token in tokens_a:
-            tokens.append(token)
-            token_type_ids.append(0)
-
-        if tokens_b is not None:
-            for token in tokens_b[1:]:
-                tokens.append(token)
-                token_type_ids.append(1)
-            # tokens.append("[SEP]")
-            # token_type_ids.append(1)
-
-        tokenizer.return_token=False
-        # input_ids = tokenizer.execute_py(example.text_a).tolist() + tokenizer.execute_py(example.text_b).tolist()
-        input_ids = tokenizer.convert_tokens_to_ids(tokens)
-
-        # print(tokenizer.execute_py(np.array(tokens)).tolist())
-        # The mask has 1 for real tokens and 0 for padding tokens. Only real
-        # tokens are attended to.
-        attention_mask = [1] * len(input_ids)
-        input_len = len(input_ids)
-
-        # Zero-pad up to the sequence length.
-        while len(input_ids) < max_seq_length:
-            input_ids.append(0)
-            attention_mask.append(0)
-            token_type_ids.append(0)
-
-        assert len(input_ids) == max_seq_length
-        assert len(attention_mask) == max_seq_length
-        assert len(token_type_ids) == max_seq_length
-        
-        label_id = example.label
-
-        # if ex_index < 5:
-        #     print("*** Example ***")
-        #     print("guid: %s" % (example.guid))
-        #     print("tokens: %s"%" ".join([str(x) for x in tokens]))
-        #     print("input_ids: %s" % " ".join([str(x) for x in input_ids]))
-        #     print("attention_mask: %s" % " ".join([str(x) for x in attention_mask]))
-        #     print("token_type_ids: %s" % " ".join([str(x) for x in token_type_ids]))
-        #     print("label: %s (id = %d)" % (example.label, label_id))
-        #     print("input length: %d" % (input_len))
-
-        features.append(
-            InputFeatures(input_ids=input_ids,
-                          attention_mask=attention_mask,
-                          token_type_ids=token_type_ids,
-                          label=label_id,
-                          input_len=input_len)
-        )
-    return features
-
-def load_examples(tokenizer, max_seq_length, data_type="train"):
-    """load_examples using load_dataset"""
-    mrpc_train, mrpc_test = MRPC()
-    mrpc = mrpc_train if data_type == "train" else mrpc_test
-    train_examples = convert_dataset_to_examples(mrpc)
-    # test_examples = convert_dataset_to_examples(mrpc_test)
-
-    features = convert_examples_to_features(train_examples, tokenizer, max_seq_length=max_seq_length)
-
-    # Convert to Tensors and build dataset
-    all_input_ids = [f.input_ids for f in features]
-    all_attention_mask = [f.attention_mask for f in features]
-    all_token_type_ids = [f.token_type_ids for f in features]
-    all_lens = [f.input_len for f in features]
-    all_labels = [f.label for f in features]
-    dataset = ((all_input_ids, all_attention_mask, all_token_type_ids, all_lens, all_labels))
-
-    return dataset
-
-def get_dataloader_from_ds(ds, batch_size):
-    train_sampler = SequentialSampler()
-    train_dataloader = NumpySlicesDataset(ds, sampler=train_sampler)
-    train_dataloader = train_dataloader.batch(batch_size)
-
-    return train_dataloader
\ No newline at end of file
diff --git a/Season2.step_into_llm/09.PEFT/train_llama_lora/train.py b/Season2.step_into_llm/09.PEFT/train_llama_lora/train.py
deleted file mode 100644
index 6ef20d2..0000000
--- a/Season2.step_into_llm/09.PEFT/train_llama_lora/train.py
+++ /dev/null
@@ -1,184 +0,0 @@
-import os
-import copy
-import argparse
-import logging
-import mindspore
-import numpy as np
-
-from sentencepiece import SentencePieceProcessor
-from typing import List
-from tqdm import tqdm 
-from mindspore import nn
