ag-im-pred.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "192e29db",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting torchxrayvision\n",
      "  Downloading torchxrayvision-0.0.37-py3-none-any.whl (29.0 MB)\n",
      "     |████████████████████████████████| 29.0 MB 33.1 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: tqdm>=4 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (4.61.1)\n",
      "Requirement already satisfied: pillow>=5.3.0 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (8.3.2)\n",
      "Requirement already satisfied: requests>=1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (2.25.1)\n",
      "Requirement already satisfied: numpy>=1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (1.19.5)\n",
      "Requirement already satisfied: scikit-image>=0.16 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (0.17.2)\n",
      "Requirement already satisfied: torchvision>=0.5 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (0.11.2)\n",
      "Requirement already satisfied: torch>=1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (1.10.1)\n",
      "Requirement already satisfied: pandas>=1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torchxrayvision) (1.1.5)\n",
      "Requirement already satisfied: python-dateutil>=2.7.3 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from pandas>=1->torchxrayvision) (2.8.1)\n",
      "Requirement already satisfied: pytz>=2017.2 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from pandas>=1->torchxrayvision) (2021.1)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from requests>=1->torchxrayvision) (2021.5.30)\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from requests>=1->torchxrayvision) (1.26.8)\n",
      "Requirement already satisfied: chardet<5,>=3.0.2 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from requests>=1->torchxrayvision) (4.0.0)\n",
      "Requirement already satisfied: idna<3,>=2.5 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from requests>=1->torchxrayvision) (2.10)\n",
      "Requirement already satisfied: scipy>=1.0.1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from scikit-image>=0.16->torchxrayvision) (1.5.4)\n",
      "Requirement already satisfied: matplotlib!=3.0.0,>=2.0.0 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from scikit-image>=0.16->torchxrayvision) (3.3.4)\n",
      "Requirement already satisfied: networkx>=2.0 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from scikit-image>=0.16->torchxrayvision) (2.5)\n",
      "Requirement already satisfied: imageio>=2.3.0 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from scikit-image>=0.16->torchxrayvision) (2.9.0)\n",
      "Requirement already satisfied: tifffile>=2019.7.26 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from scikit-image>=0.16->torchxrayvision) (2020.10.1)\n",
      "Requirement already satisfied: PyWavelets>=1.1.1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from scikit-image>=0.16->torchxrayvision) (1.1.1)\n",
      "Requirement already satisfied: dataclasses in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torch>=1->torchxrayvision) (0.8)\n",
      "Requirement already satisfied: typing-extensions in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from torch>=1->torchxrayvision) (3.10.0.0)\n",
      "Requirement already satisfied: cycler>=0.10 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/cycler-0.10.0-py3.6.egg (from matplotlib!=3.0.0,>=2.0.0->scikit-image>=0.16->torchxrayvision) (0.10.0)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image>=0.16->torchxrayvision) (1.3.1)\n",
      "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.3 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image>=0.16->torchxrayvision) (2.4.7)\n",
      "Requirement already satisfied: decorator>=4.3.0 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from networkx>=2.0->scikit-image>=0.16->torchxrayvision) (5.0.9)\n",
      "Requirement already satisfied: six>=1.5 in /home/ec2-user/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages (from python-dateutil>=2.7.3->pandas>=1->torchxrayvision) (1.16.0)\n",
      "Installing collected packages: torchxrayvision\n",
      "Successfully installed torchxrayvision-0.0.37\n"
     ]
    }
   ],
   "source": [
    "# !pip install -U pip \n",
    "# !pip install -U setuptools wheel\n",
    "# !pip install -U autogluon\n",
    "# !pip install -U torchxrayvision"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "b0d5790a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd \n",
    "import numpy as np\n",
    "import os\n",
    "import autogluon.core as ag\n",
    "from autogluon.vision import ImageDataset, ImagePredictor\n",
    "from sklearn.model_selection import train_test_split\n",
    "import torchxrayvision as xrv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "8c2011b3",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Constants:\n",
    "    DATA_DIR = \"./tum-ai-hackathon-model-implementation/data\"\n",
    "    FULL_DIR = \"./full_data\"\n",
    "    SUBSET_NAME=\"Dataset_Chest_X_Ray_Sample.csv\"\n",
    "    FULLSET_NAME=\"Data_Entry_2017.csv\"\n",
    "    IMG_DIR = \"images_sample\"\n",
    "    FULLIMG_DIR = \"images\"\n",
    "    IMG = \"image\"\n",
    "    LABEL = \"label\"\n",
    "    AGE = \"Patient_Age\"\n",
    "    GENDER = \"Patient_Gender\"\n",
    "    \n",
    "const = Constants()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6ad8dfa",
   "metadata": {},
   "source": [
    "## Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "fab38877",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "No Finding            6621\n",
       "Infiltration          1053\n",
       "Atelectasis            510\n",
       "Effusion               453\n",
       "Nodule                 292\n",
       "Pneumothorax           231\n",
       "Mass                   207\n",
       "Consolidation          141\n",
       "Pleural_Thickening     119\n",
       "Cardiomegaly           113\n",
       "Emphysema               88\n",
       "Fibrosis                87\n",
       "Edema                   55\n",
       "Pneumonia               36\n",
       "Hernia                  10\n",
       "Name: Finding_Labels, dtype: int64"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = pd.read_csv(os.path.join(const.DATA_DIR, const.SUBSET_NAME), sep=';')\n",
    "dataset.head(3)\n",
    "dataset['Finding_Labels'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "7a640e0c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image</th>\n",
       "      <th>label</th>\n",
       "      <th>Follow_up</th>\n",
       "      <th>Patient_ID</th>\n",
       "      <th>Patient_Age</th>\n",
       "      <th>Patient_Gender</th>\n",
       "      <th>View_Position</th>\n",
       "      <th>Original_Image_Width</th>\n",
       "      <th>Original_Image_Height</th>\n",
       "      <th>OriginalImagePixelSpacing_x</th>\n",
       "      <th>OriginalImagePixelSpacing_y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>00000005_000.png</td>\n",
       "      <td>No Finding</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>69</td>\n",
       "      <td>F</td>\n",
       "      <td>PA</td>\n",
       "      <td>2048</td>\n",
       "      <td>2500</td>\n",
       "      <td>0,168</td>\n",
       "      <td>0,168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>00000005_001.png</td>\n",
       "      <td>No Finding</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>69</td>\n",
       "      <td>F</td>\n",
       "      <td>AP</td>\n",
       "      <td>2500</td>\n",
       "      <td>2048</td>\n",
       "      <td>0,168</td>\n",
       "      <td>0,168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>00000005_003.png</td>\n",
       "      <td>No Finding</td>\n",
       "      <td>3</td>\n",
       "      <td>5</td>\n",
       "      <td>69</td>\n",
       "      <td>F</td>\n",
       "      <td>PA</td>\n",
       "      <td>2992</td>\n",
       "      <td>2991</td>\n",
       "      <td>0,143</td>\n",
       "      <td>0,143</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              image       label  Follow_up  Patient_ID  Patient_Age  \\\n",
       "0  00000005_000.png  No Finding          0           5           69   \n",
       "1  00000005_001.png  No Finding          1           5           69   \n",
       "2  00000005_003.png  No Finding          3           5           69   \n",
       "\n",
       "  Patient_Gender View_Position  Original_Image_Width  Original_Image_Height  \\\n",
       "0              F            PA                  2048                   2500   \n",
       "1              F            AP                  2500                   2048   \n",
       "2              F            PA                  2992                   2991   \n",
       "\n",
       "  OriginalImagePixelSpacing_x OriginalImagePixelSpacing_y  \n",
       "0                       0,168                       0,168  \n",
       "1                       0,168                       0,168  \n",
       "2                       0,143                       0,143  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.rename(columns={\"Image_Index\": \"image\", \"Finding_Labels\": \"label\"}, inplace=True)\n",
    "dataset.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "5869aa36",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image</th>\n",
       "      <th>label</th>\n",
       "      <th>Follow_up</th>\n",
       "      <th>Patient_ID</th>\n",
       "      <th>Patient_Age</th>\n",
       "      <th>Patient_Gender</th>\n",
       "      <th>View_Position</th>\n",
       "      <th>Original_Image_Width</th>\n",
       "      <th>Original_Image_Height</th>\n",
       "      <th>OriginalImagePixelSpacing_x</th>\n",
       "      <th>OriginalImagePixelSpacing_y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>./data/images_sample/00000005_000.png</td>\n",
       "      <td>No Finding</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>69</td>\n",
       "      <td>F</td>\n",
       "      <td>PA</td>\n",
       "      <td>2048</td>\n",
       "      <td>2500</td>\n",
       "      <td>0,168</td>\n",
       "      <td>0,168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>./data/images_sample/00000005_001.png</td>\n",
       "      <td>No Finding</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>69</td>\n",
       "      <td>F</td>\n",
       "      <td>AP</td>\n",
       "      <td>2500</td>\n",
       "      <td>2048</td>\n",
       "      <td>0,168</td>\n",
       "      <td>0,168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>./data/images_sample/00000005_003.png</td>\n",
       "      <td>No Finding</td>\n",
       "      <td>3</td>\n",
       "      <td>5</td>\n",
       "      <td>69</td>\n",
       "      <td>F</td>\n",
       "      <td>PA</td>\n",
       "      <td>2992</td>\n",
       "      <td>2991</td>\n",
       "      <td>0,143</td>\n",
       "      <td>0,143</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                   image       label  Follow_up  Patient_ID  \\\n",
