{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from huggingface_hub import HfApi\n",
    "from huggingface_hub import HfApi\n",
    "api = HfApi()\n",
    "# 替换为你的用户名/仓库名\n",
    "repo_id = \"vincentchao/qllava-next\"\n",
    "# # 创建仓库(如果还不存在)\n",
    "api.create_repo(repo_id, repo_type=\"model\", private=False)\n",
    "\n",
    "# 上传整个目录\n",
    "api.upload_folder(\n",
    "    folder_path=\"/common/home/users/w/wzhao/vqclip/qllava_next_newest\",\n",
    "    repo_id=repo_id,\n",
    "    repo_type=\"model\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "classes2id = { 'neutral':0, 'porn':1,'gun':2,'cigarette':3,'alcohol':4, 'knife':5,'blood':6,'insulting_gesture':7}\n",
    "id2class = ['neutral','porn','gun','cigarette','alcohol',\"knife\",'blood','insulting_gesture']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "56dbb817e0a24197b2749a8ff82fe593",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/7 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from transformers import LlavaProcessor\n",
    "import torch\n",
    "import logging\n",
    "from transformers import TrainerCallback\n",
    "\n",
    "import datetime\n",
    "import os\n",
    "#os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"  # 只显示第一个GPU设备\n",
    "import torch  # 或 tensorflow 等\n",
    "import torch.nn as nn\n",
    "import os\n",
    "import torch.distributed as dist\n",
    "from transformers import LlavaProcessor\n",
    "from qllava3_test import LlavaNextForConditionalGeneration\n",
    "# %%\n",
    "model_id = \"/common/public/llava/llama3-llava-next-8b-hf\"\n",
    "processor = LlavaProcessor.from_pretrained(\n",
    "    model_id ,\n",
    ")\n",
    "\n",
    "# 加载模型\n",
    "model = LlavaNextForConditionalGeneration.from_pretrained(\n",
    "    \"/common/home/users/w/wzhao/vqclip/qllava_next_newest\", \n",
    "    #\"/common/home/users/w/wzhao/qllava/vqllava/checkpoint-260\",\n",
    "    # \"/common/home/users/w/wzhao/\"+path + \"/checkpoint-5000\",\n",
    "    torch_dtype=torch.float32,  # 仍然使用float16以节省内存\n",
    ").cuda()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total images to process: 21213\n",
      "Processing folder: /common/home/users/w/wzhao/llava_helper/nsfw_dataset_v1/porn/train (Category: porn)\n",
      "this is q_latent_loss tensor(8.8170, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2042, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.6122, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9030, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.9424, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4856, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(11.4321, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.8580, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.0276, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0069, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.3328, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0832, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.4713, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8678, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.5524, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3881, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.4501, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6125, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.5718, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8930, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 10/21213 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.9837, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2459, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.0015, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2504, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(11.5371, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.8843, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.5971, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1493, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.3564, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5891, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.8574, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.7143, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.5497, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6374, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.1719, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2930, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.7135, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9284, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.7845, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9461, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 20/21213 images\n",
      "this is q_latent_loss tensor(7.4899, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8725, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.1208, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0302, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.0984, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.7746, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.0674, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.7669, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(13.3643, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.3411, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.2910, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5727, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.9972, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2493, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.9211, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.7303, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.6993, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6748, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(6.4199, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.6050, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 30/21213 images\n",