-from mindspore.nn import AdamWeightDecay
-from mindnlp.transformers import LlamaForCausalLM, LlamaConfig, LlamaForSequenceClassification, LlamaPreTrainedModel
-from mindnlp.transformers import LlamaTokenizer
-from mindnlp.peft import (
-    get_peft_config,
-    get_peft_model,
-    LoraConfig,
-    PeftType,
-    PeftConfig,
-    PeftModel,
-)
-
-from mrpc_dataset import load_examples, get_dataloader_from_ds
-
-def load_llama_model(config_path, model_path, token_path):
-    config = LlamaConfig.from_pretrained(config_path, problem_type="single_label_classification")
-    # print(config.to_dict())
-
-    # here we build model from config to make model smaller and test the correctness of training process
-    # mindspore.load_checkpoint(ckpt_path, net=model)
-    config.num_hidden_layers = 2
-    # model = LlamaForCausalLM(config)  
-    model = LlamaForSequenceClassification(config)
-    state_dict = mindspore.load_checkpoint(model_path)
-    param_not_load, ckpt_not_load = mindspore.load_param_into_net(model, state_dict)
-    logging.info("params in model not load\n", param_not_load)
-    logging.info("ckpt not load:\n", ckpt_not_load)
-    # print(model)
-    tokenizer = LlamaTokenizer.from_pretrained(token_path)
-
-    return model, config, tokenizer
-
-def forward_fn(input_ids, attention_mask, token_type_ids, lens, labels):
-    # _, _ = hidden_states, attentions``
-    loss, logits, _ = model(
-        input_ids=input_ids,
-        attention_mask=attention_mask,
-        # token_type_ids=token_type_ids,
-        labels=mindspore.Tensor(labels.asnumpy(), mindspore.int32)
-    )
-    # RobertaForSequenceClassification Model has loss_fn (mse or cross_entropy)
-    # loss = criterion(logits, label)
-    return loss, logits
-
-
-def train_one_epoch(model, optimizer, criterion, train_dataloader):
-    model.set_train(True)
-    total_loss, total_step = 0, 0
-    num_batches = len(train_dataloader)
-    preds, label_ids = None, None
-
-    grad_fn = mindspore.value_and_grad(forward_fn, None, optimizer.parameters, has_aux=True)
-
-    with tqdm(total=num_batches) as t:
-        for i, (input_ids, attention_mask, token_type_ids, lens, labels) in enumerate(train_dataloader):
-            # forward + grad
-            (loss, logits), grad = grad_fn(input_ids, attention_mask, token_type_ids, lens, labels)
-            # update model params
-            optimizer(grad)
-            total_loss += loss.asnumpy()
-            total_step += 1
-            curr_loss = total_loss / total_step  # 当前的平均loss
-
-            if preds is None:
-                preds = np.argmax(logits.asnumpy(), axis=1)
-                label_ids = labels.asnumpy()
-            else:
-                preds = np.append(preds, np.argmax(logits.asnumpy(), axis=1), axis=0)
-                label_ids = np.append(label_ids, labels.asnumpy(), axis=0)
-
-            t.set_postfix({'train-loss': f'{curr_loss:.2f}'})
-            t.update(1)
-    acc = (preds == label_ids).mean()
-
-    return total_loss / total_step, acc
-
-def eval_one_epoch(model, optimizer, criterion, eval_dataloader):
-    model.set_train(False)
-    total_loss, total_step = 0, 0
-    preds, label_ids = None, None
-    num_batches = len(eval_dataloader)
-
-    with tqdm(total=num_batches) as t:
-        for i, (input_ids, attention_mask, token_type_ids, lens, labels) in enumerate(eval_dataloader):
-            loss, logits, _ = model(
-                input_ids=input_ids,
-                attention_mask=attention_mask,
-                # token_typ_ids=token_type_ids,
-                labels=mindspore.Tensor(labels.asnumpy(), mindspore.int32)
-            )
-            # (loss, logits), grad = grad_fn(input_ids, attention_mask, token_type_ids, lens, labels)