       "0  ./data/images_sample/00000005_000.png  No Finding          0           5   \n",
       "1  ./data/images_sample/00000005_001.png  No Finding          1           5   \n",
       "2  ./data/images_sample/00000005_003.png  No Finding          3           5   \n",
       "\n",
       "   Patient_Age Patient_Gender View_Position  Original_Image_Width  \\\n",
       "0           69              F            PA                  2048   \n",
       "1           69              F            AP                  2500   \n",
       "2           69              F            PA                  2992   \n",
       "\n",
       "   Original_Image_Height OriginalImagePixelSpacing_x  \\\n",
       "0                   2500                       0,168   \n",
       "1                   2048                       0,168   \n",
       "2                   2991                       0,143   \n",
       "\n",
       "  OriginalImagePixelSpacing_y  \n",
       "0                       0,168  \n",
       "1                       0,168  \n",
       "2                       0,143  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset['image'] = dataset['image'].apply(lambda x: os.path.join(const.DATA_DIR, const.IMG_DIR, x))\n",
    "dataset.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "272403aa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Image Index</th>\n",
       "      <th>Finding Labels</th>\n",
       "      <th>Follow-up #</th>\n",
       "      <th>Patient ID</th>\n",
       "      <th>Patient Age</th>\n",
       "      <th>Patient Gender</th>\n",
       "      <th>View Position</th>\n",
       "      <th>OriginalImage[Width</th>\n",
       "      <th>Height]</th>\n",
       "      <th>OriginalImagePixelSpacing[x</th>\n",
       "      <th>y]</th>\n",
       "      <th>Unnamed: 11</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>00000001_000.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2682</td>\n",
       "      <td>2749</td>\n",
       "      <td>0.143</td>\n",
       "      <td>0.143</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>00000001_001.png</td>\n",
       "      <td>Cardiomegaly|Emphysema</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2894</td>\n",
       "      <td>2729</td>\n",
       "      <td>0.143</td>\n",
       "      <td>0.143</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>00000001_002.png</td>\n",
       "      <td>Cardiomegaly|Effusion</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2500</td>\n",
       "      <td>2048</td>\n",
       "      <td>0.168</td>\n",
       "      <td>0.168</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Image Index          Finding Labels  Follow-up #  Patient ID  \\\n",
       "0  00000001_000.png            Cardiomegaly            0           1   \n",
       "1  00000001_001.png  Cardiomegaly|Emphysema            1           1   \n",
       "2  00000001_002.png   Cardiomegaly|Effusion            2           1   \n",
       "\n",
       "   Patient Age Patient Gender View Position  OriginalImage[Width  Height]  \\\n",
       "0           58              M            PA                 2682     2749   \n",
       "1           58              M            PA                 2894     2729   \n",
       "2           58              M            PA                 2500     2048   \n",
       "\n",
       "   OriginalImagePixelSpacing[x     y]  Unnamed: 11  \n",
       "0                        0.143  0.143          NaN  \n",
       "1                        0.143  0.143          NaN  \n",
       "2                        0.168  0.168          NaN  "
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = pd.read_csv(os.path.join(const.FULL_DIR, const.FULLSET_NAME), sep=',')\n",
    "dataset.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "ad24793a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image</th>\n",
       "      <th>label</th>\n",
       "      <th>Follow-up #</th>\n",
       "      <th>Patient ID</th>\n",
       "      <th>Patient Age</th>\n",
       "      <th>Patient Gender</th>\n",
       "      <th>View Position</th>\n",
       "      <th>OriginalImage[Width</th>\n",
       "      <th>Height]</th>\n",
       "      <th>OriginalImagePixelSpacing[x</th>\n",
       "      <th>y]</th>\n",
       "      <th>Unnamed: 11</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>00000001_000.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2682</td>\n",
       "      <td>2749</td>\n",
       "      <td>0.143</td>\n",
       "      <td>0.143</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>00000001_001.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2894</td>\n",
       "      <td>2729</td>\n",
       "      <td>0.143</td>\n",
       "      <td>0.143</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>00000001_002.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2500</td>\n",
       "      <td>2048</td>\n",
       "      <td>0.168</td>\n",
       "      <td>0.168</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              image         label  Follow-up #  Patient ID  Patient Age  \\\n",
       "0  00000001_000.png  Cardiomegaly            0           1           58   \n",
       "1  00000001_001.png  Cardiomegaly            1           1           58   \n",
       "2  00000001_002.png  Cardiomegaly            2           1           58   \n",
       "\n",
       "  Patient Gender View Position  OriginalImage[Width  Height]  \\\n",
       "0              M            PA                 2682     2749   \n",
       "1              M            PA                 2894     2729   \n",
       "2              M            PA                 2500     2048   \n",
       "\n",
       "   OriginalImagePixelSpacing[x     y]  Unnamed: 11  \n",
       "0                        0.143  0.143          NaN  \n",
       "1                        0.143  0.143          NaN  \n",
       "2                        0.168  0.168          NaN  "
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def split_labels(text):\n",
    "    return text.split('|')[0]\n",
    "\n",
    "dataset.rename(columns={\"Image Index\": \"image\", \"Finding Labels\": \"label\"}, inplace=True)\n",
    "dataset['label'] = dataset['label'].apply(split_labels)\n",
    "dataset.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "845083c6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image</th>\n",
       "      <th>label</th>\n",
       "      <th>Follow-up #</th>\n",
       "      <th>Patient ID</th>\n",
       "      <th>Patient Age</th>\n",
       "      <th>Patient Gender</th>\n",
       "      <th>View Position</th>\n",
       "      <th>OriginalImage[Width</th>\n",
       "      <th>Height]</th>\n",
       "      <th>OriginalImagePixelSpacing[x</th>\n",
       "      <th>y]</th>\n",
       "      <th>Unnamed: 11</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>./full_data/images/00000001_000.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2682</td>\n",
       "      <td>2749</td>\n",
       "      <td>0.143</td>\n",
       "      <td>0.143</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>./full_data/images/00000001_001.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2894</td>\n",
       "      <td>2729</td>\n",
       "      <td>0.143</td>\n",
       "      <td>0.143</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>./full_data/images/00000001_002.png</td>\n",
       "      <td>Cardiomegaly</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>58</td>\n",
       "      <td>M</td>\n",
       "      <td>PA</td>\n",
       "      <td>2500</td>\n",
       "      <td>2048</td>\n",
       "      <td>0.168</td>\n",
       "      <td>0.168</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                 image         label  Follow-up #  Patient ID  \\\n",
       "0  ./full_data/images/00000001_000.png  Cardiomegaly            0           1   \n",
       "1  ./full_data/images/00000001_001.png  Cardiomegaly            1           1   \n",
       "2  ./full_data/images/00000001_002.png  Cardiomegaly            2           1   \n",
       "\n",
       "   Patient Age Patient Gender View Position  OriginalImage[Width  Height]  \\\n",
       "0           58              M            PA                 2682     2749   \n",
       "1           58              M            PA                 2894     2729   \n",
       "2           58              M            PA                 2500     2048   \n",
       "\n",
       "   OriginalImagePixelSpacing[x     y]  Unnamed: 11  \n",
       "0                        0.143  0.143          NaN  \n",
       "1                        0.143  0.143          NaN  \n",
       "2                        0.168  0.168          NaN  "
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset['image'] = dataset['image'].apply(lambda x: os.path.join(const.FULL_DIR, const.FULLIMG_DIR, x))\n",
    "dataset.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "f33e13ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = train_test_split(dataset, test_size=0.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ab6dc6d9",
   "metadata": {},
   "source": [
    "## Training the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6d92aab8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch[31] Batch [9699]\tSpeed: 125.874570 samples/sec\taccuracy=0.540477\tlr=0.000002\n",
      "Epoch[31] Batch [9749]\tSpeed: 127.745702 samples/sec\taccuracy=0.540462\tlr=0.000002\n",
      "Epoch[31] Batch [9799]\tSpeed: 129.187876 samples/sec\taccuracy=0.540485\tlr=0.000002\n",
      "Epoch[31] Batch [9849]\tSpeed: 127.441907 samples/sec\taccuracy=0.540508\tlr=0.000002\n",
      "Epoch[31] Batch [9899]\tSpeed: 130.249808 samples/sec\taccuracy=0.540467\tlr=0.000002\n",
      "Epoch[31] Batch [9949]\tSpeed: 127.696386 samples/sec\taccuracy=0.540653\tlr=0.000002\n",
      "Epoch[31] Batch [9999]\tSpeed: 129.244095 samples/sec\taccuracy=0.540550\tlr=0.000002\n",
      "Epoch[31] Batch [10049]\tSpeed: 128.370454 samples/sec\taccuracy=0.540398\tlr=0.000002\n",
      "Epoch[31] Batch [10099]\tSpeed: 130.263329 samples/sec\taccuracy=0.540507\tlr=0.000002\n",
      "Epoch[31] Batch [10149]\tSpeed: 126.805117 samples/sec\taccuracy=0.540628\tlr=0.000002\n",
      "Epoch[31] Batch [10199]\tSpeed: 129.771738 samples/sec\taccuracy=0.540772\tlr=0.000002\n",
      "Epoch[31] Batch [10249]\tSpeed: 126.737412 samples/sec\taccuracy=0.540805\tlr=0.000002\n",
      "Epoch[31] Batch [10299]\tSpeed: 129.487638 samples/sec\taccuracy=0.540862\tlr=0.000002\n",
      "Epoch[31] Batch [10349]\tSpeed: 128.811865 samples/sec\taccuracy=0.540652\tlr=0.000002\n",
      "Epoch[31] Batch [10399]\tSpeed: 129.311854 samples/sec\taccuracy=0.540769\tlr=0.000002\n",
      "Epoch[31] Batch [10449]\tSpeed: 127.110849 samples/sec\taccuracy=0.540598\tlr=0.000002\n",
      "Epoch[31] Batch [10499]\tSpeed: 129.520906 samples/sec\taccuracy=0.540524\tlr=0.000002\n",
      "Epoch[31] Batch [10549]\tSpeed: 129.770965 samples/sec\taccuracy=0.540616\tlr=0.000002\n",