      "this is q_latent_loss tensor(9.1634, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2909, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.9688, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9922, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.6793, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4198, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.4200, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6050, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.4276, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6069, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.8010, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.7003, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.6418, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9105, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.3313, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0828, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.5423, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3856, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.8121, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2030, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 40/21213 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(12.9253, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.2313, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.9912, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2478, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(6.2949, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.5737, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(14.0637, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.5159, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(11.6149, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9037, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.6029, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.1507, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(13.2988, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.3247, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(13.8039, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.4510, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.4613, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1153, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(12.6695, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.1674, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 50/21213 images\n",
      "this is q_latent_loss tensor(9.4005, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3501, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.8682, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.2171, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.6948, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1737, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.5272, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1318, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.7591, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1898, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.1201, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5300, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.5750, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6437, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.1127, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.7782, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.6827, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4207, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(11.4962, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.8740, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 60/21213 images\n",
      "this is q_latent_loss tensor(9.2464, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3116, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.1726, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2932, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.0186, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.7546, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.8717, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4679, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(14.1204, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.5301, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.3599, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0900, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.3845, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.0961, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.8901, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9725, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.0965, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.0241, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.8583, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2146, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 70/21213 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.9591, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9898, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.9921, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2480, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.3415, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8354, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.7903, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4476, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.8003, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2001, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(6.3422, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.5856, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.4356, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1089, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.7969, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4492, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.9973, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.7493, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.3360, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0840, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 