-            total_loss += loss.asnumpy()
-            total_step += 1
-            curr_loss = total_loss / total_step  # 当前的平均loss
-            if preds is None:
-                preds = np.argmax(logits.asnumpy(), axis=1)
-                label_ids = labels.asnumpy()
-            else:
-                preds = np.append(preds, np.argmax(logits.asnumpy(), axis=1), axis=0)
-                label_ids = np.append(label_ids, labels.asnumpy(), axis=0)
-            t.set_postfix({'eval-loss': f'{curr_loss:.2f}'})
-            t.update(1)
-
-    # compute metrics
-    acc = (preds == label_ids).mean()
-
-    return total_loss / total_step, acc
-
-
-def train(model, optimizer, criterion, train_dataloader, eval_dataloader, epochs):
-    # train start from here
-    for epoch in range(1, epochs+1):
-        train_loss, train_acc = train_one_epoch(model, optimizer, criterion, train_dataloader)
-        eval_loss, eval_acc = eval_one_epoch(model, optimizer, criterion, eval_dataloader)
-
-        print(f"epoch:{epoch} train_loss:{train_loss} eval_acc:{train_acc} \
-              eval_loss:{eval_loss} eval_acc:{eval_acc}")
-        # print(f"epoch:{epoch} train_loss:{train_loss} eval_loss:{eval_loss}")
-
-def eval_model(model, optimizer, criterion, eval_dataloader):
-    eval_loss, eval_acc = eval_one_epoch(model, optimizer, criterion, eval_dataloader)
-    print(f"eval_loss:{eval_loss} eval_acc:{eval_acc}")
-
-if __name__ == "__main__":
-    # `python llm/peft/train_llama_lora/train.py --do_train --do_eval --model_name_or_path bert-base-cased`
-    # from pretrained
-    parser = argparse.ArgumentParser()
-    # parser.add_argument("--do_train", action='store_true', help="Whether to run training.")
-    # parser.add_argument("--do_eval", action='store_true', help="Whether to run eval on the dev set.")
-    parser.add_argument("--save_dir", default=".mindnlp/peft_model/mrpc_lora", type=str, help="The output directory where the model checkpoints will be written.")
-    parser.add_argument("--batch_size", default=16, type=int, help="Batch size per GPU/CPU for training.")
-    parser.add_argument("--model_name_or_path", default="roberta-base", type=str, help="bert-base-cased")
-    parser.add_argument("--num_epochs", default=5, type=int)
-    parser.add_argument("--lr", default=1e-4, type=float, help="Set 2e-5 for full-finetuning.")
-    parser.add_argument("--max_seq_len", default=256, type=int)
-    parser.add_argument("--debug", action="store_true", help="debug mode")
-    parser.add_argument("--lora", action="store_true", help="lora mode")
-    
-    args = parser.parse_args()
-    if args.debug:
-        args.num_epochs = 1
-
-    # load model
-    model, config, tokenizer = load_llama_model(
-        config_path="/home/cjl/code/mind/collect/mindnlp-models/llama-2-7b-hf/config.json",
-        model_path="/home/cjl/code/mind/collect/mindnlp-models/convert/llama.ckpt",
-        token_path="/home/cjl/code/mind/collect/mindnlp-models/llama-2-7b-hf/tokenizer.model"
-    )
-    logging.info("model load")
-
-    if args.lora:
-        # build peft model
-        peft_config = LoraConfig(task_type="SEQ_CLS", inference_mode=False, r=8, lora_alpha=16, lora_dropout=0.1)
-        model = get_peft_model(model, peft_config)
-        # print(model)
-        model.print_trainable_parameters()
-
-    # load dataset
-    train_ds = load_examples(tokenizer, args.max_seq_len, data_type="train")
-    test_ds = load_examples(tokenizer, args.max_seq_len, data_type="test")
-    train_dataloader = get_dataloader_from_ds(train_ds, args.batch_size)
-    test_dataloader = get_dataloader_from_ds(test_ds, args.batch_size)
-    logging.info("dataset load")
-
-    # optimizer