      "Epoch[31] Batch [10599]\tSpeed: 128.387056 samples/sec\taccuracy=0.540778\tlr=0.000002\n",
      "Epoch[31] Batch [10649]\tSpeed: 129.627153 samples/sec\taccuracy=0.540915\tlr=0.000002\n",
      "Epoch[31] Batch [10699]\tSpeed: 130.332688 samples/sec\taccuracy=0.540759\tlr=0.000002\n",
      "Epoch[31] Batch [10749]\tSpeed: 127.352037 samples/sec\taccuracy=0.540849\tlr=0.000002\n",
      "Epoch[31] Batch [10799]\tSpeed: 128.443150 samples/sec\taccuracy=0.540856\tlr=0.000002\n",
      "Epoch[31] Batch [10849]\tSpeed: 129.365627 samples/sec\taccuracy=0.540737\tlr=0.000002\n",
      "Epoch[31] Batch [10899]\tSpeed: 125.412722 samples/sec\taccuracy=0.540700\tlr=0.000002\n",
      "Epoch[31] Batch [10949]\tSpeed: 126.904870 samples/sec\taccuracy=0.540639\tlr=0.000002\n",
      "Epoch[31] Batch [10999]\tSpeed: 130.934171 samples/sec\taccuracy=0.540625\tlr=0.000002\n",
      "Epoch[31] Batch [11049]\tSpeed: 128.624497 samples/sec\taccuracy=0.540498\tlr=0.000002\n",
      "Epoch[31] Batch [11099]\tSpeed: 130.460020 samples/sec\taccuracy=0.540698\tlr=0.000002\n",
      "Epoch[31] Batch [11149]\tSpeed: 129.788001 samples/sec\taccuracy=0.540549\tlr=0.000002\n",
      "Epoch[31] Batch [11199]\tSpeed: 124.698652 samples/sec\taccuracy=0.540469\tlr=0.000002\n",
      "Epoch[31] Batch [11249]\tSpeed: 129.400829 samples/sec\taccuracy=0.540544\tlr=0.000002\n",
      "Epoch[31] Batch [11299]\tSpeed: 129.547979 samples/sec\taccuracy=0.540608\tlr=0.000002\n",
      "Epoch[31] Batch [11349]\tSpeed: 134.552700 samples/sec\taccuracy=0.540407\tlr=0.000002\n",
      "[Epoch 31] training: accuracy=0.540411\n",
      "[Epoch 31] speed: 127 samples/sec\ttime cost: 710.725149\n",
      "Epoch[32] Batch [6899]\tSpeed: 127.572769 samples/sec\taccuracy=0.543424\tlr=0.000002\n",
      "Epoch[32] Batch [6949]\tSpeed: 119.394262 samples/sec\taccuracy=0.543507\tlr=0.000002\n",
      "Epoch[32] Batch [6999]\tSpeed: 119.123951 samples/sec\taccuracy=0.543286\tlr=0.000002\n",
      "Epoch[32] Batch [7049]\tSpeed: 117.303190 samples/sec\taccuracy=0.543475\tlr=0.000002\n",
      "Epoch[32] Batch [7099]\tSpeed: 120.690826 samples/sec\taccuracy=0.543468\tlr=0.000002\n",
      "Epoch[32] Batch [7149]\tSpeed: 114.000435 samples/sec\taccuracy=0.543252\tlr=0.000002\n",
      "Epoch[32] Batch [7199]\tSpeed: 117.718839 samples/sec\taccuracy=0.543125\tlr=0.000002\n",
      "Epoch[32] Batch [7249]\tSpeed: 129.196640 samples/sec\taccuracy=0.543103\tlr=0.000002\n",
      "Epoch[32] Batch [7299]\tSpeed: 128.445618 samples/sec\taccuracy=0.543014\tlr=0.000002\n",
      "Epoch[32] Batch [7349]\tSpeed: 130.316035 samples/sec\taccuracy=0.543095\tlr=0.000002\n",
      "Epoch[32] Batch [7399]\tSpeed: 126.207803 samples/sec\taccuracy=0.543024\tlr=0.000002\n",
      "Epoch[32] Batch [7449]\tSpeed: 128.698409 samples/sec\taccuracy=0.542886\tlr=0.000002\n",
      "Epoch[32] Batch [7499]\tSpeed: 128.300754 samples/sec\taccuracy=0.542900\tlr=0.000002\n",
      "Epoch[32] Batch [7549]\tSpeed: 125.904477 samples/sec\taccuracy=0.543162\tlr=0.000002\n",
      "Epoch[32] Batch [7599]\tSpeed: 125.859169 samples/sec\taccuracy=0.542878\tlr=0.000002\n",
      "Epoch[32] Batch [7649]\tSpeed: 118.058653 samples/sec\taccuracy=0.542647\tlr=0.000002\n",
      "Epoch[32] Batch [7699]\tSpeed: 129.012487 samples/sec\taccuracy=0.542532\tlr=0.000002\n",
      "Epoch[32] Batch [7749]\tSpeed: 127.672238 samples/sec\taccuracy=0.542645\tlr=0.000002\n",
      "Epoch[32] Batch [7799]\tSpeed: 127.737259 samples/sec\taccuracy=0.542644\tlr=0.000002\n",
      "Epoch[32] Batch [7849]\tSpeed: 128.614952 samples/sec\taccuracy=0.542484\tlr=0.000002\n",
      "Epoch[32] Batch [7899]\tSpeed: 128.983465 samples/sec\taccuracy=0.542421\tlr=0.000002\n",
      "Epoch[32] Batch [7949]\tSpeed: 128.643640 samples/sec\taccuracy=0.542233\tlr=0.000002\n",
      "Epoch[32] Batch [7999]\tSpeed: 125.565398 samples/sec\taccuracy=0.542125\tlr=0.000002\n",
      "Epoch[32] Batch [8049]\tSpeed: 130.282701 samples/sec\taccuracy=0.542189\tlr=0.000002\n",
      "Epoch[32] Batch [8099]\tSpeed: 127.851306 samples/sec\taccuracy=0.542068\tlr=0.000002\n",
      "Epoch[32] Batch [8149]\tSpeed: 127.313864 samples/sec\taccuracy=0.541963\tlr=0.000002\n",
      "Epoch[32] Batch [8199]\tSpeed: 130.276863 samples/sec\taccuracy=0.541951\tlr=0.000002\n",
      "Epoch[32] Batch [8249]\tSpeed: 127.491937 samples/sec\taccuracy=0.541788\tlr=0.000002\n",
      "Epoch[32] Batch [8299]\tSpeed: 128.708184 samples/sec\taccuracy=0.541657\tlr=0.000002\n",
      "Epoch[32] Batch [8349]\tSpeed: 128.085868 samples/sec\taccuracy=0.541392\tlr=0.000002\n",
      "Epoch[32] Batch [8399]\tSpeed: 127.433563 samples/sec\taccuracy=0.541235\tlr=0.000002\n",
      "Epoch[32] Batch [8449]\tSpeed: 127.089299 samples/sec\taccuracy=0.541391\tlr=0.000002\n",
      "Epoch[32] Batch [8499]\tSpeed: 127.129139 samples/sec\taccuracy=0.541118\tlr=0.000002\n",
      "Epoch[32] Batch [8549]\tSpeed: 125.375187 samples/sec\taccuracy=0.541009\tlr=0.000002\n",
      "Epoch[32] Batch [8599]\tSpeed: 125.285129 samples/sec\taccuracy=0.541134\tlr=0.000002\n",
      "Epoch[32] Batch [8649]\tSpeed: 129.677090 samples/sec\taccuracy=0.541040\tlr=0.000002\n",
      "Epoch[32] Batch [8699]\tSpeed: 128.597808 samples/sec\taccuracy=0.540948\tlr=0.000002\n",
      "Epoch[32] Batch [8749]\tSpeed: 128.229120 samples/sec\taccuracy=0.541086\tlr=0.000002\n",
      "Epoch[32] Batch [8799]\tSpeed: 129.496903 samples/sec\taccuracy=0.541136\tlr=0.000002\n",
      "Epoch[32] Batch [8849]\tSpeed: 130.445109 samples/sec\taccuracy=0.541017\tlr=0.000002\n",
      "Epoch[32] Batch [8899]\tSpeed: 129.661796 samples/sec\taccuracy=0.541390\tlr=0.000002\n",
      "Epoch[32] Batch [8949]\tSpeed: 129.964308 samples/sec\taccuracy=0.541453\tlr=0.000002\n",
      "Epoch[32] Batch [8999]\tSpeed: 128.713871 samples/sec\taccuracy=0.541444\tlr=0.000002\n",
      "Epoch[32] Batch [9049]\tSpeed: 128.060820 samples/sec\taccuracy=0.541519\tlr=0.000002\n",
      "Epoch[32] Batch [9099]\tSpeed: 127.505211 samples/sec\taccuracy=0.541607\tlr=0.000002\n",
      "Epoch[32] Batch [9149]\tSpeed: 128.127813 samples/sec\taccuracy=0.541667\tlr=0.000002\n",
      "Epoch[32] Batch [9199]\tSpeed: 131.199643 samples/sec\taccuracy=0.541916\tlr=0.000002\n",
      "Epoch[32] Batch [9249]\tSpeed: 126.913558 samples/sec\taccuracy=0.541919\tlr=0.000002\n",
      "Epoch[32] Batch [9299]\tSpeed: 126.165920 samples/sec\taccuracy=0.542043\tlr=0.000002\n",
      "Epoch[32] Batch [9349]\tSpeed: 127.339462 samples/sec\taccuracy=0.542179\tlr=0.000002\n",
      "Epoch[32] Batch [9399]\tSpeed: 127.082416 samples/sec\taccuracy=0.541782\tlr=0.000002\n",
      "Epoch[32] Batch [9449]\tSpeed: 125.989382 samples/sec\taccuracy=0.541772\tlr=0.000002\n",
      "Epoch[32] Batch [9499]\tSpeed: 126.173397 samples/sec\taccuracy=0.541684\tlr=0.000002\n",
      "Epoch[32] Batch [9549]\tSpeed: 126.347502 samples/sec\taccuracy=0.541819\tlr=0.000002\n",
      "Epoch[32] Batch [9599]\tSpeed: 127.086662 samples/sec\taccuracy=0.541576\tlr=0.000002\n",
      "Epoch[32] Batch [9649]\tSpeed: 126.519388 samples/sec\taccuracy=0.541645\tlr=0.000002\n",
      "Epoch[32] Batch [9699]\tSpeed: 128.301470 samples/sec\taccuracy=0.541495\tlr=0.000002\n",
      "Epoch[32] Batch [9749]\tSpeed: 128.636903 samples/sec\taccuracy=0.541590\tlr=0.000002\n",
      "Epoch[32] Batch [9799]\tSpeed: 127.813446 samples/sec\taccuracy=0.541556\tlr=0.000002\n",
      "Epoch[32] Batch [9849]\tSpeed: 126.583784 samples/sec\taccuracy=0.541586\tlr=0.000002\n",
      "Epoch[32] Batch [9899]\tSpeed: 127.658764 samples/sec\taccuracy=0.541705\tlr=0.000002\n",
      "Epoch[32] Batch [9949]\tSpeed: 127.498535 samples/sec\taccuracy=0.541771\tlr=0.000002\n",
      "Epoch[32] Batch [9999]\tSpeed: 129.129946 samples/sec\taccuracy=0.541687\tlr=0.000002\n",
      "Epoch[32] Batch [10049]\tSpeed: 127.120740 samples/sec\taccuracy=0.541791\tlr=0.000002\n",
      "Epoch[32] Batch [10099]\tSpeed: 127.406931 samples/sec\taccuracy=0.541931\tlr=0.000002\n",
      "Epoch[32] Batch [10149]\tSpeed: 127.345213 samples/sec\taccuracy=0.541860\tlr=0.000002\n",
      "Epoch[32] Batch [10199]\tSpeed: 127.464002 samples/sec\taccuracy=0.541752\tlr=0.000002\n",
      "Epoch[32] Batch [10249]\tSpeed: 127.505104 samples/sec\taccuracy=0.541683\tlr=0.000002\n",
      "Epoch[32] Batch [10299]\tSpeed: 128.669183 samples/sec\taccuracy=0.541650\tlr=0.000002\n",
      "Epoch[32] Batch [10349]\tSpeed: 130.369431 samples/sec\taccuracy=0.541812\tlr=0.000002\n",
      "Epoch[32] Batch [10399]\tSpeed: 127.164262 samples/sec\taccuracy=0.541815\tlr=0.000002\n",
      "Epoch[32] Batch [10449]\tSpeed: 128.278662 samples/sec\taccuracy=0.541567\tlr=0.000002\n",
      "Epoch[32] Batch [10499]\tSpeed: 122.261359 samples/sec\taccuracy=0.541321\tlr=0.000002\n",
      "Epoch[32] Batch [10549]\tSpeed: 122.724715 samples/sec\taccuracy=0.541422\tlr=0.000002\n",
      "Epoch[32] Batch [10599]\tSpeed: 128.800938 samples/sec\taccuracy=0.541509\tlr=0.000002\n",
      "Epoch[32] Batch [10649]\tSpeed: 125.362727 samples/sec\taccuracy=0.541350\tlr=0.000002\n",
      "Epoch[32] Batch [10699]\tSpeed: 129.355533 samples/sec\taccuracy=0.541355\tlr=0.000002\n",
      "Epoch[33] Batch [9749]\tSpeed: 126.857276 samples/sec\taccuracy=0.542372\tlr=0.000002\n",
      "Epoch[33] Batch [9799]\tSpeed: 128.130700 samples/sec\taccuracy=0.542334\tlr=0.000002\n",
      "Epoch[33] Batch [9849]\tSpeed: 129.194541 samples/sec\taccuracy=0.542360\tlr=0.000002\n",
      "Epoch[33] Batch [9899]\tSpeed: 127.215936 samples/sec\taccuracy=0.542311\tlr=0.000002\n",
      "Epoch[33] Batch [9949]\tSpeed: 128.030329 samples/sec\taccuracy=0.542085\tlr=0.000002\n",
      "Epoch[33] Batch [9999]\tSpeed: 127.085285 samples/sec\taccuracy=0.542300\tlr=0.000002\n",
      "Epoch[33] Batch [10049]\tSpeed: 128.688824 samples/sec\taccuracy=0.542326\tlr=0.000002\n",
      "Epoch[33] Batch [10099]\tSpeed: 128.213363 samples/sec\taccuracy=0.542401\tlr=0.000002\n",
      "Epoch[33] Batch [10149]\tSpeed: 127.129852 samples/sec\taccuracy=0.542414\tlr=0.000002\n",
      "Epoch[33] Batch [10199]\tSpeed: 129.886431 samples/sec\taccuracy=0.542353\tlr=0.000002\n",
      "Epoch[33] Batch [10249]\tSpeed: 128.724912 samples/sec\taccuracy=0.542427\tlr=0.000002\n",
      "Epoch[33] Batch [10299]\tSpeed: 127.897485 samples/sec\taccuracy=0.542597\tlr=0.000002\n",
      "Epoch[33] Batch [10349]\tSpeed: 129.218591 samples/sec\taccuracy=0.542283\tlr=0.000002\n",
      "Epoch[33] Batch [10399]\tSpeed: 127.238329 samples/sec\taccuracy=0.542200\tlr=0.000002\n",
      "Epoch[33] Batch [10449]\tSpeed: 125.576112 samples/sec\taccuracy=0.542057\tlr=0.000002\n",
      "Epoch[33] Batch [10499]\tSpeed: 128.461836 samples/sec\taccuracy=0.542000\tlr=0.000002\n",
      "Epoch[33] Batch [10549]\tSpeed: 128.088088 samples/sec\taccuracy=0.541813\tlr=0.000002\n",
      "Epoch[33] Batch [10599]\tSpeed: 126.770355 samples/sec\taccuracy=0.541651\tlr=0.000002\n",
      "Epoch[33] Batch [10649]\tSpeed: 128.191752 samples/sec\taccuracy=0.541526\tlr=0.000002\n",
      "Epoch[33] Batch [10699]\tSpeed: 126.845891 samples/sec\taccuracy=0.541530\tlr=0.000002\n",