80/21213 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.2751, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.8188, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.1294, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0324, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.8635, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9659, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.8551, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9638, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.8822, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4705, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.0291, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0073, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.0501, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2625, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(12.6947, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.1737, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.9940, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4985, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.2904, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8226, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 90/21213 images\n",
      "this is q_latent_loss tensor(9.6648, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4162, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.5542, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3886, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.8552, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4638, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.3666, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0916, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.7352, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4338, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.1629, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0407, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(14.1190, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.5297, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.3341, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.8335, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.9301, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2325, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(6.2145, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.5536, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 100/21213 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.1991, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0498, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.6093, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9023, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.7087, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4272, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.8062, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2016, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.4319, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1080, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.5253, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1313, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.1614, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0404, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.9642, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4910, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.2686, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3171, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.1140, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5285, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 110/21213 images\n",
      "this is q_latent_loss tensor(12.3632, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.0908, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.1329, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0332, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.9918, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2479, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.1709, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.7927, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.1411, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5353, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.4103, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6026, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.9719, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4930, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(6.5009, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.6252, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.8178, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9544, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.2491, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.0623, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 120/21213 images\n",
      "this is q_latent_loss tensor(11.7427, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.9357, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.0834, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0208, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.1786, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.0447, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.7889, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4472, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.3587, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0897, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.1803, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0451, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.9105, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4776, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.8691, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2173, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.4474, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8619, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.4359, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.3590, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 130/21213 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(6.5443, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.6361, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.5185, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.8796, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(15.6529, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.9132, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(11.5756, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.8939, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.6367, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.1592, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.2915, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5729, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.9234, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4809, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(12.2591, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.0648, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.1375, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2844, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.9599, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2400, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 140/21213 images\n",
      "this is q_latent_loss tensor(10.5039, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6260, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.1007, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2752, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(13.8705, device='cuda:0')\n",
      "This is e_latent_loss tensor(3.4676, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.4230, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.6058, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.0464, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5116, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(10.8465, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.7116, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.8604, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4651, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(7.0887, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.7722, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(9.0656, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.2664, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.3511, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0878, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "Processed 150/21213 images\n",
      "this is q_latent_loss tensor(11.1818, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.7954, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(7.8044, device='cuda:0')\n",
      "This is e_latent_loss tensor(1.9511, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.0274, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0068, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(8.4584, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.1146, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "this is q_latent_loss tensor(8.1515, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.0379, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(9.8069, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.4517, device='cuda:0')\n",
      "57\n",
      "[0]\n",
      "this is q_latent_loss tensor(10.3290, device='cuda:0')\n",
      "This is e_latent_loss tensor(2.5823, device='cuda:0')\n",
      "57\n",
      "[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n",
      "Setting `pad_token_id` to `eos_token_id`:None for open-end generation.\n"
     ]
    }
   ],
   "source": [
    "from PIL import Image\n",
    "import torch\n",
    "import os\n",
    "import json\n",
    "import re\n",
    "\n",
    "# 假设 processor 和 model 已经初始化好了\n",
    "# 如果没有，需要添加相关的初始化代码\n",
    "\n",
    "def get_category_from_path(path):\n",
    "    \"\"\"从路径中提取类别\"\"\"\n",
    "    if \"porn\" in path:\n",
    "        return \"porn\"\n",
    "    elif \"blood\" in path:\n",
    "        return \"blood\"\n",
    "    elif \"gun\" in path:\n",
    "        return \"gun\"\n",
    "    elif \"insulting_gesture\" in path:\n",
    "        return \"insulting_gesture\"\n",
    "    elif \"alcohol\" in path:\n",
    "        return \"alcohol\"\n",
    "    elif \"cigarette\" in path:\n",
    "        return \"cigarette\"\n",
    "    elif \"knife\" in path:\n",
    "        return \"knife\"\n",
    "    elif \"real_neutral\" in path:\n",
    "        return \"neutral\"\n",
    "    else:\n",
    "        return \"unknown\"\n",
    "\n",
    "def process_image(image_path, processor, model):\n",
    "    \"\"\"处理图像并获取描述\"\"\"\n",
    "    try:\n",
    "        image = Image.open(image_path)\n",
    "        \n",
    "        conversation = [\n",
    "            {\n",
    "              \"role\": \"user\",\n",
    "              \"content\": [\n",
    "                  {\"type\": \"text\", \"text\": \"Describe this image \"},\n",
    "                  {\"type\": \"image\"},\n",
    "                ],\n",
    "            }, \n",
    "        ]\n",
    "        \n",
    "        prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)\n",
    "        \n",
    "        inputs = processor(images=image, text=prompt, return_tensors=\"pt\").to(model.device)\n",
    "        \n",
    "        with torch.no_grad():\n",
    "            output = model.generate(**inputs, max_new_tokens=10)  # 增加 token 数量以获取更详细的描述\n",
    "        \n",
    "        description = processor.decode(output[0], skip_special_tokens=True)\n",
    "        \n",