-    optimizer = AdamWeightDecay(params=model.trainable_params(), learning_rate=args.lr)
-    loss_fn = nn.CrossEntropyLoss()  
-
-
-    logging.info("start train")
-    train(model, optimizer, loss_fn, train_dataloader, test_dataloader, epochs=args.num_epochs)
-    logging.info("end train")
-    model.save_pretrained(save_directory=args.save_dir)
\ No newline at end of file
diff --git a/Season2.step_into_llm/09.PEFT/train_llama_lora/train_mrpc_peft.py b/Season2.step_into_llm/09.PEFT/train_llama_lora/train_mrpc_peft.py
new file mode 100644
index 0000000..c459c8f
--- /dev/null
+++ b/Season2.step_into_llm/09.PEFT/train_llama_lora/train_mrpc_peft.py
@@ -0,0 +1,503 @@
+"""
+完整的MRPC数据集PEFT训练脚本,修复了标签问题
+"""
+
+import os
+import argparse
+import json
+import copy
+import mindspore
+from mindspore import context, Tensor, ops
+from mindspore.dataset import NumpySlicesDataset, SequentialSampler
+from mindspore.common.parameter import Parameter
+from mindspore.nn import AdamWeightDecay
+from mindnlp.engine import Evaluator
+from mindnlp.metrics import Accuracy
+from mindnlp.common.grad import value_and_grad
+from mindnlp.dataset import load_dataset
+from mindnlp.transformers import GPT2Tokenizer, GPT2ForSequenceClassification
+from mindnlp.peft import LoraConfig, get_peft_model, TaskType
+from tqdm.auto import tqdm
+
+# 导入我们的辅助函数
+from fix_mrpc_training import (
+    print_dataset_keys,
+    improved_forward_fn,
+    improved_train_step,
+    examine_batch,
+    prepare_mrpc_batch
+)
+
+class InputExample(object):
+    def __init__(self, guid, text_a, text_b=None, label=None):
+        """
+            InputExample表示单个输入示例
+            包含一个全局唯一标识符（guid）、文本 A（text_a）、可选的文本 B（text_b）和标签（label）
+        """
+        self.guid = guid
+        self.text_a = text_a
+        self.text_b = text_b
+        self.label = label
+
+    def __repr__(self):
+        return str(self.to_json_string())
+
+    def to_dict(self):
+        """Serializes this instance to a Python dictionary."""
+        output = copy.deepcopy(self.__dict__)
+        return output
+
+    def to_json_string(self):
+        """Serializes this instance to a JSON string."""
+        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n"    
+
+class InputFeatures(object):
+    def __init__(self, input_ids, attention_mask, token_type_ids, label, input_len):
+        """
+            InputFeatures 表示模型输入特征
+            包含输入 ID（input_ids）、注意力掩码（attention_mask）、标记类型 ID（token_type_ids）、标签（label）和输入长度（input_len）
+        """
+        self.input_ids = input_ids
+        self.attention_mask = attention_mask
+        self.token_type_ids = token_type_ids
+        self.input_len = input_len
+        self.label = label
+
+    def __repr__(self):
+        return str(self.to_json_string())
+
+    def to_dict(self):
+        """Serializes this instance to a Python dictionary."""
+        output = copy.deepcopy(self.__dict__)
+        return output
+
+    def to_json_string(self):
+        """Serializes this instance to a JSON string."""
+        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n"
+
+def convert_dataset_to_examples(ds):
+    """
+        Convert dataset to examples.
+        将数据集ds转换为 InputExample 实例列表examples
+        使用 mindspore.dataset 的迭代器遍历数据集，并将每个样本转换为 InputExample 对象。
+    """
+    examples = []
+    iter0 = ds.create_tuple_iterator()
+    for i, (text_a, text_b, label, idx, label_text) in enumerate(iter0):
+        examples.append(
+            InputExample(guid=i, text_a=str(text_a.asnumpy()), text_b=str(text_b.asnumpy()), label=int(label))
+        )
+    
+    return examples
+
+def _truncate_seq_pair(tokens_a, tokens_b, max_length):
+    """
+        Truncates a sequence pair in place to the maximum length.