      "Epoch[33] Batch [10749]\tSpeed: 128.150391 samples/sec\taccuracy=0.541326\tlr=0.000002\n",
      "Epoch[33] Batch [10799]\tSpeed: 125.702001 samples/sec\taccuracy=0.541377\tlr=0.000002\n",
      "Epoch[33] Batch [10849]\tSpeed: 128.407229 samples/sec\taccuracy=0.541509\tlr=0.000002\n",
      "Epoch[33] Batch [10899]\tSpeed: 129.192850 samples/sec\taccuracy=0.541502\tlr=0.000002\n",
      "Epoch[33] Batch [10949]\tSpeed: 128.766397 samples/sec\taccuracy=0.541301\tlr=0.000002\n",
      "Epoch[33] Batch [10999]\tSpeed: 131.327720 samples/sec\taccuracy=0.541182\tlr=0.000002\n",
      "Epoch[33] Batch [11049]\tSpeed: 126.716650 samples/sec\taccuracy=0.541029\tlr=0.000002\n",
      "Epoch[33] Batch [11099]\tSpeed: 127.085141 samples/sec\taccuracy=0.540935\tlr=0.000002\n",
      "Epoch[33] Batch [11149]\tSpeed: 127.828307 samples/sec\taccuracy=0.541043\tlr=0.000002\n",
      "Epoch[33] Batch [11199]\tSpeed: 129.875642 samples/sec\taccuracy=0.541116\tlr=0.000002\n",
      "Epoch[33] Batch [11249]\tSpeed: 127.519545 samples/sec\taccuracy=0.541211\tlr=0.000002\n",
      "Epoch[33] Batch [11299]\tSpeed: 130.429452 samples/sec\taccuracy=0.541239\tlr=0.000002\n",
      "Epoch[33] Batch [11349]\tSpeed: 134.930691 samples/sec\taccuracy=0.541300\tlr=0.000002\n",
      "[Epoch 33] training: accuracy=0.541314\n",
      "[Epoch 33] speed: 127 samples/sec\ttime cost: 710.880873\n",
      "[Epoch 33] validation: top1=0.542860 top5=0.878109\n",
      "Epoch[34] Batch [49]\tSpeed: 105.628157 samples/sec\taccuracy=0.540000\tlr=0.000002\n",
      "Epoch[34] Batch [99]\tSpeed: 127.560101 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[34] Batch [149]\tSpeed: 122.439088 samples/sec\taccuracy=0.550000\tlr=0.000002\n",
      "Epoch[34] Batch [199]\tSpeed: 120.335351 samples/sec\taccuracy=0.541875\tlr=0.000002\n",
      "Epoch[34] Batch [249]\tSpeed: 127.075178 samples/sec\taccuracy=0.533500\tlr=0.000002\n",
      "Epoch[34] Batch [299]\tSpeed: 128.423398 samples/sec\taccuracy=0.537083\tlr=0.000002\n",
      "Epoch[34] Batch [349]\tSpeed: 126.990159 samples/sec\taccuracy=0.537500\tlr=0.000002\n",
      "Epoch[34] Batch [399]\tSpeed: 126.580718 samples/sec\taccuracy=0.533750\tlr=0.000002\n",
      "Epoch[34] Batch [449]\tSpeed: 127.898596 samples/sec\taccuracy=0.533889\tlr=0.000002\n",
      "Epoch[34] Batch [499]\tSpeed: 129.184404 samples/sec\taccuracy=0.537000\tlr=0.000002\n",
      "Epoch[34] Batch [549]\tSpeed: 127.438538 samples/sec\taccuracy=0.538864\tlr=0.000002\n",
      "Epoch[34] Batch [599]\tSpeed: 128.777111 samples/sec\taccuracy=0.538750\tlr=0.000002\n",
      "Epoch[34] Batch [649]\tSpeed: 130.235835 samples/sec\taccuracy=0.539808\tlr=0.000002\n",
      "Epoch[34] Batch [699]\tSpeed: 127.304097 samples/sec\taccuracy=0.541786\tlr=0.000002\n",
      "Epoch[34] Batch [749]\tSpeed: 126.049877 samples/sec\taccuracy=0.540333\tlr=0.000002\n",
      "Epoch[34] Batch [799]\tSpeed: 128.097379 samples/sec\taccuracy=0.542969\tlr=0.000002\n",
      "Epoch[34] Batch [849]\tSpeed: 128.842670 samples/sec\taccuracy=0.542206\tlr=0.000002\n",
      "Epoch[34] Batch [899]\tSpeed: 129.172816 samples/sec\taccuracy=0.543194\tlr=0.000002\n",
      "Epoch[34] Batch [949]\tSpeed: 130.235501 samples/sec\taccuracy=0.541842\tlr=0.000002\n",
      "Epoch[34] Batch [999]\tSpeed: 128.071153 samples/sec\taccuracy=0.539750\tlr=0.000002\n",
      "Epoch[34] Batch [1049]\tSpeed: 126.459222 samples/sec\taccuracy=0.539643\tlr=0.000002\n",
      "Epoch[34] Batch [1099]\tSpeed: 128.369933 samples/sec\taccuracy=0.540114\tlr=0.000002\n",
      "Epoch[34] Batch [1149]\tSpeed: 127.397740 samples/sec\taccuracy=0.542065\tlr=0.000002\n",
      "Epoch[34] Batch [1199]\tSpeed: 130.070427 samples/sec\taccuracy=0.542604\tlr=0.000002\n",
      "Epoch[34] Batch [1249]\tSpeed: 126.968440 samples/sec\taccuracy=0.542900\tlr=0.000002\n",
      "Epoch[34] Batch [1299]\tSpeed: 127.348644 samples/sec\taccuracy=0.543942\tlr=0.000002\n",
      "Epoch[34] Batch [1349]\tSpeed: 131.429581 samples/sec\taccuracy=0.543889\tlr=0.000002\n",
      "Epoch[34] Batch [1399]\tSpeed: 127.383907 samples/sec\taccuracy=0.543661\tlr=0.000002\n",
      "Epoch[34] Batch [1449]\tSpeed: 127.582790 samples/sec\taccuracy=0.544052\tlr=0.000002\n",
      "Epoch[34] Batch [1499]\tSpeed: 129.062507 samples/sec\taccuracy=0.543500\tlr=0.000002\n",
      "Epoch[34] Batch [1549]\tSpeed: 130.781634 samples/sec\taccuracy=0.543065\tlr=0.000002\n",
      "Epoch[34] Batch [1599]\tSpeed: 128.311283 samples/sec\taccuracy=0.542969\tlr=0.000002\n",
      "Epoch[34] Batch [1649]\tSpeed: 129.191576 samples/sec\taccuracy=0.542803\tlr=0.000002\n",
      "Epoch[34] Batch [1699]\tSpeed: 125.383057 samples/sec\taccuracy=0.543676\tlr=0.000002\n",
      "Epoch[34] Batch [1749]\tSpeed: 129.042633 samples/sec\taccuracy=0.545429\tlr=0.000002\n",
      "Epoch[34] Batch [1799]\tSpeed: 127.599664 samples/sec\taccuracy=0.545278\tlr=0.000002\n",
      "Epoch[34] Batch [1849]\tSpeed: 127.841564 samples/sec\taccuracy=0.545270\tlr=0.000002\n",
      "Epoch[34] Batch [1899]\tSpeed: 128.234374 samples/sec\taccuracy=0.545987\tlr=0.000002\n",
      "Epoch[34] Batch [1949]\tSpeed: 127.781137 samples/sec\taccuracy=0.545897\tlr=0.000002\n",
      "Epoch[34] Batch [1999]\tSpeed: 131.525083 samples/sec\taccuracy=0.545562\tlr=0.000002\n",
      "Epoch[34] Batch [2049]\tSpeed: 127.409824 samples/sec\taccuracy=0.545000\tlr=0.000002\n",
      "Epoch[34] Batch [2099]\tSpeed: 127.914608 samples/sec\taccuracy=0.545952\tlr=0.000002\n",
      "Epoch[34] Batch [2149]\tSpeed: 129.655103 samples/sec\taccuracy=0.545058\tlr=0.000002\n",
      "Epoch[34] Batch [2199]\tSpeed: 129.186831 samples/sec\taccuracy=0.544716\tlr=0.000002\n",
      "Epoch[34] Batch [2249]\tSpeed: 130.589858 samples/sec\taccuracy=0.545278\tlr=0.000002\n",
      "Epoch[34] Batch [2299]\tSpeed: 126.109038 samples/sec\taccuracy=0.544620\tlr=0.000002\n",
      "Epoch[34] Batch [2349]\tSpeed: 127.452837 samples/sec\taccuracy=0.544840\tlr=0.000002\n",
      "Epoch[34] Batch [2399]\tSpeed: 126.852710 samples/sec\taccuracy=0.544375\tlr=0.000002\n",
      "Epoch[34] Batch [2449]\tSpeed: 129.550280 samples/sec\taccuracy=0.544592\tlr=0.000002\n",
      "Epoch[34] Batch [2499]\tSpeed: 127.405731 samples/sec\taccuracy=0.544800\tlr=0.000002\n",
      "Epoch[34] Batch [2549]\tSpeed: 127.415504 samples/sec\taccuracy=0.545392\tlr=0.000002\n",
      "Epoch[34] Batch [2599]\tSpeed: 129.717325 samples/sec\taccuracy=0.545481\tlr=0.000002\n",
      "Epoch[34] Batch [2649]\tSpeed: 126.980336 samples/sec\taccuracy=0.545425\tlr=0.000002\n",
      "Epoch[34] Batch [2699]\tSpeed: 128.630779 samples/sec\taccuracy=0.545000\tlr=0.000002\n",
      "Epoch[34] Batch [2749]\tSpeed: 128.853604 samples/sec\taccuracy=0.545273\tlr=0.000002\n",
      "Epoch[34] Batch [2799]\tSpeed: 127.740644 samples/sec\taccuracy=0.544643\tlr=0.000002\n",
      "Epoch[34] Batch [2849]\tSpeed: 129.622847 samples/sec\taccuracy=0.543816\tlr=0.000002\n",
      "Epoch[34] Batch [2899]\tSpeed: 130.140701 samples/sec\taccuracy=0.543405\tlr=0.000002\n",
      "Epoch[34] Batch [2949]\tSpeed: 126.588273 samples/sec\taccuracy=0.543771\tlr=0.000002\n",
      "Epoch[34] Batch [2999]\tSpeed: 126.426660 samples/sec\taccuracy=0.543708\tlr=0.000002\n",
      "Epoch[34] Batch [3049]\tSpeed: 121.563238 samples/sec\taccuracy=0.543770\tlr=0.000002\n",
      "Epoch[34] Batch [3099]\tSpeed: 122.549380 samples/sec\taccuracy=0.543831\tlr=0.000002\n",
      "Epoch[34] Batch [3149]\tSpeed: 128.269767 samples/sec\taccuracy=0.544048\tlr=0.000002\n",
      "Epoch[34] Batch [3199]\tSpeed: 129.882761 samples/sec\taccuracy=0.544414\tlr=0.000002\n",
      "Epoch[34] Batch [3249]\tSpeed: 129.198700 samples/sec\taccuracy=0.544538\tlr=0.000002\n",
      "Epoch[34] Batch [3299]\tSpeed: 128.954536 samples/sec\taccuracy=0.544545\tlr=0.000002\n",
      "Epoch[34] Batch [3349]\tSpeed: 128.673900 samples/sec\taccuracy=0.544590\tlr=0.000002\n",
      "Epoch[34] Batch [3399]\tSpeed: 129.364131 samples/sec\taccuracy=0.544743\tlr=0.000002\n",
      "Epoch[34] Batch [3449]\tSpeed: 126.359263 samples/sec\taccuracy=0.544964\tlr=0.000002\n",
      "Epoch[34] Batch [3499]\tSpeed: 126.272501 samples/sec\taccuracy=0.544679\tlr=0.000002\n",
      "Epoch[34] Batch [3549]\tSpeed: 126.323528 samples/sec\taccuracy=0.544507\tlr=0.000002\n",
      "Epoch[34] Batch [3599]\tSpeed: 127.979046 samples/sec\taccuracy=0.544583\tlr=0.000002\n",
      "Epoch[34] Batch [3649]\tSpeed: 126.673357 samples/sec\taccuracy=0.544281\tlr=0.000002\n",
      "Epoch[34] Batch [3699]\tSpeed: 127.830722 samples/sec\taccuracy=0.543885\tlr=0.000002\n",
      "Epoch[34] Batch [3749]\tSpeed: 129.673081 samples/sec\taccuracy=0.543033\tlr=0.000002\n",
      "Epoch[34] Batch [3799]\tSpeed: 127.562031 samples/sec\taccuracy=0.542632\tlr=0.000002\n",
      "Epoch[34] Batch [3849]\tSpeed: 125.486592 samples/sec\taccuracy=0.543019\tlr=0.000002\n",
      "Epoch[34] Batch [3899]\tSpeed: 128.132804 samples/sec\taccuracy=0.543301\tlr=0.000002\n",
      "Epoch[34] Batch [3949]\tSpeed: 127.678660 samples/sec\taccuracy=0.543449\tlr=0.000002\n",
      "Epoch[34] Batch [3999]\tSpeed: 129.189378 samples/sec\taccuracy=0.543594\tlr=0.000002\n",
      "Epoch[34] Batch [4049]\tSpeed: 127.543208 samples/sec\taccuracy=0.543488\tlr=0.000002\n",
      "Epoch[34] Batch [4099]\tSpeed: 129.495354 samples/sec\taccuracy=0.543079\tlr=0.000002\n",
      "Epoch[34] Batch [4149]\tSpeed: 128.182937 samples/sec\taccuracy=0.542892\tlr=0.000002\n",
      "Epoch[34] Batch [4199]\tSpeed: 129.254629 samples/sec\taccuracy=0.543452\tlr=0.000002\n",
      "Epoch[34] Batch [4249]\tSpeed: 128.576708 samples/sec\taccuracy=0.544059\tlr=0.000002\n",
      "Epoch[34] Batch [4299]\tSpeed: 126.748490 samples/sec\taccuracy=0.543750\tlr=0.000002\n",
      "Epoch[34] Batch [4349]\tSpeed: 127.346479 samples/sec\taccuracy=0.543621\tlr=0.000002\n",
      "Epoch[34] Batch [4399]\tSpeed: 127.747793 samples/sec\taccuracy=0.543438\tlr=0.000002\n",
      "Epoch[34] Batch [4449]\tSpeed: 128.037863 samples/sec\taccuracy=0.543624\tlr=0.000002\n",
      "Epoch[34] Batch [4499]\tSpeed: 128.958184 samples/sec\taccuracy=0.543417\tlr=0.000002\n",
      "Epoch[34] Batch [4549]\tSpeed: 128.909840 samples/sec\taccuracy=0.543681\tlr=0.000002\n",
      "Epoch[34] Batch [4599]\tSpeed: 127.738066 samples/sec\taccuracy=0.543832\tlr=0.000002\n",
      "Epoch[34] Batch [4649]\tSpeed: 129.813971 samples/sec\taccuracy=0.543790\tlr=0.000002\n",
      "Epoch[34] Batch [4699]\tSpeed: 126.047548 samples/sec\taccuracy=0.543963\tlr=0.000002\n",
      "Epoch[34] Batch [4749]\tSpeed: 126.097673 samples/sec\taccuracy=0.544053\tlr=0.000002\n",
      "Epoch[34] Batch [4799]\tSpeed: 127.656938 samples/sec\taccuracy=0.544141\tlr=0.000002\n",
      "Epoch[34] Batch [4849]\tSpeed: 127.825999 samples/sec\taccuracy=0.544175\tlr=0.000002\n",
      "Epoch[34] Batch [4899]\tSpeed: 127.802298 samples/sec\taccuracy=0.544311\tlr=0.000002\n",
      "Epoch[34] Batch [4949]\tSpeed: 126.444163 samples/sec\taccuracy=0.544268\tlr=0.000002\n",
      "Epoch[34] Batch [4999]\tSpeed: 127.157612 samples/sec\taccuracy=0.544525\tlr=0.000002\n",