    "        # 清理描述文本，移除提示和可能的系统消息\n",
    "        # 这部分可能需要根据实际输出进行调整\n",
    "        description = re.sub(r\"Describe this image\\s*\", \"\", description)\n",
    "        \n",
    "        return description\n",
    "    except ValueError as e:\n",
    "        error_data = e.args[0]  # 这将获取到 [indices, categories[0]]\n",
    "        \n",
    "        # 提取 indices 和 categories[0]\n",
    "        if isinstance(error_data, list) and len(error_data) == 2:\n",
    "            captured_indices = error_data[0]\n",
    "            captured_category = error_data[1]\n",
    "\n",
    "        return f\"Error processing image: {id2class[captured_category]} index {captured_indices}\"\n",
    "\n",
    "def main():\n",
    "    image_folders = [\n",
    "        \"/common/home/users/w/wzhao/llava_helper/nsfw_dataset_v1/porn/train\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/nsfw_dataset_v1/porn/test\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/blood/testing/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/blood/training/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/gun/testing/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/gun/training/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/insulting_gesture/training/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/insulting_gesture/testing/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/alcohol/testing/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/alcohol/training/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/cigarette/testing/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/cigarette/training/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/knife/testing/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/dataset/class/knife/training/jpg\",\n",
    "        \"/common/home/users/w/wzhao/llava_helper/real_neutral\"\n",
    "    ]\n",
    "    \n",
    "    results = []\n",
    "    total_images = 0\n",
    "    \n",
    "    # 计算总图片数量\n",
    "    for folder in image_folders:\n",
    "        if os.path.exists(folder):\n",
    "            for root, _, files in os.walk(folder):\n",
    "                total_images += sum(1 for f in files if f.lower().endswith(('.png', '.jpg', '.jpeg')))\n",
    "    \n",
    "    print(f\"Total images to process: {total_images}\")\n",
    "    \n",
    "    processed_count = 0\n",
    "    \n",
    "    # 处理每个文件夹中的图片\n",
    "    for folder in image_folders:\n",
    "        if not os.path.exists(folder):\n",
    "            print(f\"Folder not found: {folder}\")\n",
    "            continue\n",
    "            \n",
    "        category = get_category_from_path(folder)\n",
    "        print(f\"Processing folder: {folder} (Category: {category})\")\n",
    "        \n",
    "        for root, _, files in os.walk(folder):\n",
    "            for file in files:\n",
    "                if file.lower().endswith(('.png', '.jpg', '.jpeg')):\n",
    "                    image_path = os.path.join(root, file)\n",
    "                    \n",
    "                    # 处理图像\n",
    "                    description = process_image(image_path, processor, model)\n",
    "                    \n",
    "                    # 存储结果\n",
    "                    results.append({\n",
    "                        \"path\": image_path,\n",
    "                        \"response\": description,\n",
    "                        \"category\": category\n",
    "                    })\n",
    "                    \n",
    "                    processed_count += 1\n",
    "                    if processed_count % 10 == 0:\n",
    "                        print(f\"Processed {processed_count}/{total_images} images\")\n",
    "                    \n",
    "                    # 每处理100张图片保存一次中间结果\n",
    "                    if processed_count % 100 == 0:\n",
    "                        with open(f\"/common/home/users/w/wzhao/vqclip/json_results_newest/image_descriptions_checkpoint_{processed_count}.json\", \"w\") as f:\n",
    "                            json.dump(results, f, indent=2)\n",
    "    \n",
    "    # 保存最终结果\n",
    "    with open(\"/common/home/users/w/wzhao/vqclip/json_results_newest/image_descriptions_complete.json\", \"w\") as f:\n",
    "        json.dump(results, f, indent=2)\n",
    "    \n",
    "    print(f\"Processing complete. Results saved to image_descriptions_complete.json\")\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    # 确保已经初始化了processor和model\n",
    "    # 如果没有，需要在这里添加初始化代码\n",
    "    main()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['54', 'porn'], dtype='<U21')"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "np.load(\"/common/home/users/w/wzhao/vqclip/classified_results/1.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**************************************************************\n",
      "torch.Size([1, 3, 336, 336])\n",
      "this is q_latent_loss tensor(0.6994, device='cuda:0')\n",
      "This is e_latent_loss tensor(0.1748, device='cuda:0')\n",
      "34\n",
      "[1]\n",
      "1\n",
      "检测到不适当内容，生成已停止\n",
      "**************************************************************\n"
     ]
    }
   ],
   "source": [
    "from PIL import Image\n",
    "import time\n",
    "image = Image.open('/common/home/users/w/wzhao/qllava/00008.jpg')\n",
    "#image = Image.open('/common/home/users/w/wzhao/qllava/94915.jpg')\n",
    "#image = Image.open(\"\")\n",
    "# Define a chat histiry and use `apply_chat_template` to get correctly formatted prompt\n",
    "# Each value in \"content\" has to be a list of dicts with types (\"text\", \"image\") \n",
    "print(\"**************************************************************\")\n",
    "\n",
    "\n",
    "conversation = [\n",
    "    {\n",
    "      \"role\": \"user\",\n",
    "      \"content\": [\n",
    "          {\"type\": \"text\", \"text\": \"Describe this image \"},\n",
    "          {\"type\": \"image\"},\n",
    "        ],\n",