+        即保持文本对的意义，同时截断文本对，使其总长度不超过指定的最大长度max_length
+    """
+    while True:
+        total_length = len(tokens_a) + len(tokens_b)
+        if total_length <= max_length:
+            break
+        # 优先选择文本更长的文本进行截断
+        if len(tokens_a) > len(tokens_b):
+            tokens_a.pop()
+        else:
+            tokens_b.pop()
+
+def convert_examples_to_features(examples, tokenizer, max_seq_length=512):
+    """
+        将 InputExample 实例列表转换为 InputFeatures 实例列表。
+        使用 tokenizer 对文本进行编码，生成模型输入所需的特征。
+    """
+    features = []
+
+    for ex_index, example in enumerate(examples):
+        tokenizer.return_token = True
+        tokens_a = tokenizer(example.text_a)
+        tokens_b  = None
+        if example.text_b:
+            tokens_b = tokenizer(example.text_b)
+        if tokens_b is not None:
+            # Modifies `tokens_a` and `tokens_b` in place so that the total
+            # length is less than the specified length.
+            # Account for [CLS], [SEP], [SEP] with "- 3"
+            _truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3)
+        else:
+            # Account for [CLS] and [SEP] with "- 2"
+            if len(tokens_a) > max_seq_length - 2:
+                tokens_a = tokens_a[0:(max_seq_length - 2)]
+
+        tokens = []
+        token_type_ids = []
+        for token in tokens_a:
+            tokens.append(token)
+            token_type_ids.append(0)
+
+        if tokens_b is not None:
+            for token in tokens_b[1:]:
+                tokens.append(token)
+                token_type_ids.append(1)
+
+        tokenizer.return_token=False
+        input_ids = tokenizer.convert_tokens_to_ids(tokens)
+
+        # The mask has 1 for real tokens and 0 for padding tokens. Only real
+        # tokens are attended to.
+        attention_mask = [1] * len(input_ids)
+        input_len = len(input_ids)
+
+        # Zero-pad up to the sequence length.
+        while len(input_ids) < max_seq_length:
+            input_ids.append(0)
+            attention_mask.append(0)
+            token_type_ids.append(0)
+
+        assert len(input_ids) == max_seq_length
+        assert len(attention_mask) == max_seq_length
+        assert len(token_type_ids) == max_seq_length
+        
+        label_id = example.label
+
+        features.append(
+            InputFeatures(input_ids=input_ids,
+                          attention_mask=attention_mask,
+                          token_type_ids=token_type_ids,
+                          label=label_id,
+                          input_len=input_len)
+        )
+    return features
+
+def load_examples(tokenizer, max_seq_length, mrpc_datas):
+    """load_examples using load_dataset
+        加载数据集并转换为模型训练所需的特征：
+            首先加载 MRPC 数据集的指定部分（训练或测试）
+            然后调用 convert_examples_to_features 函数转换为模型输入所需的特征
+    """
+    
+    train_examples = convert_dataset_to_examples(mrpc_datas)
+
+    features = convert_examples_to_features(train_examples, tokenizer, max_seq_length=max_seq_length)
+
+    # Convert to Tensors and build dataset
+    all_input_ids = [f.input_ids for f in features]
+    all_attention_mask = [f.attention_mask for f in features]
+    all_token_type_ids = [f.token_type_ids for f in features]
+    all_lens = [f.input_len for f in features]
+    all_labels = [f.label for f in features]
+    dataset = ((all_input_ids, all_attention_mask, all_token_type_ids, all_lens, all_labels))
+
+    return dataset
+
+def get_dataloader_from_ds(ds, batch_size):
+    train_sampler = SequentialSampler()  # 应用 SequentialSampler 以顺序方式采样数据
+    col_names = ['input_ids', 'attention_mask', 'token_type_ids', 'lens', 'labels']
+    train_dataloader = NumpySlicesDataset(ds, sampler=train_sampler, column_names=col_names)  # 使用 NumpySlicesDataset 包装数据集
+    train_dataloader = train_dataloader.batch(batch_size)  # 根据指定批次大小 进行 批处理
+
+    return train_dataloader
+
+def main():
+    parser = argparse.ArgumentParser(description="Train PEFT model on MRPC dataset")
+    parser.add_argument("--save_dir", default="./saved_models", help="Directory to save model checkpoints")
+    parser.add_argument("--lr", type=float, default=1e-4, help="Learning rate")
+    parser.add_argument("--num_epochs", type=int, default=1, help="Number of training epochs")
+    parser.add_argument("--debug", action="store_true", help="Enable debug mode")
+    parser.add_argument("--batch_size", type=int, default=32, help="Batch size")
+    parser.add_argument("--max_seq_len", type=int, default=256, help="Maximum sequence length")
+    parser.add_argument("--model_name", default="gpt2", help="Base model name")
+    parser.add_argument("--use_lora", action="store_true", help="Use LoRA for fine-tuning")
+    args = parser.parse_args()
+    
+    # 设置运行模式
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="GPU")
+    
+    # 测试后保存模型，所以tokenizer需要在这个作用域可用
+    global tokenizer
+    
+    print("加载MRPC数据集...")