      "Epoch[34] Batch [5049]\tSpeed: 127.674482 samples/sec\taccuracy=0.544975\tlr=0.000002\n",
      "Epoch[34] Batch [5099]\tSpeed: 127.738669 samples/sec\taccuracy=0.544975\tlr=0.000002\n",
      "Epoch[34] Batch [5149]\tSpeed: 126.898554 samples/sec\taccuracy=0.544903\tlr=0.000002\n",
      "Epoch[34] Batch [5199]\tSpeed: 126.790742 samples/sec\taccuracy=0.544904\tlr=0.000002\n",
      "Epoch[34] Batch [5249]\tSpeed: 128.502030 samples/sec\taccuracy=0.545000\tlr=0.000002\n",
      "Epoch[34] Batch [5299]\tSpeed: 127.327565 samples/sec\taccuracy=0.544646\tlr=0.000002\n",
      "Epoch[34] Batch [5349]\tSpeed: 126.957554 samples/sec\taccuracy=0.544322\tlr=0.000002\n",
      "Epoch[34] Batch [5399]\tSpeed: 126.992582 samples/sec\taccuracy=0.544167\tlr=0.000002\n",
      "Epoch[34] Batch [5449]\tSpeed: 128.233609 samples/sec\taccuracy=0.543945\tlr=0.000002\n",
      "Epoch[34] Batch [5499]\tSpeed: 128.440770 samples/sec\taccuracy=0.543955\tlr=0.000002\n",
      "Epoch[34] Batch [5549]\tSpeed: 127.281922 samples/sec\taccuracy=0.543874\tlr=0.000002\n",
      "Epoch[34] Batch [5599]\tSpeed: 128.776933 samples/sec\taccuracy=0.543638\tlr=0.000002\n",
      "Epoch[34] Batch [5649]\tSpeed: 128.331481 samples/sec\taccuracy=0.543717\tlr=0.000002\n",
      "Epoch[34] Batch [5699]\tSpeed: 130.283439 samples/sec\taccuracy=0.543882\tlr=0.000002\n",
      "Epoch[34] Batch [5749]\tSpeed: 125.658504 samples/sec\taccuracy=0.544065\tlr=0.000002\n",
      "Epoch[34] Batch [5799]\tSpeed: 129.097961 samples/sec\taccuracy=0.544052\tlr=0.000002\n",
      "Epoch[34] Batch [5849]\tSpeed: 126.805395 samples/sec\taccuracy=0.544038\tlr=0.000002\n",
      "Epoch[34] Batch [5899]\tSpeed: 119.389334 samples/sec\taccuracy=0.544089\tlr=0.000002\n",
      "Epoch[34] Batch [5949]\tSpeed: 121.237472 samples/sec\taccuracy=0.543824\tlr=0.000002\n",
      "Epoch[34] Batch [5999]\tSpeed: 125.844734 samples/sec\taccuracy=0.543354\tlr=0.000002\n",
      "Epoch[34] Batch [6049]\tSpeed: 130.410128 samples/sec\taccuracy=0.543079\tlr=0.000002\n",
      "Epoch[34] Batch [6099]\tSpeed: 128.253421 samples/sec\taccuracy=0.542643\tlr=0.000002\n",
      "Epoch[34] Batch [6149]\tSpeed: 130.221653 samples/sec\taccuracy=0.542541\tlr=0.000002\n",
      "Epoch[34] Batch [6199]\tSpeed: 129.352302 samples/sec\taccuracy=0.542379\tlr=0.000002\n",
      "Epoch[34] Batch [6249]\tSpeed: 126.876473 samples/sec\taccuracy=0.542280\tlr=0.000002\n",
      "Epoch[34] Batch [6299]\tSpeed: 128.299017 samples/sec\taccuracy=0.542222\tlr=0.000002\n",
      "Epoch[34] Batch [6349]\tSpeed: 130.705636 samples/sec\taccuracy=0.542244\tlr=0.000002\n",
      "Epoch[34] Batch [6399]\tSpeed: 127.935862 samples/sec\taccuracy=0.542402\tlr=0.000002\n",
      "Epoch[34] Batch [6449]\tSpeed: 126.918436 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[34] Batch [6499]\tSpeed: 126.372017 samples/sec\taccuracy=0.542788\tlr=0.000002\n",
      "Epoch[34] Batch [6549]\tSpeed: 129.311873 samples/sec\taccuracy=0.542672\tlr=0.000002\n",
      "Epoch[34] Batch [6599]\tSpeed: 126.086633 samples/sec\taccuracy=0.542652\tlr=0.000002\n",
      "Epoch[34] Batch [6649]\tSpeed: 127.882471 samples/sec\taccuracy=0.542594\tlr=0.000002\n",
      "Epoch[34] Batch [6699]\tSpeed: 132.222272 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[34] Batch [6749]\tSpeed: 126.475943 samples/sec\taccuracy=0.542444\tlr=0.000002\n",
      "Epoch[34] Batch [6799]\tSpeed: 129.313897 samples/sec\taccuracy=0.542243\tlr=0.000002\n",
      "Epoch[34] Batch [6849]\tSpeed: 125.464970 samples/sec\taccuracy=0.541971\tlr=0.000002\n",
      "Epoch[34] Batch [6899]\tSpeed: 127.303865 samples/sec\taccuracy=0.541830\tlr=0.000002\n",
      "Epoch[34] Batch [6949]\tSpeed: 125.900112 samples/sec\taccuracy=0.541942\tlr=0.000002\n",
      "Epoch[34] Batch [6999]\tSpeed: 127.785390 samples/sec\taccuracy=0.541875\tlr=0.000002\n",
      "Epoch[34] Batch [7049]\tSpeed: 127.690360 samples/sec\taccuracy=0.541915\tlr=0.000002\n",
      "Epoch[34] Batch [7099]\tSpeed: 131.567661 samples/sec\taccuracy=0.541989\tlr=0.000002\n",
      "Epoch[34] Batch [7149]\tSpeed: 125.613504 samples/sec\taccuracy=0.541853\tlr=0.000002\n",
      "Epoch[34] Batch [7199]\tSpeed: 126.508836 samples/sec\taccuracy=0.541962\tlr=0.000002\n",
      "Epoch[34] Batch [7249]\tSpeed: 127.424416 samples/sec\taccuracy=0.541776\tlr=0.000002\n",
      "Epoch[34] Batch [7299]\tSpeed: 127.316492 samples/sec\taccuracy=0.541558\tlr=0.000002\n",
      "Epoch[34] Batch [7349]\tSpeed: 130.314597 samples/sec\taccuracy=0.541548\tlr=0.000002\n",
      "Epoch[34] Batch [7399]\tSpeed: 128.500238 samples/sec\taccuracy=0.541639\tlr=0.000002\n",
      "Epoch[34] Batch [7449]\tSpeed: 127.691011 samples/sec\taccuracy=0.541292\tlr=0.000002\n",
      "Epoch[34] Batch [7499]\tSpeed: 128.160171 samples/sec\taccuracy=0.541317\tlr=0.000002\n",
      "Epoch[34] Batch [7549]\tSpeed: 127.419452 samples/sec\taccuracy=0.541209\tlr=0.000002\n",
      "Epoch[34] Batch [7599]\tSpeed: 128.033749 samples/sec\taccuracy=0.541184\tlr=0.000002\n",
      "Epoch[34] Batch [7649]\tSpeed: 131.701616 samples/sec\taccuracy=0.541422\tlr=0.000002\n",
      "Epoch[34] Batch [7699]\tSpeed: 126.015472 samples/sec\taccuracy=0.541429\tlr=0.000002\n",
      "Epoch[34] Batch [7749]\tSpeed: 125.349005 samples/sec\taccuracy=0.541484\tlr=0.000002\n",
      "Epoch[34] Batch [7799]\tSpeed: 129.156826 samples/sec\taccuracy=0.541314\tlr=0.000002\n",
      "Epoch[34] Batch [7849]\tSpeed: 128.384236 samples/sec\taccuracy=0.541417\tlr=0.000002\n",
      "Epoch[34] Batch [7899]\tSpeed: 127.411643 samples/sec\taccuracy=0.541155\tlr=0.000002\n",
      "Epoch[34] Batch [7949]\tSpeed: 124.239645 samples/sec\taccuracy=0.541085\tlr=0.000002\n",
      "Epoch[34] Batch [7999]\tSpeed: 129.282249 samples/sec\taccuracy=0.541000\tlr=0.000002\n",
      "Epoch[34] Batch [8049]\tSpeed: 127.453631 samples/sec\taccuracy=0.541149\tlr=0.000002\n",
      "Epoch[34] Batch [8099]\tSpeed: 128.440751 samples/sec\taccuracy=0.541219\tlr=0.000002\n",
      "Epoch[34] Batch [8149]\tSpeed: 128.581093 samples/sec\taccuracy=0.541396\tlr=0.000002\n",
      "Epoch[34] Batch [8199]\tSpeed: 126.815152 samples/sec\taccuracy=0.541585\tlr=0.000002\n",
      "Epoch[34] Batch [8249]\tSpeed: 129.883857 samples/sec\taccuracy=0.541561\tlr=0.000002\n",
      "Epoch[34] Batch [8299]\tSpeed: 126.736206 samples/sec\taccuracy=0.541536\tlr=0.000002\n",
      "Epoch[34] Batch [8349]\tSpeed: 129.670655 samples/sec\taccuracy=0.541257\tlr=0.000002\n",
      "Epoch[34] Batch [8399]\tSpeed: 127.252883 samples/sec\taccuracy=0.541116\tlr=0.000002\n",
      "Epoch[34] Batch [8449]\tSpeed: 128.831054 samples/sec\taccuracy=0.541272\tlr=0.000002\n",
      "Epoch[34] Batch [8499]\tSpeed: 126.685189 samples/sec\taccuracy=0.541397\tlr=0.000002\n",
      "Epoch[34] Batch [8549]\tSpeed: 128.517287 samples/sec\taccuracy=0.541477\tlr=0.000002\n",
      "Epoch[34] Batch [8599]\tSpeed: 129.325739 samples/sec\taccuracy=0.541555\tlr=0.000002\n",
      "Epoch[34] Batch [8649]\tSpeed: 128.663885 samples/sec\taccuracy=0.541474\tlr=0.000002\n",
      "Epoch[34] Batch [8699]\tSpeed: 129.912782 samples/sec\taccuracy=0.541451\tlr=0.000002\n",
      "Epoch[34] Batch [8749]\tSpeed: 128.359729 samples/sec\taccuracy=0.541143\tlr=0.000002\n",
      "Epoch[34] Batch [8799]\tSpeed: 121.540984 samples/sec\taccuracy=0.540937\tlr=0.000002\n",
      "Epoch[34] Batch [8849]\tSpeed: 122.715012 samples/sec\taccuracy=0.540862\tlr=0.000002\n",
      "Epoch[34] Batch [8899]\tSpeed: 128.444635 samples/sec\taccuracy=0.540758\tlr=0.000002\n",
      "Epoch[34] Batch [8949]\tSpeed: 126.224078 samples/sec\taccuracy=0.540908\tlr=0.000002\n",
      "Epoch[34] Batch [8999]\tSpeed: 128.806277 samples/sec\taccuracy=0.540944\tlr=0.000002\n",
      "Epoch[34] Batch [9049]\tSpeed: 129.448803 samples/sec\taccuracy=0.541008\tlr=0.000002\n",
      "Epoch[34] Batch [9099]\tSpeed: 128.145213 samples/sec\taccuracy=0.541044\tlr=0.000002\n",
      "Epoch[34] Batch [9149]\tSpeed: 127.698038 samples/sec\taccuracy=0.540847\tlr=0.000002\n",
      "Epoch[34] Batch [9199]\tSpeed: 127.599955 samples/sec\taccuracy=0.540707\tlr=0.000002\n",
      "Epoch[34] Batch [9249]\tSpeed: 127.555426 samples/sec\taccuracy=0.540851\tlr=0.000002\n",
      "Epoch[34] Batch [9299]\tSpeed: 126.421963 samples/sec\taccuracy=0.540672\tlr=0.000002\n",
      "Epoch[34] Batch [9349]\tSpeed: 128.123322 samples/sec\taccuracy=0.540455\tlr=0.000002\n",
      "Epoch[34] Batch [9399]\tSpeed: 129.425086 samples/sec\taccuracy=0.540505\tlr=0.000002\n",
      "Epoch[34] Batch [9449]\tSpeed: 128.281457 samples/sec\taccuracy=0.540476\tlr=0.000002\n",
      "Epoch[34] Batch [9499]\tSpeed: 128.971914 samples/sec\taccuracy=0.540592\tlr=0.000002\n",
      "Epoch[34] Batch [9549]\tSpeed: 128.987611 samples/sec\taccuracy=0.540393\tlr=0.000002\n",
      "Epoch[34] Batch [9599]\tSpeed: 128.632317 samples/sec\taccuracy=0.540273\tlr=0.000002\n",
      "Epoch[34] Batch [9649]\tSpeed: 128.195915 samples/sec\taccuracy=0.540285\tlr=0.000002\n",
      "Epoch[34] Batch [9699]\tSpeed: 128.502384 samples/sec\taccuracy=0.540399\tlr=0.000002\n",
      "Epoch[34] Batch [9749]\tSpeed: 125.528880 samples/sec\taccuracy=0.540436\tlr=0.000002\n",
      "Epoch[34] Batch [9799]\tSpeed: 131.625041 samples/sec\taccuracy=0.540548\tlr=0.000002\n",
      "Epoch[34] Batch [9849]\tSpeed: 129.242581 samples/sec\taccuracy=0.540470\tlr=0.000002\n",
      "Epoch[34] Batch [9899]\tSpeed: 126.897931 samples/sec\taccuracy=0.540404\tlr=0.000002\n",
      "Epoch[34] Batch [9949]\tSpeed: 129.280605 samples/sec\taccuracy=0.540226\tlr=0.000002\n",
      "Epoch[34] Batch [9999]\tSpeed: 128.757681 samples/sec\taccuracy=0.540075\tlr=0.000002\n",
      "Epoch[34] Batch [10049]\tSpeed: 125.950177 samples/sec\taccuracy=0.539963\tlr=0.000002\n",
      "Epoch[34] Batch [10099]\tSpeed: 127.602867 samples/sec\taccuracy=0.539901\tlr=0.000002\n",
      "Epoch[34] Batch [10149]\tSpeed: 126.390429 samples/sec\taccuracy=0.540086\tlr=0.000002\n",
      "Epoch[34] Batch [10199]\tSpeed: 126.731467 samples/sec\taccuracy=0.540368\tlr=0.000002\n",
      "Epoch[34] Batch [10249]\tSpeed: 127.165795 samples/sec\taccuracy=0.540329\tlr=0.000002\n",
      "Epoch[34] Batch [10299]\tSpeed: 128.608563 samples/sec\taccuracy=0.540473\tlr=0.000002\n",
      "Epoch[34] Batch [10349]\tSpeed: 128.332934 samples/sec\taccuracy=0.540640\tlr=0.000002\n",
      "Epoch[34] Batch [10399]\tSpeed: 128.316621 samples/sec\taccuracy=0.540769\tlr=0.000002\n",
      "Epoch[34] Batch [10449]\tSpeed: 127.889285 samples/sec\taccuracy=0.540610\tlr=0.000002\n",
      "Epoch[34] Batch [10499]\tSpeed: 125.708245 samples/sec\taccuracy=0.540619\tlr=0.000002\n",
      "Epoch[34] Batch [10549]\tSpeed: 129.038435 samples/sec\taccuracy=0.540782\tlr=0.000002\n",
      "Epoch[34] Batch [10599]\tSpeed: 128.382232 samples/sec\taccuracy=0.540590\tlr=0.000002\n",
      "Epoch[34] Batch [10649]\tSpeed: 129.784758 samples/sec\taccuracy=0.540552\tlr=0.000002\n",
      "Epoch[34] Batch [10699]\tSpeed: 126.530637 samples/sec\taccuracy=0.540561\tlr=0.000002\n",