    "    }, \n",
    "]\n",
    "\n",
    "prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)\n",
    "#print(prompt)\n",
    "\n",
    "inputs = processor(images=image, text=prompt, return_tensors=\"pt\").to(model.device)\n",
    "print(inputs[\"pixel_values\"].shape)\n",
    "\n",
    "# autoregressively complete prompt\n",
    "try:\n",
    "    output = model.generate(**inputs, max_new_tokens=50)\n",
    "    print(processor.decode(output[0], skip_special_tokens=True))\n",
    "except ValueError as e:\n",
    "    # 记录错误信息\n",
    "    \n",
    "    harm_class = int(str(e)))\n",
    "\n",
    "    \n",
    "    # 可以选择继续执行其他代码或者返回一个预设的安全输出\n",
    "    print(\"检测到不适当内容，生成已停止\")\n",
    "print(\"**************************************************************\")\n",
    "#time.sleep(100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LlavaForConditionalGeneration(\n",
       "  (vision_tower): CLIPVisionModel(\n",
       "    (vision_model): CLIPVisionTransformer(\n",
       "      (embeddings): CLIPVisionEmbeddings(\n",
       "        (patch_embedding): Conv2d(3, 1024, kernel_size=(14, 14), stride=(14, 14), bias=False)\n",
       "        (position_embedding): Embedding(577, 1024)\n",
       "      )\n",
       "      (pre_layrnorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
       "      (encoder): CLIPEncoder(\n",
       "        (layers): ModuleList(\n",
       "          (0-23): 24 x CLIPEncoderLayer(\n",
       "            (self_attn): CLIPSdpaAttention(\n",
       "              (k_proj): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (v_proj): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (q_proj): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (out_proj): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "            )\n",
       "            (layer_norm1): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
       "            (mlp): CLIPMLP(\n",
       "              (activation_fn): QuickGELUActivation()\n",
       "              (fc1): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "              (fc2): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            )\n",
       "            (layer_norm2): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (post_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
       "    )\n",
       "  )\n",
       "  (multi_modal_projector): LlavaMultiModalProjector(\n",
       "    (linear_1): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "    (act): GELUActivation()\n",
       "    (linear_2): Linear(in_features=4096, out_features=4096, bias=True)\n",
       "    (vq): VectorQuantizer(\n",
       "      (embedding): Embedding(16000, 4096)\n",
       "    )\n",
       "    (vq_cls): VectorQuantizerCLS(\n",
       "      (embedding): Embedding(128, 4096)\n",
       "    )\n",
       "  )\n",
       "  (language_model): LlamaForCausalLM(\n",
       "    (model): LlamaModel(\n",
       "      (embed_tokens): Embedding(32064, 4096)\n",
       "      (layers): ModuleList(\n",
       "        (0-31): 32 x LlamaDecoderLayer(\n",
       "          (self_attn): LlamaSdpaAttention(\n",
       "            (q_proj): Linear(in_features=4096, out_features=4096, bias=False)\n",
       "            (k_proj): Linear(in_features=4096, out_features=4096, bias=False)\n",
       "            (v_proj): Linear(in_features=4096, out_features=4096, bias=False)\n",
       "            (o_proj): Linear(in_features=4096, out_features=4096, bias=False)\n",
       "            (rotary_emb): LlamaRotaryEmbedding()\n",
       "          )\n",
       "          (mlp): LlamaMLP(\n",
       "            (gate_proj): Linear(in_features=4096, out_features=11008, bias=False)\n",
       "            (up_proj): Linear(in_features=4096, out_features=11008, bias=False)\n",
       "            (down_proj): Linear(in_features=11008, out_features=4096, bias=False)\n",
       "            (act_fn): SiLU()\n",
       "          )\n",
       "          (input_layernorm): LlamaRMSNorm((4096,), eps=1e-05)\n",
       "          (post_attention_layernorm): LlamaRMSNorm((4096,), eps=1e-05)\n",
       "        )\n",
       "      )\n",
       "      (norm): LlamaRMSNorm((4096,), eps=1e-05)\n",
       "      (rotary_emb): LlamaRotaryEmbedding()\n",
       "    )\n",
       "    (lm_head): Linear(in_features=4096, out_features=32064, bias=False)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2025-03-28 01:00:17,048] [WARNING] [real_accelerator.py:174:get_accelerator] Setting accelerator to CPU. If you have GPU or other accelerator, we were unable to detect it.\n",
      "[2025-03-28 01:00:17,049] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cpu (auto detect)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/apps/software/Anaconda3/2024.06-1/compiler_compat/ld: cannot find -laio: No such file or directory\n",
      "collect2: error: ld returned 1 exit status\n"
     ]
    },
    {
     "ename": "AttributeError",
     "evalue": "'list' object has no attribute 'device'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[6], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[43mmodel\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msave_pretrained\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m      2\u001b[0m \u001b[43m    \u001b[49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43m/common/home/users/w/wzhao/vqclip/VQLLMfinal\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\n\u001b[1;32m      3\u001b[0m \u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/.local/lib/python3.12/site-packages/transformers/modeling_utils.py:2971\u001b[0m, in \u001b[0;36mPreTrainedModel.save_pretrained\u001b[0;34m(self, save_directory, is_main_process, state_dict, save_function, push_to_hub, max_shard_size, safe_serialization, variant, token, save_peft_format, **kwargs)\u001b[0m\n\u001b[1;32m   2968\u001b[0m     weights_name \u001b[38;5;241m=\u001b[39m ADAPTER_SAFE_WEIGHTS_NAME \u001b[38;5;28;01mif\u001b[39;00m safe_serialization \u001b[38;5;28;01melse\u001b[39;00m ADAPTER_WEIGHTS_NAME\n\u001b[1;32m   2970\u001b[0m filename_pattern \u001b[38;5;241m=\u001b[39m weights_name\u001b[38;5;241m.