+    # 加载MRPC数据集
+    mrpc_dict = load_dataset("SetFit/mrpc")
+    mrpc_train = mrpc_dict['train']
+    mrpc_valid = mrpc_dict['validation']
+    mrpc_test = mrpc_dict['test']
+    
+    print("加载tokenizer...")
+    # 加载tokenizer
+    tokenizer = GPT2Tokenizer.from_pretrained(args.model_name)
+    # 添加特殊token
+    special_tokens_dict = {
+        "bos_token": "<bos>",
+        "eos_token": "<eos>",
+        "pad_token": "<pad>",
+    }
+    num_added_toks = tokenizer.add_special_tokens(special_tokens_dict)
+    print(f"添加了 {num_added_toks} 个特殊token")
+    
+    print("处理数据集...")
+    # 处理数据集
+    train_ds = load_examples(tokenizer, args.max_seq_len, mrpc_train)
+    valid_ds = load_examples(tokenizer, args.max_seq_len, mrpc_valid)
+    test_ds = load_examples(tokenizer, args.max_seq_len, mrpc_test)
+    
+    # 转换为dataloader
+    train_dataloader = get_dataloader_from_ds(train_ds, args.batch_size)
+    valid_dataloader = get_dataloader_from_ds(valid_ds, args.batch_size)
+    test_dataloader = get_dataloader_from_ds(test_ds, args.batch_size)
+    
+    print("加载模型...")
+    # 加载模型
+    model = GPT2ForSequenceClassification.from_pretrained(args.model_name, num_labels=2, force_download=False, use_safetensors=False)
+    model.config.pad_token_id = tokenizer.pad_token_id
+    model.resize_token_embeddings(model.config.vocab_size + num_added_toks)
+    
+    # 配置PEFT (LoRA)
+    if args.use_lora:
+        print("应用LoRA配置...")
+        peft_config = LoraConfig(task_type=TaskType.SEQ_CLS, inference_mode=False, r=8, lora_alpha=32, lora_dropout=0.1, fan_in_fan_out=True)
+        model = get_peft_model(model, peft_config)
+        print("可训练参数信息:")
+        model.print_trainable_parameters()
+    
+    # 获取参数
+    params = []
+    if hasattr(model, "get_trainable_parameters"):
+        params = model.get_trainable_parameters()
+    elif hasattr(model, "trainable_params"):
+        params = model.trainable_params()
+
+    print(f"找到 {len(params)} 个可训练参数")
+
+    # 转换参数
+    converted_params = []
+    for param in params:
+        if hasattr(param, "data") and hasattr(param, "name"):
+            converted_param = Parameter(param.data, name=param.name)
+        else:
+            converted_param = param
+        converted_params.append(converted_param)
+
+    print(f"转换后的参数数量: {len(converted_params)}")
+
+    # 使用转换后的参数
+    optimizer = AdamWeightDecay(params=converted_params, learning_rate=args.lr)
+    metric = Accuracy()
+
+    # 创建保存目录
+    os.makedirs(args.save_dir, exist_ok=True)
+    ckpt_path = os.path.join(args.save_dir, 'gpt2_mrpc_finetune.ckpt')
+    best_ckpt_path = os.path.join(args.save_dir, 'gpt2_mrpc_finetune_best.ckpt')
+
+    # 预处理数据集，移除不需要的'lens'列
+    def remove_lens_column(dataset):
+        """移除数据集中的'lens'列"""