      "Epoch[34] Batch [10749]\tSpeed: 127.727175 samples/sec\taccuracy=0.540651\tlr=0.000002\n",
      "Epoch[34] Batch [10799]\tSpeed: 131.720694 samples/sec\taccuracy=0.540509\tlr=0.000002\n",
      "Epoch[34] Batch [10849]\tSpeed: 127.740858 samples/sec\taccuracy=0.540622\tlr=0.000002\n",
      "Epoch[34] Batch [10899]\tSpeed: 128.559200 samples/sec\taccuracy=0.540768\tlr=0.000002\n",
      "Epoch[34] Batch [10949]\tSpeed: 127.299509 samples/sec\taccuracy=0.540742\tlr=0.000002\n",
      "Epoch[34] Batch [10999]\tSpeed: 131.107348 samples/sec\taccuracy=0.540784\tlr=0.000002\n",
      "Epoch[34] Batch [11049]\tSpeed: 127.331411 samples/sec\taccuracy=0.540520\tlr=0.000002\n",
      "Epoch[34] Batch [11099]\tSpeed: 127.171530 samples/sec\taccuracy=0.540552\tlr=0.000002\n",
      "Epoch[34] Batch [11149]\tSpeed: 131.869688 samples/sec\taccuracy=0.540706\tlr=0.000002\n",
      "Epoch[34] Batch [11199]\tSpeed: 127.806085 samples/sec\taccuracy=0.540603\tlr=0.000002\n",
      "Epoch[34] Batch [11249]\tSpeed: 126.598007 samples/sec\taccuracy=0.540767\tlr=0.000002\n",
      "Epoch[34] Batch [11299]\tSpeed: 126.498754 samples/sec\taccuracy=0.540730\tlr=0.000002\n",
      "Epoch[34] Batch [11349]\tSpeed: 136.352019 samples/sec\taccuracy=0.540716\tlr=0.000002\n",
      "[Epoch 34] training: accuracy=0.540720\n",
      "[Epoch 34] speed: 127 samples/sec\ttime cost: 711.315840\n",
      "[Epoch 34] validation: top1=0.543752 top5=0.877812\n",
      "Epoch[35] Batch [49]\tSpeed: 105.330098 samples/sec\taccuracy=0.527500\tlr=0.000002\n",
      "Epoch[35] Batch [99]\tSpeed: 129.540607 samples/sec\taccuracy=0.502500\tlr=0.000002\n",
      "Epoch[35] Batch [149]\tSpeed: 128.885993 samples/sec\taccuracy=0.518333\tlr=0.000002\n",
      "Epoch[35] Batch [199]\tSpeed: 129.277138 samples/sec\taccuracy=0.523750\tlr=0.000002\n",
      "Epoch[35] Batch [249]\tSpeed: 125.710289 samples/sec\taccuracy=0.530500\tlr=0.000002\n",
      "Epoch[35] Batch [299]\tSpeed: 126.263929 samples/sec\taccuracy=0.529583\tlr=0.000002\n",
      "Epoch[35] Batch [349]\tSpeed: 128.922232 samples/sec\taccuracy=0.529286\tlr=0.000002\n",
      "Epoch[35] Batch [399]\tSpeed: 127.425645 samples/sec\taccuracy=0.528438\tlr=0.000002\n",
      "Epoch[35] Batch [449]\tSpeed: 127.558501 samples/sec\taccuracy=0.525000\tlr=0.000002\n",
      "Epoch[35] Batch [499]\tSpeed: 129.059370 samples/sec\taccuracy=0.530250\tlr=0.000002\n",
      "Epoch[35] Batch [549]\tSpeed: 128.052316 samples/sec\taccuracy=0.535000\tlr=0.000002\n",
      "Epoch[35] Batch [599]\tSpeed: 129.748545 samples/sec\taccuracy=0.534583\tlr=0.000002\n",
      "Epoch[35] Batch [649]\tSpeed: 130.443933 samples/sec\taccuracy=0.534231\tlr=0.000002\n",
      "Epoch[35] Batch [699]\tSpeed: 128.529033 samples/sec\taccuracy=0.531786\tlr=0.000002\n",
      "Epoch[35] Batch [749]\tSpeed: 126.271997 samples/sec\taccuracy=0.530833\tlr=0.000002\n",
      "Epoch[35] Batch [799]\tSpeed: 128.505632 samples/sec\taccuracy=0.528750\tlr=0.000002\n",
      "Epoch[35] Batch [849]\tSpeed: 127.400323 samples/sec\taccuracy=0.528088\tlr=0.000002\n",
      "Epoch[35] Batch [899]\tSpeed: 126.435692 samples/sec\taccuracy=0.528750\tlr=0.000002\n",
      "Epoch[35] Batch [949]\tSpeed: 128.015949 samples/sec\taccuracy=0.530263\tlr=0.000002\n",
      "Epoch[35] Batch [999]\tSpeed: 129.494364 samples/sec\taccuracy=0.531875\tlr=0.000002\n",
      "Epoch[35] Batch [1049]\tSpeed: 130.486097 samples/sec\taccuracy=0.531905\tlr=0.000002\n",
      "Epoch[35] Batch [1099]\tSpeed: 130.267112 samples/sec\taccuracy=0.532955\tlr=0.000002\n",
      "Epoch[35] Batch [1149]\tSpeed: 126.952875 samples/sec\taccuracy=0.534022\tlr=0.000002\n",
      "Epoch[35] Batch [1199]\tSpeed: 128.577545 samples/sec\taccuracy=0.534583\tlr=0.000002\n",
      "Epoch[35] Batch [1249]\tSpeed: 127.406873 samples/sec\taccuracy=0.535700\tlr=0.000002\n",
      "Epoch[35] Batch [1299]\tSpeed: 126.991505 samples/sec\taccuracy=0.535096\tlr=0.000002\n",
      "Epoch[35] Batch [1349]\tSpeed: 126.266827 samples/sec\taccuracy=0.535185\tlr=0.000002\n",
      "Epoch[35] Batch [1399]\tSpeed: 127.936984 samples/sec\taccuracy=0.534554\tlr=0.000002\n",
      "Epoch[35] Batch [1449]\tSpeed: 128.357185 samples/sec\taccuracy=0.534224\tlr=0.000002\n",
      "Epoch[35] Batch [1499]\tSpeed: 125.797583 samples/sec\taccuracy=0.534500\tlr=0.000002\n",
      "Epoch[35] Batch [1549]\tSpeed: 127.242015 samples/sec\taccuracy=0.533952\tlr=0.000002\n",
      "Epoch[35] Batch [1599]\tSpeed: 128.726532 samples/sec\taccuracy=0.534766\tlr=0.000002\n",
      "Epoch[35] Batch [1649]\tSpeed: 126.346978 samples/sec\taccuracy=0.534015\tlr=0.000002\n",
      "Epoch[35] Batch [1699]\tSpeed: 129.271401 samples/sec\taccuracy=0.534559\tlr=0.000002\n",
      "Epoch[35] Batch [1749]\tSpeed: 129.272716 samples/sec\taccuracy=0.535357\tlr=0.000002\n",
      "Epoch[35] Batch [1799]\tSpeed: 126.261582 samples/sec\taccuracy=0.534306\tlr=0.000002\n",
      "Epoch[35] Batch [1849]\tSpeed: 127.545109 samples/sec\taccuracy=0.534865\tlr=0.000002\n",
      "Epoch[35] Batch [1899]\tSpeed: 124.503589 samples/sec\taccuracy=0.535395\tlr=0.000002\n",
      "Epoch[35] Batch [1949]\tSpeed: 127.611466 samples/sec\taccuracy=0.536026\tlr=0.000002\n",
      "Epoch[35] Batch [1999]\tSpeed: 125.729055 samples/sec\taccuracy=0.536250\tlr=0.000002\n",
      "Epoch[35] Batch [2049]\tSpeed: 125.951520 samples/sec\taccuracy=0.536098\tlr=0.000002\n",
      "Epoch[35] Batch [2099]\tSpeed: 126.459317 samples/sec\taccuracy=0.535714\tlr=0.000002\n",
      "Epoch[35] Batch [2149]\tSpeed: 125.929483 samples/sec\taccuracy=0.534419\tlr=0.000002\n",
      "Epoch[35] Batch [2199]\tSpeed: 117.569699 samples/sec\taccuracy=0.533977\tlr=0.000002\n",
      "Epoch[35] Batch [2249]\tSpeed: 128.227425 samples/sec\taccuracy=0.534611\tlr=0.000002\n",
      "Epoch[35] Batch [2299]\tSpeed: 130.174003 samples/sec\taccuracy=0.534402\tlr=0.000002\n",
      "Epoch[35] Batch [2349]\tSpeed: 127.190099 samples/sec\taccuracy=0.534681\tlr=0.000002\n",
      "Epoch[35] Batch [2399]\tSpeed: 128.614676 samples/sec\taccuracy=0.535990\tlr=0.000002\n",
      "Epoch[35] Batch [2449]\tSpeed: 128.852911 samples/sec\taccuracy=0.536173\tlr=0.000002\n",
      "Epoch[35] Batch [2499]\tSpeed: 128.858998 samples/sec\taccuracy=0.536000\tlr=0.000002\n",
      "Epoch[35] Batch [2549]\tSpeed: 128.170539 samples/sec\taccuracy=0.535588\tlr=0.000002\n",
      "Epoch[35] Batch [2599]\tSpeed: 128.549143 samples/sec\taccuracy=0.535192\tlr=0.000002\n",
      "Epoch[35] Batch [2649]\tSpeed: 130.069167 samples/sec\taccuracy=0.535708\tlr=0.000002\n",
      "Epoch[35] Batch [2699]\tSpeed: 128.420065 samples/sec\taccuracy=0.535556\tlr=0.000002\n",
      "Epoch[35] Batch [2749]\tSpeed: 126.131451 samples/sec\taccuracy=0.535818\tlr=0.000002\n",
      "Epoch[35] Batch [2799]\tSpeed: 126.236976 samples/sec\taccuracy=0.537143\tlr=0.000002\n",
      "Epoch[35] Batch [2849]\tSpeed: 128.330833 samples/sec\taccuracy=0.537193\tlr=0.000002\n",
      "Epoch[35] Batch [2899]\tSpeed: 126.820051 samples/sec\taccuracy=0.537457\tlr=0.000002\n",
      "Epoch[35] Batch [2949]\tSpeed: 129.796918 samples/sec\taccuracy=0.537924\tlr=0.000002\n",
      "Epoch[35] Batch [2999]\tSpeed: 129.691104 samples/sec\taccuracy=0.537333\tlr=0.000002\n",
      "Epoch[35] Batch [3049]\tSpeed: 129.131676 samples/sec\taccuracy=0.537623\tlr=0.000002\n",
      "Epoch[35] Batch [3099]\tSpeed: 124.869489 samples/sec\taccuracy=0.537137\tlr=0.000002\n",
      "Epoch[35] Batch [3149]\tSpeed: 128.810975 samples/sec\taccuracy=0.536825\tlr=0.000002\n",
      "Epoch[35] Batch [3199]\tSpeed: 125.710647 samples/sec\taccuracy=0.537188\tlr=0.000002\n",
      "Epoch[35] Batch [3249]\tSpeed: 125.953203 samples/sec\taccuracy=0.537077\tlr=0.000002\n",
      "Epoch[35] Batch [3299]\tSpeed: 128.202801 samples/sec\taccuracy=0.536477\tlr=0.000002\n",
      "Epoch[35] Batch [3349]\tSpeed: 129.993420 samples/sec\taccuracy=0.536418\tlr=0.000002\n",
      "Epoch[35] Batch [3399]\tSpeed: 127.390049 samples/sec\taccuracy=0.536397\tlr=0.000002\n",
      "Epoch[35] Batch [3449]\tSpeed: 129.718178 samples/sec\taccuracy=0.537101\tlr=0.000002\n",
      "Epoch[35] Batch [3499]\tSpeed: 130.417265 samples/sec\taccuracy=0.537393\tlr=0.000002\n",
      "Epoch[35] Batch [3549]\tSpeed: 128.230414 samples/sec\taccuracy=0.538275\tlr=0.000002\n",
      "Epoch[35] Batch [3599]\tSpeed: 128.536625 samples/sec\taccuracy=0.538785\tlr=0.000002\n",
      "Epoch[35] Batch [3649]\tSpeed: 127.466210 samples/sec\taccuracy=0.538664\tlr=0.000002\n",
      "Epoch[35] Batch [3699]\tSpeed: 128.085115 samples/sec\taccuracy=0.538851\tlr=0.000002\n",
      "Epoch[35] Batch [3749]\tSpeed: 128.554018 samples/sec\taccuracy=0.539167\tlr=0.000002\n",
      "Epoch[35] Batch [3799]\tSpeed: 126.529960 samples/sec\taccuracy=0.539507\tlr=0.000002\n",
      "Epoch[35] Batch [3849]\tSpeed: 126.178720 samples/sec\taccuracy=0.539188\tlr=0.000002\n",
      "Epoch[35] Batch [3899]\tSpeed: 130.647660 samples/sec\taccuracy=0.539167\tlr=0.000002\n",
      "Epoch[35] Batch [3949]\tSpeed: 131.382844 samples/sec\taccuracy=0.539335\tlr=0.000002\n",
      "Epoch[35] Batch [3999]\tSpeed: 128.360632 samples/sec\taccuracy=0.539813\tlr=0.000002\n",
      "Epoch[35] Batch [4049]\tSpeed: 128.800097 samples/sec\taccuracy=0.539815\tlr=0.000002\n",
      "Epoch[35] Batch [4099]\tSpeed: 126.998849 samples/sec\taccuracy=0.539634\tlr=0.000002\n",
      "Epoch[35] Batch [4149]\tSpeed: 128.207357 samples/sec\taccuracy=0.539488\tlr=0.000002\n",
      "Epoch[35] Batch [4199]\tSpeed: 129.608237 samples/sec\taccuracy=0.539613\tlr=0.000002\n",
      "Epoch[35] Batch [4249]\tSpeed: 130.818294 samples/sec\taccuracy=0.539324\tlr=0.000002\n",
      "Epoch[35] Batch [4299]\tSpeed: 128.249185 samples/sec\taccuracy=0.539273\tlr=0.000002\n",
      "Epoch[35] Batch [4349]\tSpeed: 128.796735 samples/sec\taccuracy=0.539167\tlr=0.000002\n",
      "Epoch[35] Batch [4399]\tSpeed: 127.702101 samples/sec\taccuracy=0.538892\tlr=0.000002\n",
      "Epoch[35] Batch [4449]\tSpeed: 129.700047 samples/sec\taccuracy=0.539129\tlr=0.000002\n",
      "Epoch[35] Batch [4499]\tSpeed: 127.189607 samples/sec\taccuracy=0.539611\tlr=0.000002\n",
      "Epoch[35] Batch [4549]\tSpeed: 127.541715 samples/sec\taccuracy=0.539588\tlr=0.000002\n",
      "Epoch[35] Batch [4599]\tSpeed: 127.996397 samples/sec\taccuracy=0.539755\tlr=0.000002\n",
      "Epoch[35] Batch [4649]\tSpeed: 127.044760 samples/sec\taccuracy=0.539946\tlr=0.000002\n",
      "Epoch[35] Batch [4699]\tSpeed: 126.803545 samples/sec\taccuracy=0.539521\tlr=0.000002\n",
      "Epoch[35] Batch [4749]\tSpeed: 126.047955 samples/sec\taccuracy=0.539789\tlr=0.000002\n",
      "Epoch[35] Batch [4799]\tSpeed: 129.235095 samples/sec\taccuracy=0.539583\tlr=0.000002\n",
      "Epoch[35] Batch [4849]\tSpeed: 129.614175 samples/sec\taccuracy=0.539562\tlr=0.000002\n",
      "Epoch[35] Batch [4899]\tSpeed: 128.760112 samples/sec\taccuracy=0.539847\tlr=0.000002\n",
      "Epoch[35] Batch [4949]\tSpeed: 129.447774 samples/sec\taccuracy=0.539545\tlr=0.000002\n",
      "Epoch[35] Batch [4999]\tSpeed: 128.878963 samples/sec\taccuracy=0.539450\tlr=0.000002\n",