\u001b[39mreplace(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m.bin\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{suffix}\u001b[39;00m\u001b[38;5;124m.bin\u001b[39m\u001b[38;5;124m\"\u001b[39m)\u001b[38;5;241m.\u001b[39mreplace(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m.safetensors\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{suffix}\u001b[39;00m\u001b[38;5;124m.safetensors\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[0;32m-> 2971\u001b[0m state_dict_split \u001b[38;5;241m=\u001b[39m \u001b[43msplit_torch_state_dict_into_shards\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m   2972\u001b[0m \u001b[43m    \u001b[49m\u001b[43mstate_dict\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mfilename_pattern\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mfilename_pattern\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mmax_shard_size\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmax_shard_size\u001b[49m\n\u001b[1;32m   2973\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   2974\u001b[0m \u001b[38;5;66;03m# Save index if sharded\u001b[39;00m\n\u001b[1;32m   2975\u001b[0m index \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/.local/lib/python3.12/site-packages/huggingface_hub/serialization/_torch.py:369\u001b[0m, in \u001b[0;36msplit_torch_state_dict_into_shards\u001b[0;34m(state_dict, filename_pattern, max_shard_size)\u001b[0m\n\u001b[1;32m    302\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21msplit_torch_state_dict_into_shards\u001b[39m(\n\u001b[1;32m    303\u001b[0m     state_dict: Dict[\u001b[38;5;28mstr\u001b[39m, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mtorch.Tensor\u001b[39m\u001b[38;5;124m\"\u001b[39m],\n\u001b[1;32m    304\u001b[0m     \u001b[38;5;241m*\u001b[39m,\n\u001b[1;32m    305\u001b[0m     filename_pattern: \u001b[38;5;28mstr\u001b[39m \u001b[38;5;241m=\u001b[39m constants\u001b[38;5;241m.\u001b[39mSAFETENSORS_WEIGHTS_FILE_PATTERN,\n\u001b[1;32m    306\u001b[0m     max_shard_size: Union[\u001b[38;5;28mint\u001b[39m, \u001b[38;5;28mstr\u001b[39m] \u001b[38;5;241m=\u001b[39m MAX_SHARD_SIZE,\n\u001b[1;32m    307\u001b[0m ) \u001b[38;5;241m-\u001b[39m\u001b[38;5;241m>\u001b[39m StateDictSplit:\n\u001b[1;32m    308\u001b[0m \u001b[38;5;250m    \u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m    309\u001b[0m \u001b[38;5;124;03m    Split a model state dictionary in shards so that each shard is smaller than a given size.\u001b[39;00m\n\u001b[1;32m    310\u001b[0m \n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    367\u001b[0m \u001b[38;5;124;03m    ```\u001b[39;00m\n\u001b[1;32m    368\u001b[0m \u001b[38;5;124;03m    \"\"\"\u001b[39;00m\n\u001b[0;32m--> 369\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43msplit_state_dict_into_shards_factory\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    370\u001b[0m \u001b[43m        \u001b[49m\u001b[43mstate_dict\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    371\u001b[0m \u001b[43m        \u001b[49m\u001b[43mmax_shard_size\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmax_shard_size\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    372\u001b[0m \u001b[43m        \u001b[49m\u001b[43mfilename_pattern\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mfilename_pattern\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    373\u001b[0m \u001b[43m        \u001b[49m\u001b[43mget_storage_size\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mget_torch_storage_size\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    374\u001b[0m \u001b[43m        \u001b[49m\u001b[43mget_storage_id\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mget_torch_storage_id\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    375\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/.local/lib/python3.12/site-packages/huggingface_hub/serialization/_base.py:108\u001b[0m, in \u001b[0;36msplit_state_dict_into_shards_factory\u001b[0;34m(state_dict, get_storage_size, filename_pattern, get_storage_id, max_shard_size)\u001b[0m\n\u001b[1;32m    105\u001b[0m     \u001b[38;5;28;01mcontinue\u001b[39;00m\n\u001b[1;32m    107\u001b[0m \u001b[38;5;66;03m# If a `tensor` shares the same underlying storage as another tensor, we put `tensor` in the same `block`\u001b[39;00m\n\u001b[0;32m--> 108\u001b[0m storage_id \u001b[38;5;241m=\u001b[39m \u001b[43mget_storage_id\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtensor\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    109\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m storage_id \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m    110\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m storage_id \u001b[38;5;129;01min\u001b[39;00m storage_id_to_tensors:\n\u001b[1;32m    111\u001b[0m         \u001b[38;5;66;03m# We skip this tensor for now and will reassign to correct shard later\u001b[39;00m\n",