+        columns_to_keep = [col for col in dataset.get_col_names() if col != 'lens']
+        return dataset.project(columns=columns_to_keep)
+
+    # 打印数据集列名
+    print("训练数据集列名:", train_dataloader.get_col_names())
+    
+    # 处理数据集
+    train_dataset = remove_lens_column(train_dataloader)
+    valid_dataset = remove_lens_column(valid_dataloader)
+    test_dataset = remove_lens_column(test_dataloader)
+
+    print("处理后训练数据集列名:", train_dataset.get_col_names())
+    
+    # 定义前向函数(使用改进的版本)
+    def forward_fn(model, data):
+        """前向计算函数"""
+        # 只在第一个批次时启用详细输出
+        verbose = getattr(forward_fn, 'first_batch', False)
+        if verbose:
+            forward_fn.first_batch = False
+        
+        # 使用我们的改进版本
+        return improved_forward_fn(model, data, verbose=verbose)
+    
+    # 设置第一个批次标志
+    forward_fn.first_batch = True
+
+    # 使用value_and_grad包装前向计算函数
+    grad_fn = value_and_grad(forward_fn, None, optimizer.parameters)
+
+    # 定义训练步骤(使用改进的版本)
+    def train_step(model, optimizer, data):
+        """执行一个训练步骤"""
+        # 首先处理批次数据,确保包含所需字段
+        processed_data = prepare_mrpc_batch(data)
+        # 使用改进的训练步骤
+        return improved_train_step(model, optimizer, processed_data, grad_fn)
+
+    # 定义评估函数
+    def evaluate(model, dataset, metric):
+        """评估模型"""
+        metric.clear()
+        model.set_train(False)
+        
+        for data in dataset.create_dict_iterator():
+            # 处理批次数据,确保包含所需字段
+            processed_data = prepare_mrpc_batch(data)
+            
+            # 确保标签单独传递
+            if 'labels' in processed_data:
+                labels = processed_data.pop('labels')
+            elif 'label' in processed_data:
+                labels = processed_data.pop('label')
+            else:
+                print("警告: 评估数据中没有标签!")
+                continue
+            
+            try:
+                # 前向计算
+                outputs = model(**processed_data)
+                
+                # 更新指标
+                if isinstance(outputs, tuple):
+                    logits = outputs[0]
+                elif hasattr(outputs, "logits"):
+                    # 处理SequenceClassifierOutputWithPast对象
+                    logits = outputs.logits
+                else:
+                    logits = outputs
+                    
+                metric.update(logits, labels)
+            except Exception as e:
+                print(f"评估时出错: {e}")
+                continue
+        
+        # 计算指标
+        result = metric.eval()
+        model.set_train(True)
+        
+        return result
+
+    # 训练循环
+    model.set_train(True)
+    best_acc = 0
+    total_steps = args.num_epochs * train_dataset.get_dataset_size()
+
+    print(f"开始训练: {args.num_epochs}个epoch, 共{total_steps}步")
+
+    # 在开始训练前检查一个批次数据
+    print("检查数据格式...")