      "Epoch[35] Batch [5049]\tSpeed: 122.306780 samples/sec\taccuracy=0.539455\tlr=0.000002\n",
      "Epoch[35] Batch [5099]\tSpeed: 121.170645 samples/sec\taccuracy=0.539583\tlr=0.000002\n",
      "Epoch[35] Batch [5149]\tSpeed: 127.693956 samples/sec\taccuracy=0.539733\tlr=0.000002\n",
      "Epoch[35] Batch [5199]\tSpeed: 126.664969 samples/sec\taccuracy=0.539736\tlr=0.000002\n",
      "Epoch[35] Batch [5249]\tSpeed: 131.013269 samples/sec\taccuracy=0.539143\tlr=0.000002\n",
      "Epoch[35] Batch [5299]\tSpeed: 128.603023 samples/sec\taccuracy=0.539340\tlr=0.000002\n",
      "Epoch[35] Batch [5349]\tSpeed: 130.030688 samples/sec\taccuracy=0.539276\tlr=0.000002\n",
      "Epoch[35] Batch [5399]\tSpeed: 126.637979 samples/sec\taccuracy=0.539444\tlr=0.000002\n",
      "Epoch[35] Batch [5449]\tSpeed: 128.124692 samples/sec\taccuracy=0.539541\tlr=0.000002\n",
      "Epoch[35] Batch [5499]\tSpeed: 128.887023 samples/sec\taccuracy=0.540023\tlr=0.000002\n",
      "Epoch[35] Batch [5549]\tSpeed: 132.522301 samples/sec\taccuracy=0.539482\tlr=0.000002\n",
      "Epoch[35] Batch [5599]\tSpeed: 127.131095 samples/sec\taccuracy=0.539844\tlr=0.000002\n",
      "Epoch[35] Batch [5649]\tSpeed: 127.675298 samples/sec\taccuracy=0.540177\tlr=0.000002\n",
      "Epoch[35] Batch [5699]\tSpeed: 126.295523 samples/sec\taccuracy=0.540044\tlr=0.000002\n",
      "Epoch[35] Batch [5749]\tSpeed: 127.964609 samples/sec\taccuracy=0.540196\tlr=0.000002\n",
      "Epoch[35] Batch [5799]\tSpeed: 126.694851 samples/sec\taccuracy=0.540108\tlr=0.000002\n",
      "Epoch[35] Batch [5849]\tSpeed: 128.717239 samples/sec\taccuracy=0.540449\tlr=0.000002\n",
      "Epoch[35] Batch [5899]\tSpeed: 125.391707 samples/sec\taccuracy=0.540593\tlr=0.000002\n",
      "Epoch[35] Batch [5949]\tSpeed: 129.791566 samples/sec\taccuracy=0.540525\tlr=0.000002\n",
      "Epoch[35] Batch [5999]\tSpeed: 127.664126 samples/sec\taccuracy=0.540333\tlr=0.000002\n",
      "Epoch[35] Batch [6049]\tSpeed: 127.336349 samples/sec\taccuracy=0.540434\tlr=0.000002\n",
      "Epoch[35] Batch [6099]\tSpeed: 124.816184 samples/sec\taccuracy=0.540451\tlr=0.000002\n",
      "Epoch[35] Batch [6149]\tSpeed: 130.971950 samples/sec\taccuracy=0.540488\tlr=0.000002\n",
      "Epoch[35] Batch [6199]\tSpeed: 126.242447 samples/sec\taccuracy=0.540464\tlr=0.000002\n",
      "Epoch[35] Batch [6249]\tSpeed: 128.156186 samples/sec\taccuracy=0.540560\tlr=0.000002\n",
      "Epoch[35] Batch [6299]\tSpeed: 127.409175 samples/sec\taccuracy=0.540238\tlr=0.000002\n",
      "Epoch[35] Batch [6349]\tSpeed: 128.605231 samples/sec\taccuracy=0.540630\tlr=0.000002\n",
      "Epoch[35] Batch [6399]\tSpeed: 127.980150 samples/sec\taccuracy=0.540840\tlr=0.000002\n",
      "Epoch[35] Batch [6449]\tSpeed: 128.066881 samples/sec\taccuracy=0.540601\tlr=0.000002\n",
      "Epoch[35] Batch [6499]\tSpeed: 129.488527 samples/sec\taccuracy=0.540538\tlr=0.000002\n",
      "Epoch[35] Batch [6549]\tSpeed: 128.716311 samples/sec\taccuracy=0.540305\tlr=0.000002\n",
      "Epoch[35] Batch [6599]\tSpeed: 127.367438 samples/sec\taccuracy=0.540511\tlr=0.000002\n",
      "Epoch[35] Batch [6649]\tSpeed: 129.094027 samples/sec\taccuracy=0.540489\tlr=0.000002\n",
      "Epoch[35] Batch [6699]\tSpeed: 127.086373 samples/sec\taccuracy=0.540187\tlr=0.000002\n",
      "Epoch[35] Batch [6749]\tSpeed: 127.817370 samples/sec\taccuracy=0.540296\tlr=0.000002\n",
      "Epoch[35] Batch [6799]\tSpeed: 130.694843 samples/sec\taccuracy=0.540368\tlr=0.000002\n",
      "Epoch[35] Batch [6849]\tSpeed: 129.237703 samples/sec\taccuracy=0.540657\tlr=0.000002\n",
      "Epoch[35] Batch [6899]\tSpeed: 128.182937 samples/sec\taccuracy=0.540743\tlr=0.000002\n",
      "Epoch[35] Batch [6949]\tSpeed: 127.423420 samples/sec\taccuracy=0.540881\tlr=0.000002\n",
      "Epoch[35] Batch [6999]\tSpeed: 131.016124 samples/sec\taccuracy=0.541161\tlr=0.000002\n",
      "Epoch[35] Batch [7049]\tSpeed: 129.604472 samples/sec\taccuracy=0.541259\tlr=0.000002\n",
      "Epoch[35] Batch [7099]\tSpeed: 125.988275 samples/sec\taccuracy=0.541092\tlr=0.000002\n",
      "Epoch[35] Batch [7149]\tSpeed: 128.886587 samples/sec\taccuracy=0.540822\tlr=0.000002\n",
      "Epoch[35] Batch [7199]\tSpeed: 129.369458 samples/sec\taccuracy=0.540764\tlr=0.000002\n",
      "Epoch[35] Batch [7249]\tSpeed: 129.332319 samples/sec\taccuracy=0.540879\tlr=0.000002\n",
      "Epoch[35] Batch [7299]\tSpeed: 128.749342 samples/sec\taccuracy=0.541250\tlr=0.000002\n",
      "Epoch[35] Batch [7349]\tSpeed: 126.742218 samples/sec\taccuracy=0.541378\tlr=0.000002\n",
      "Epoch[35] Batch [7399]\tSpeed: 127.086623 samples/sec\taccuracy=0.541351\tlr=0.000002\n",
      "Epoch[35] Batch [7449]\tSpeed: 129.454996 samples/sec\taccuracy=0.541644\tlr=0.000002\n",
      "Epoch[35] Batch [7499]\tSpeed: 129.251702 samples/sec\taccuracy=0.541717\tlr=0.000002\n",
      "Epoch[35] Batch [7549]\tSpeed: 127.876857 samples/sec\taccuracy=0.541788\tlr=0.000002\n",
      "Epoch[35] Batch [7599]\tSpeed: 129.596814 samples/sec\taccuracy=0.541826\tlr=0.000002\n",
      "Epoch[35] Batch [7649]\tSpeed: 127.142194 samples/sec\taccuracy=0.541961\tlr=0.000002\n",
      "Epoch[35] Batch [7699]\tSpeed: 126.827250 samples/sec\taccuracy=0.541997\tlr=0.000002\n",
      "Epoch[35] Batch [7749]\tSpeed: 129.315571 samples/sec\taccuracy=0.542113\tlr=0.000002\n",
      "Epoch[35] Batch [7799]\tSpeed: 126.881203 samples/sec\taccuracy=0.541971\tlr=0.000002\n",
      "Epoch[35] Batch [7849]\tSpeed: 128.456308 samples/sec\taccuracy=0.541990\tlr=0.000002\n",
      "Epoch[35] Batch [7899]\tSpeed: 128.595039 samples/sec\taccuracy=0.541820\tlr=0.000002\n",
      "Epoch[35] Batch [7949]\tSpeed: 118.595226 samples/sec\taccuracy=0.541903\tlr=0.000002\n",
      "Epoch[35] Batch [7999]\tSpeed: 125.826245 samples/sec\taccuracy=0.541750\tlr=0.000002\n",
      "Epoch[35] Batch [8049]\tSpeed: 126.855665 samples/sec\taccuracy=0.541646\tlr=0.000002\n",
      "Epoch[35] Batch [8099]\tSpeed: 128.251293 samples/sec\taccuracy=0.541682\tlr=0.000002\n",
      "Epoch[35] Batch [8149]\tSpeed: 128.750458 samples/sec\taccuracy=0.541702\tlr=0.000002\n",
      "Epoch[35] Batch [8199]\tSpeed: 129.577286 samples/sec\taccuracy=0.541814\tlr=0.000002\n",
      "Epoch[35] Batch [8249]\tSpeed: 130.665823 samples/sec\taccuracy=0.542030\tlr=0.000002\n",
      "Epoch[35] Batch [8299]\tSpeed: 129.062467 samples/sec\taccuracy=0.542274\tlr=0.000002\n",
      "Epoch[35] Batch [8349]\tSpeed: 130.838719 samples/sec\taccuracy=0.542216\tlr=0.000002\n",
      "Epoch[35] Batch [8399]\tSpeed: 128.842650 samples/sec\taccuracy=0.542440\tlr=0.000002\n",
      "Epoch[35] Batch [8449]\tSpeed: 126.678053 samples/sec\taccuracy=0.542396\tlr=0.000002\n",
      "Epoch[35] Batch [8499]\tSpeed: 128.412752 samples/sec\taccuracy=0.542382\tlr=0.000002\n",
      "Epoch[35] Batch [8549]\tSpeed: 127.163530 samples/sec\taccuracy=0.542427\tlr=0.000002\n",
      "Epoch[35] Batch [8599]\tSpeed: 128.381711 samples/sec\taccuracy=0.542297\tlr=0.000002\n",
      "Epoch[35] Batch [8649]\tSpeed: 129.040470 samples/sec\taccuracy=0.542283\tlr=0.000002\n",
      "Epoch[35] Batch [8699]\tSpeed: 128.046179 samples/sec\taccuracy=0.542399\tlr=0.000002\n",
      "Epoch[35] Batch [8749]\tSpeed: 128.986054 samples/sec\taccuracy=0.542300\tlr=0.000002\n",
      "Epoch[35] Batch [8799]\tSpeed: 130.900428 samples/sec\taccuracy=0.542074\tlr=0.000002\n",
      "Epoch[35] Batch [8849]\tSpeed: 127.604594 samples/sec\taccuracy=0.542161\tlr=0.000002\n",
      "Epoch[35] Batch [8899]\tSpeed: 127.703637 samples/sec\taccuracy=0.542135\tlr=0.000002\n",
      "Epoch[35] Batch [8949]\tSpeed: 125.250298 samples/sec\taccuracy=0.542123\tlr=0.000002\n",
      "Epoch[35] Batch [8999]\tSpeed: 127.599470 samples/sec\taccuracy=0.542472\tlr=0.000002\n",
      "Epoch[35] Batch [9049]\tSpeed: 128.533710 samples/sec\taccuracy=0.542555\tlr=0.000002\n",
      "Epoch[35] Batch [9099]\tSpeed: 129.397336 samples/sec\taccuracy=0.542294\tlr=0.000002\n",
      "Epoch[35] Batch [9149]\tSpeed: 127.787560 samples/sec\taccuracy=0.542377\tlr=0.000002\n",
      "Epoch[35] Batch [9199]\tSpeed: 127.238271 samples/sec\taccuracy=0.542323\tlr=0.000002\n",
      "Epoch[35] Batch [9249]\tSpeed: 127.725677 samples/sec\taccuracy=0.542324\tlr=0.000002\n",
      "Epoch[35] Batch [9299]\tSpeed: 126.317745 samples/sec\taccuracy=0.542446\tlr=0.000002\n",
      "Epoch[35] Batch [9349]\tSpeed: 127.887716 samples/sec\taccuracy=0.542286\tlr=0.000002\n",
      "Epoch[35] Batch [9399]\tSpeed: 126.337312 samples/sec\taccuracy=0.542221\tlr=0.000002\n",
      "Epoch[35] Batch [9449]\tSpeed: 129.263672 samples/sec\taccuracy=0.542302\tlr=0.000002\n",
      "Epoch[35] Batch [9499]\tSpeed: 126.071170 samples/sec\taccuracy=0.542158\tlr=0.000002\n",
      "Epoch[35] Batch [9549]\tSpeed: 128.022250 samples/sec\taccuracy=0.542147\tlr=0.000002\n",
      "Epoch[35] Batch [9599]\tSpeed: 126.535466 samples/sec\taccuracy=0.542057\tlr=0.000002\n",
      "Epoch[35] Batch [9649]\tSpeed: 128.102357 samples/sec\taccuracy=0.542228\tlr=0.000002\n",
      "Epoch[35] Batch [9699]\tSpeed: 127.908201 samples/sec\taccuracy=0.542268\tlr=0.000002\n",
      "Epoch[35] Batch [9749]\tSpeed: 129.265265 samples/sec\taccuracy=0.542321\tlr=0.000002\n",
      "Epoch[35] Batch [9799]\tSpeed: 129.598145 samples/sec\taccuracy=0.542130\tlr=0.000002\n",
      "Epoch[35] Batch [9849]\tSpeed: 128.630552 samples/sec\taccuracy=0.542246\tlr=0.000002\n",
      "Epoch[35] Batch [9899]\tSpeed: 125.022782 samples/sec\taccuracy=0.542247\tlr=0.000002\n",
      "Epoch[35] Batch [9949]\tSpeed: 128.983138 samples/sec\taccuracy=0.542198\tlr=0.000002\n",
      "Epoch[35] Batch [9999]\tSpeed: 129.222125 samples/sec\taccuracy=0.542100\tlr=0.000002\n",
      "Epoch[35] Batch [10049]\tSpeed: 128.837772 samples/sec\taccuracy=0.542164\tlr=0.000002\n",
      "Epoch[35] Batch [10099]\tSpeed: 126.587795 samples/sec\taccuracy=0.542079\tlr=0.000002\n",
      "Epoch[35] Batch [10149]\tSpeed: 129.241665 samples/sec\taccuracy=0.541872\tlr=0.000002\n",
      "Epoch[35] Batch [10199]\tSpeed: 128.340091 samples/sec\taccuracy=0.541863\tlr=0.000002\n",
      "Epoch[35] Batch [10249]\tSpeed: 128.587006 samples/sec\taccuracy=0.541878\tlr=0.000002\n",
      "Epoch[35] Batch [10299]\tSpeed: 127.920167 samples/sec\taccuracy=0.541687\tlr=0.000002\n",
      "Epoch[35] Batch [10349]\tSpeed: 126.409294 samples/sec\taccuracy=0.541679\tlr=0.000002\n",
      "Epoch[35] Batch [10399]\tSpeed: 127.809045 samples/sec\taccuracy=0.541719\tlr=0.000002\n",
      "Epoch[35] Batch [10449]\tSpeed: 129.269897 samples/sec\taccuracy=0.541794\tlr=0.000002\n",
      "Epoch[35] Batch [10499]\tSpeed: 127.183349 samples/sec\taccuracy=0.541595\tlr=0.000002\n",
      "Epoch[35] Batch [10549]\tSpeed: 129.973319 samples/sec\taccuracy=0.541434\tlr=0.000002\n",
      "Epoch[35] Batch [10599]\tSpeed: 130.240344 samples/sec\taccuracy=0.541450\tlr=0.000002\n",
      "Epoch[35] Batch [10649]\tSpeed: 127.761860 samples/sec\taccuracy=0.541467\tlr=0.000002\n",
      "Epoch[35] Batch [10699]\tSpeed: 131.329334 samples/sec\taccuracy=0.541542\tlr=0.000002\n",
      "Epoch[35] Batch [10749]\tSpeed: 125.684702 samples/sec\taccuracy=0.541593\tlr=0.000002\n",