      "File \u001b[0;32m~/.local/lib/python3.12/site-packages/huggingface_hub/serialization/_torch.py:746\u001b[0m, in \u001b[0;36mget_torch_storage_id\u001b[0;34m(tensor)\u001b[0m\n\u001b[1;32m    735\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mget_torch_storage_id\u001b[39m(tensor: \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mtorch.Tensor\u001b[39m\u001b[38;5;124m\"\u001b[39m) \u001b[38;5;241m-\u001b[39m\u001b[38;5;241m>\u001b[39m Optional[Tuple[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mtorch.device\u001b[39m\u001b[38;5;124m\"\u001b[39m, Union[\u001b[38;5;28mint\u001b[39m, Tuple[Any, \u001b[38;5;241m.\u001b[39m\u001b[38;5;241m.\u001b[39m\u001b[38;5;241m.\u001b[39m]], \u001b[38;5;28mint\u001b[39m]]:\n\u001b[1;32m    736\u001b[0m \u001b[38;5;250m    \u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m    737\u001b[0m \u001b[38;5;124;03m    Return unique identifier to a tensor storage.\u001b[39;00m\n\u001b[1;32m    738\u001b[0m \n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    744\u001b[0m \u001b[38;5;124;03m    Taken from https://github.com/huggingface/transformers/blob/1ecf5f7c982d761b4daaa96719d162c324187c64/src/transformers/pytorch_utils.py#L278.\u001b[39;00m\n\u001b[1;32m    745\u001b[0m \u001b[38;5;124;03m    \"\"\"\u001b[39;00m\n\u001b[0;32m--> 746\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[43mtensor\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdevice\u001b[49m\u001b[38;5;241m.\u001b[39mtype \u001b[38;5;241m==\u001b[39m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mmeta\u001b[39m\u001b[38;5;124m\"\u001b[39m:\n\u001b[1;32m    747\u001b[0m         \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[1;32m    748\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n",
      "\u001b[0;31mAttributeError\u001b[0m: 'list' object has no attribute 'device'"
     ]
    }
   ],
   "source": [
    "model.save_pretrained(\n",
    "    \"/common/home/users/w/wzhao/vqclip/VQLLMfinal\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'qllava_new.LlavaForConditionalGeneration'>\n"
     ]
    }
   ],
   "source": [
    "print(type(model))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([56,  1])"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "np.load(\"/common/home/users/w/wzhao/vqclip/classified_results_llama2/1.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "统计结果已保存到字典并导出到：\n",
      "- JSON: /common/home/users/w/wzhao/vqclip/classification_stats.json\n",
      "- Pickle: /common/home/users/w/wzhao/vqclip/classification_stats.pkl\n",
      "\n",
      "基本统计信息:\n",
      "总共发现 8 个不同的类别\n",
      "总共处理了 21213 个元素\n",
      "\n",
      "各类别的统计摘要:\n",
      "类别 0: 包含 5413 个元素，有 43 个不同的元素ID\n",
      "  出现频率最高的元素: ID 16: 615次, ID 60: 609次, ID 14: 566次\n",
      "类别 1: 包含 5554 个元素，有 4 个不同的元素ID\n",
      "  出现频率最高的元素: ID 56: 2067次, ID 7: 1430次, ID 34: 1180次\n",
      "类别 2: 包含 1473 个元素，有 3 个不同的元素ID\n",
      "  出现频率最高的元素: ID 13: 778次, ID 24: 488次, ID 23: 207次\n",
      "类别 3: 包含 2134 个元素，有 4 个不同的元素ID\n",
      "  出现频率最高的元素: ID 46: 840次, ID 33: 641次, ID 39: 351次\n",
      "类别 4: 包含 1416 个元素，有 2 个不同的元素ID\n",
      "  出现频率最高的元素: ID 22: 748次, ID 42: 668次\n",
      "类别 5: 包含 2785 个元素，有 2 个不同的元素ID\n",
      "  出现频率最高的元素: ID 52: 1654次, ID 4: 1131次\n",
      "类别 6: 包含 1723 个元素，有 5 个不同的元素ID\n",
      "  出现频率最高的元素: ID 20: 555次, ID 57: 481次, ID 17: 350次\n",
      "类别 7: 包含 715 个元素，有 1 个不同的元素ID\n",
      "  出现频率最高的元素: ID 40: 715次\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import json\n",
    "from collections import defaultdict\n",
    "import pickle\n",
    "\n",
    "# 定义目录路径\n",
    "directory_path = '/common/home/users/w/wzhao/vqclip/classified_results_llama2'\n",
    "\n",
    "# 创建一个嵌套字典，用于存储每个类别中每个元素ID出现的次数\n",
    "class_element_counts = defaultdict(lambda: defaultdict(int))\n",
    "# 创建一个字典用于存储每个类别的总计数\n",
    "class_total_counts = defaultdict(int)\n",
    "\n",
    "# 遍历目录中的所有.npy文件\n",
    "try:\n",
    "    for filename in os.listdir(directory_path):\n",
    "        if filename.endswith('.npy'):\n",
    "            file_path = os.path.join(directory_path, filename)\n",
    "            \n",
    "            # 加载.npy文件\n",
    "            data = np.load(file_path)\n",
    "            \n",
    "            # 确保数据格式正确\n",
    "            if data.size == 2:\n",
    "                element_id = int(data[0])  # 确保是整数\n",
    "                class_id = int(data[1])    # 确保是整数\n",
    "                \n",
    "                # 增加该元素在对应类别中的计数\n",
    "                class_element_counts[class_id][element_id] += 1\n",
    "                # 增加该类别的总计数\n",
    "                class_total_counts[class_id] += 1\n",
    "            else:\n",
    "                print(f\"警告: 文件 {filename} 的数据格式不符合预期，已跳过\")\n",
    "    \n",
    "    # 将defaultdict转换为普通dict以便序列化\n",
    "    result_dict = {\n",
    "        \"class_totals\": dict(class_total_counts),\n",
    "        \"class_elements\": {\n",
    "            class_id: dict(elements) \n",
    "            for class_id, elements in class_element_counts.items()\n",
    "        }\n",
    "    }\n",
    "    \n",
    "    # 保存结果到JSON文件\n",
    "    output_json_path = os.path.join(os.path.dirname(directory_path), \"classification_stats.json\")\n",
    "    with open(output_json_path, 'w') as f:\n",
    "        json.dump(result_dict, f, indent=2)\n",
    "    \n",
    "    # 也保存为Python pickle格式，这样在后续Python处理中更方便\n",
    "    output_pickle_path = os.path.join(os.path.dirname(directory_path), \"classification_stats.pkl\")\n",
    "    with open(output_pickle_path, 'wb') as f:\n",
    "        pickle.dump(result_dict, f)\n",
    "    \n",
    "    # 打印一些基本统计信息\n",
    "    print(f\"统计结果已保存到字典并导出到：\")\n",
    "    print(f\"- JSON: {output_json_path}\")\n",
    "    print(f\"- Pickle: {output_pickle_path}\")\n",
    "    print(\"\\n基本统计信息:\")\n",
    "    print(f\"总共发现 {len(class_total_counts)} 个不同的类别\")\n",
    "    total_elements = sum(class_total_counts.values())\n",
    "    print(f\"总共处理了 {total_elements} 个元素\")\n",
    "    \n",
    "    # 打印每个类别的样本统计\n",
    "    print(\"\\n各类别的统计摘要:\")\n",
    "    for class_id in sorted(class_total_counts.keys()):\n",
    "        print(f\"类别 {class_id}: 包含 {class_total_counts[class_id]} 个元素，有 {len(class_element_counts[class_id])} 个不同的元素ID\")\n",
    "        \n",
    "        # 获取该类别中出现次数最多的3个元素\n",
    "        sorted_elements = sorted(class_element_counts[class_id].items(), \n",
    "                                key=lambda x: x[1], reverse=True)[:3]\n",
    "        \n",
    "        # 打印这些元素及其出现次数\n",
    "        print(f\"  出现频率最高的元素: \" + \", \".join([f\"ID {e_id}: {count}次\" for e_id, count in sorted_elements]))\n",
    "\n",
    "except Exception as e:\n",
    "    print(f\"发生错误: {e}\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}