+    first_batch = next(train_dataset.create_dict_iterator())
+    examine_batch(first_batch)
+
+    for epoch in range(args.num_epochs):
+        # 训练一个epoch
+        model.set_train(True)
+        train_loss = 0
+        train_steps = 0
+        
+        # 重置第一个批次标志（每个epoch只在第一个批次详细输出）
+        forward_fn.first_batch = epoch == 0
+        
+        progress_bar = tqdm(train_dataset.create_dict_iterator(), total=train_dataset.get_dataset_size())
+        for batch in progress_bar:
+            loss = train_step(model, optimizer, batch)
+            
+            # 错误处理
+            if isinstance(loss, mindspore.Tensor):
+                try:
+                    loss_value = loss.asnumpy()
+                    train_loss += loss_value
+                except:
+                    print(f"警告: 无法转换损失值 {loss}")
+                    train_loss += 1000.0  # 使用一个大的默认值
+            else:
+                # 处理SequenceClassifierOutputWithPast类型的返回值
+                try:
+                    if hasattr(loss, "loss"):
+                        loss_value = loss.loss.asnumpy()
+                        train_loss += loss_value
+                    else:
+                        print(f"警告: 返回对象没有loss属性 {type(loss)}")
+                        train_loss += 1000.0  # 使用一个大的默认值
+                except Exception as e:
+                    print(f"警告: 处理损失值时出错 {e}")
+                    train_loss += 1000.0  # 使用一个大的默认值
+                
+            train_steps += 1
+            
+            # 更新进度条
+            progress_bar.set_description(f"Epoch {epoch+1}/{args.num_epochs}")
+            progress_bar.set_postfix(loss=train_loss/train_steps)
+            
+            # 在debug模式下,训练几个批次后停止
+            if args.debug and train_steps >= 5:
+                print("Debug模式: 提前停止训练")
+                break
+        
+        # 计算该epoch的平均损失
+        avg_loss = train_loss / train_steps
+        print(f"Epoch {epoch+1}/{args.num_epochs}, 平均损失: {avg_loss:.4f}")
+        
+        # 保存模型检查点
+        try:
+            # 使用模型自带的save_pretrained方法替代mindspore.save_checkpoint
+            save_dir = os.path.join(args.save_dir, f"epoch_{epoch+1}")
+            os.makedirs(save_dir, exist_ok=True)
+            model.save_pretrained(save_dir)
+            print(f"模型已保存至 {save_dir}")
+        except Exception as e:
+            print(f"保存模型时出错: {e}")
+        
+        # 评估
+        acc = evaluate(model, valid_dataset, metric)
+        print(f"验证集准确率: {acc:.4f}")
+        
+        # 保存最佳模型
+        if acc > best_acc:
+            best_acc = acc
+            try:
+                # 使用模型自带的save_pretrained方法替代mindspore.save_checkpoint
+                best_save_dir = os.path.join(args.save_dir, "best_model")
+                os.makedirs(best_save_dir, exist_ok=True)
+                model.save_pretrained(best_save_dir)
+                print(f"找到更好的模型! 已保存至 {best_save_dir}")
+            except Exception as e:
+                print(f"保存最佳模型时出错: {e}")
+        
+        # 在debug模式下,只运行一个epoch
+        if args.debug:
+            print("Debug模式: 提前停止训练")
+            break
+
+    # 在测试集上评估
+    print("在测试集上评估...")
+    test_acc = evaluate(model, test_dataset, metric)
+    print(f"测试集准确率: {test_acc:.4f}")
+
+    # 保存最终模型和tokenizer
+    try:
+        final_save_dir = os.path.join(args.save_dir, "final_model")
+        os.makedirs(final_save_dir, exist_ok=True)
+        
+        # 保存模型
+        model.save_pretrained(final_save_dir)
+        
+        # 保存tokenizer
+        tokenizer.save_pretrained(final_save_dir)
+        
+        print(f"最终模型和tokenizer已保存至 {final_save_dir}")
+        
+        # 创建README文件，记录训练信息
+        readme_path = os.path.join(final_save_dir, "README.md")
+        with open(readme_path, "w", encoding="utf-8") as f:
+            f.write(f"# MRPC数据集上的GPT2模型\n\n")
+            f.write(f"## 训练信息\n")
+            f.write(f"- 模型: {args.model_name}\n")
+            f.write(f"- 数据集: MRPC\n")
+            f.write(f"- 训练轮数: {args.num_epochs}\n")
+            f.write(f"- 批次大小: {args.batch_size}\n")
+            f.write(f"- 学习率: {args.lr}\n")
+            f.write(f"- 验证集准确率: {best_acc:.4f}\n")
+            f.write(f"- 测试集准确率: {test_acc:.4f}\n")
+            
+        print("训练信息已记录到README.md")
+    except Exception as e:
+        print(f"保存最终模型时出错: {e}")
+
+if __name__ == "__main__":
+    main() 
\ No newline at end of file