      "Epoch[35] Batch [10799]\tSpeed: 125.108563 samples/sec\taccuracy=0.541806\tlr=0.000002\n",
      "Epoch[35] Batch [10849]\tSpeed: 120.892981 samples/sec\taccuracy=0.541682\tlr=0.000002\n",
      "Epoch[35] Batch [10899]\tSpeed: 127.401039 samples/sec\taccuracy=0.541594\tlr=0.000002\n",
      "Epoch[35] Batch [10949]\tSpeed: 127.031524 samples/sec\taccuracy=0.541644\tlr=0.000002\n",
      "Epoch[35] Batch [10999]\tSpeed: 128.206230 samples/sec\taccuracy=0.541739\tlr=0.000002\n",
      "Epoch[35] Batch [11049]\tSpeed: 129.330704 samples/sec\taccuracy=0.541889\tlr=0.000002\n",
      "Epoch[35] Batch [11099]\tSpeed: 127.674026 samples/sec\taccuracy=0.541858\tlr=0.000002\n",
      "Epoch[35] Batch [11149]\tSpeed: 127.695375 samples/sec\taccuracy=0.541906\tlr=0.000002\n",
      "Epoch[35] Batch [11199]\tSpeed: 127.899756 samples/sec\taccuracy=0.541752\tlr=0.000002\n",
      "Epoch[35] Batch [11249]\tSpeed: 128.431361 samples/sec\taccuracy=0.541778\tlr=0.000002\n",
      "Epoch[35] Batch [11299]\tSpeed: 129.283743 samples/sec\taccuracy=0.541881\tlr=0.000002\n",
      "Epoch[35] Batch [11349]\tSpeed: 137.003749 samples/sec\taccuracy=0.541927\tlr=0.000002\n",
      "[Epoch 35] training: accuracy=0.541909\n",
      "[Epoch 35] speed: 127 samples/sec\ttime cost: 710.725572\n",
      "[Epoch 35] validation: top1=0.543752 top5=0.876821\n",
      "Epoch[36] Batch [49]\tSpeed: 106.668769 samples/sec\taccuracy=0.540000\tlr=0.000002\n",
      "Epoch[36] Batch [99]\tSpeed: 128.908334 samples/sec\taccuracy=0.517500\tlr=0.000002\n",
      "Epoch[36] Batch [149]\tSpeed: 127.526989 samples/sec\taccuracy=0.527500\tlr=0.000002\n",
      "Epoch[36] Batch [199]\tSpeed: 129.223906 samples/sec\taccuracy=0.526875\tlr=0.000002\n",
      "Epoch[36] Batch [249]\tSpeed: 128.688192 samples/sec\taccuracy=0.538500\tlr=0.000002\n",
      "Epoch[36] Batch [299]\tSpeed: 130.338297 samples/sec\taccuracy=0.547500\tlr=0.000002\n",
      "Epoch[36] Batch [349]\tSpeed: 126.147014 samples/sec\taccuracy=0.537143\tlr=0.000002\n",
      "Epoch[36] Batch [399]\tSpeed: 129.462188 samples/sec\taccuracy=0.537188\tlr=0.000002\n",
      "Epoch[36] Batch [449]\tSpeed: 128.019759 samples/sec\taccuracy=0.534722\tlr=0.000002\n",
      "Epoch[36] Batch [499]\tSpeed: 127.747501 samples/sec\taccuracy=0.536750\tlr=0.000002\n",
      "Epoch[36] Batch [549]\tSpeed: 128.908740 samples/sec\taccuracy=0.540000\tlr=0.000002\n",
      "Epoch[36] Batch [599]\tSpeed: 126.680502 samples/sec\taccuracy=0.541250\tlr=0.000002\n",
      "Epoch[36] Batch [649]\tSpeed: 125.850474 samples/sec\taccuracy=0.543269\tlr=0.000002\n",
      "Epoch[36] Batch [699]\tSpeed: 127.122445 samples/sec\taccuracy=0.541786\tlr=0.000002\n",
      "Epoch[36] Batch [749]\tSpeed: 127.165390 samples/sec\taccuracy=0.539000\tlr=0.000002\n",
      "Epoch[36] Batch [799]\tSpeed: 128.252587 samples/sec\taccuracy=0.540781\tlr=0.000002\n",
      "Epoch[36] Batch [849]\tSpeed: 129.547179 samples/sec\taccuracy=0.542059\tlr=0.000002\n",
      "Epoch[36] Batch [899]\tSpeed: 127.764866 samples/sec\taccuracy=0.541111\tlr=0.000002\n",
      "Epoch[36] Batch [949]\tSpeed: 128.853030 samples/sec\taccuracy=0.543684\tlr=0.000002\n",
      "Epoch[36] Batch [999]\tSpeed: 126.840521 samples/sec\taccuracy=0.542875\tlr=0.000002\n",
      "Epoch[36] Batch [1049]\tSpeed: 126.182810 samples/sec\taccuracy=0.542619\tlr=0.000002\n",
      "Epoch[36] Batch [1099]\tSpeed: 129.576335 samples/sec\taccuracy=0.542727\tlr=0.000002\n",
      "Epoch[36] Batch [1149]\tSpeed: 126.856230 samples/sec\taccuracy=0.543478\tlr=0.000002\n",
      "Epoch[36] Batch [1199]\tSpeed: 127.678068 samples/sec\taccuracy=0.543646\tlr=0.000002\n",
      "Epoch[36] Batch [1249]\tSpeed: 130.656207 samples/sec\taccuracy=0.542400\tlr=0.000002\n",
      "Epoch[36] Batch [1299]\tSpeed: 126.049906 samples/sec\taccuracy=0.542115\tlr=0.000002\n",
      "Epoch[36] Batch [1349]\tSpeed: 118.332749 samples/sec\taccuracy=0.541667\tlr=0.000002\n",
      "Epoch[36] Batch [1399]\tSpeed: 125.325652 samples/sec\taccuracy=0.541518\tlr=0.000002\n",
      "Epoch[36] Batch [1449]\tSpeed: 130.010727 samples/sec\taccuracy=0.541724\tlr=0.000002\n",
      "Epoch[36] Batch [1499]\tSpeed: 127.980999 samples/sec\taccuracy=0.540750\tlr=0.000002\n",
      "Epoch[36] Batch [1549]\tSpeed: 127.359945 samples/sec\taccuracy=0.540565\tlr=0.000002\n",
      "Epoch[36] Batch [1599]\tSpeed: 128.679417 samples/sec\taccuracy=0.539141\tlr=0.000002\n",
      "Epoch[36] Batch [1649]\tSpeed: 132.491010 samples/sec\taccuracy=0.538712\tlr=0.000002\n",
      "Epoch[36] Batch [1699]\tSpeed: 126.170512 samples/sec\taccuracy=0.538529\tlr=0.000002\n",
      "Epoch[36] Batch [1749]\tSpeed: 129.617479 samples/sec\taccuracy=0.537857\tlr=0.000002\n",
      "Epoch[36] Batch [1799]\tSpeed: 125.177041 samples/sec\taccuracy=0.538542\tlr=0.000002\n",
      "Epoch[36] Batch [1849]\tSpeed: 126.171034 samples/sec\taccuracy=0.538851\tlr=0.000002\n",
      "Epoch[36] Batch [1899]\tSpeed: 127.618941 samples/sec\taccuracy=0.539079\tlr=0.000002\n",
      "Epoch[36] Batch [1949]\tSpeed: 128.693700 samples/sec\taccuracy=0.539167\tlr=0.000002\n",
      "Epoch[36] Batch [1999]\tSpeed: 128.815277 samples/sec\taccuracy=0.539875\tlr=0.000002\n",
      "Epoch[36] Batch [2049]\tSpeed: 128.942118 samples/sec\taccuracy=0.540183\tlr=0.000002\n",
      "Epoch[36] Batch [2099]\tSpeed: 128.047049 samples/sec\taccuracy=0.540179\tlr=0.000002\n",
      "Epoch[36] Batch [2149]\tSpeed: 131.086993 samples/sec\taccuracy=0.539884\tlr=0.000002\n",
      "Epoch[36] Batch [2199]\tSpeed: 128.139644 samples/sec\taccuracy=0.539773\tlr=0.000002\n",
      "Epoch[36] Batch [2249]\tSpeed: 127.921259 samples/sec\taccuracy=0.540278\tlr=0.000002\n",
      "Epoch[36] Batch [2299]\tSpeed: 126.077518 samples/sec\taccuracy=0.539728\tlr=0.000002\n",
      "Epoch[36] Batch [2349]\tSpeed: 130.430253 samples/sec\taccuracy=0.540213\tlr=0.000002\n",
      "Epoch[36] Batch [2399]\tSpeed: 125.904987 samples/sec\taccuracy=0.540052\tlr=0.000002\n",
      "Epoch[36] Batch [2449]\tSpeed: 127.878494 samples/sec\taccuracy=0.540255\tlr=0.000002\n",
      "Epoch[36] Batch [2499]\tSpeed: 128.092332 samples/sec\taccuracy=0.541050\tlr=0.000002\n",
      "Epoch[36] Batch [2549]\tSpeed: 131.958151 samples/sec\taccuracy=0.541176\tlr=0.000002\n",
      "Epoch[36] Batch [2599]\tSpeed: 127.560605 samples/sec\taccuracy=0.541010\tlr=0.000002\n",
      "Epoch[36] Batch [2649]\tSpeed: 130.974813 samples/sec\taccuracy=0.541415\tlr=0.000002\n",
      "Epoch[36] Batch [2699]\tSpeed: 127.375126 samples/sec\taccuracy=0.541528\tlr=0.000002\n",
      "Epoch[36] Batch [2749]\tSpeed: 128.559614 samples/sec\taccuracy=0.542136\tlr=0.000002\n",
      "Epoch[36] Batch [2799]\tSpeed: 125.808770 samples/sec\taccuracy=0.541518\tlr=0.000002\n",
      "Epoch[36] Batch [2849]\tSpeed: 126.548905 samples/sec\taccuracy=0.541360\tlr=0.000002\n",
      "Epoch[36] Batch [2899]\tSpeed: 130.233217 samples/sec\taccuracy=0.540819\tlr=0.000002\n",
      "Epoch[36] Batch [2949]\tSpeed: 131.055751 samples/sec\taccuracy=0.540847\tlr=0.000002\n",
      "Epoch[36] Batch [2999]\tSpeed: 126.667293 samples/sec\taccuracy=0.540875\tlr=0.000002\n",
      "Epoch[36] Batch [3049]\tSpeed: 130.501667 samples/sec\taccuracy=0.541721\tlr=0.000002\n",
      "Epoch[36] Batch [3099]\tSpeed: 128.946994 samples/sec\taccuracy=0.540726\tlr=0.000002\n",
      "Epoch[36] Batch [3149]\tSpeed: 129.505929 samples/sec\taccuracy=0.540476\tlr=0.000002\n",
      "Epoch[36] Batch [3199]\tSpeed: 128.139684 samples/sec\taccuracy=0.539883\tlr=0.000002\n",
      "Epoch[36] Batch [3249]\tSpeed: 126.541212 samples/sec\taccuracy=0.540615\tlr=0.000002\n",
      "Epoch[36] Batch [3299]\tSpeed: 127.954274 samples/sec\taccuracy=0.540530\tlr=0.000002\n",
      "Epoch[36] Batch [3349]\tSpeed: 126.371760 samples/sec\taccuracy=0.541343\tlr=0.000002\n",
      "Epoch[36] Batch [3399]\tSpeed: 126.266428 samples/sec\taccuracy=0.541434\tlr=0.000002\n",
      "Epoch[36] Batch [3449]\tSpeed: 128.190155 samples/sec\taccuracy=0.541486\tlr=0.000002\n",
      "Epoch[36] Batch [3499]\tSpeed: 129.148813 samples/sec\taccuracy=0.541357\tlr=0.000002\n",
      "Epoch[36] Batch [3549]\tSpeed: 129.116838 samples/sec\taccuracy=0.541549\tlr=0.000002\n",
      "Epoch[36] Batch [3599]\tSpeed: 126.635379 samples/sec\taccuracy=0.541875\tlr=0.000002\n",
      "Epoch[36] Batch [3649]\tSpeed: 128.523982 samples/sec\taccuracy=0.541678\tlr=0.000002\n",
      "Epoch[36] Batch [3699]\tSpeed: 127.264910 samples/sec\taccuracy=0.541791\tlr=0.000002\n",
      "Epoch[36] Batch [3749]\tSpeed: 130.531569 samples/sec\taccuracy=0.541700\tlr=0.000002\n",
      "Epoch[36] Batch [3799]\tSpeed: 127.540910 samples/sec\taccuracy=0.540987\tlr=0.000002\n",
      "Epoch[36] Batch [3849]\tSpeed: 128.769649 samples/sec\taccuracy=0.540909\tlr=0.000002\n",
      "Epoch[36] Batch [3899]\tSpeed: 128.695250 samples/sec\taccuracy=0.540929\tlr=0.000002\n",
      "Epoch[36] Batch [3949]\tSpeed: 130.673904 samples/sec\taccuracy=0.540728\tlr=0.000002\n",
      "Epoch[36] Batch [3999]\tSpeed: 126.153113 samples/sec\taccuracy=0.541031\tlr=0.000002\n",
      "Epoch[36] Batch [4049]\tSpeed: 128.917239 samples/sec\taccuracy=0.540648\tlr=0.000002\n",
      "Epoch[36] Batch [4099]\tSpeed: 126.858158 samples/sec\taccuracy=0.540854\tlr=0.000002\n",
      "Epoch[36] Batch [4149]\tSpeed: 126.793090 samples/sec\taccuracy=0.540663\tlr=0.000002\n",
      "Epoch[36] Batch [4199]\tSpeed: 122.928811 samples/sec\taccuracy=0.540685\tlr=0.000002\n",
      "Epoch[36] Batch [4249]\tSpeed: 119.487049 samples/sec\taccuracy=0.540882\tlr=0.000002\n",
      "Epoch[36] Batch [4299]\tSpeed: 128.268551 samples/sec\taccuracy=0.540552\tlr=0.000002\n",
      "Epoch[36] Batch [4349]\tSpeed: 131.108198 samples/sec\taccuracy=0.540632\tlr=0.000002\n",
      "Epoch[36] Batch [4399]\tSpeed: 126.592590 samples/sec\taccuracy=0.540341\tlr=0.000002\n",
      "Epoch[36] Batch [4449]\tSpeed: 125.721857 samples/sec\taccuracy=0.540140\tlr=0.000002\n",
      "Epoch[36] Batch [4499]\tSpeed: 128.095569 samples/sec\taccuracy=0.540167\tlr=0.000002\n",
      "Epoch[36] Batch [4549]\tSpeed: 129.189517 samples/sec\taccuracy=0.540357\tlr=0.000002\n",
      "Epoch[36] Batch [4599]\tSpeed: 126.387459 samples/sec\taccuracy=0.540625\tlr=0.000002\n",
      "Epoch[36] Batch [4649]\tSpeed: 129.342878 samples/sec\taccuracy=0.540134\tlr=0.000002\n",
      "Epoch[36] Batch [4699]\tSpeed: 129.802692 samples/sec\taccuracy=0.539867\tlr=0.000002\n",
      "Epoch[36] Batch [4749]\tSpeed: 129.413026 samples/sec\taccuracy=0.539579\tlr=0.000002\n",
      "Epoch[36] Batch [4799]\tSpeed: 126.998042 samples/sec\taccuracy=0.539714\tlr=0.000002\n",
      "Epoch[36] Batch [4849]\tSpeed: 129.792138 samples/sec\taccuracy=0.540155\tlr=0.000002\n",
      "Epoch[36] Batch [4899]\tSpeed: 126.876607 samples/sec\taccuracy=0.540357\tlr=0.000002\n",
      "Epoch[36] Batch [4949]\tSpeed: 129.924483 samples/sec\taccuracy=0.540530\tlr=0.000002\n",
      "Epoch[36] Batch [4999]\tSpeed: 127.996865 samples/sec\taccuracy=0.540575\tlr=0.000002\n",
      "Epoch[36] Batch [5049]\tSpeed: 125.729263 samples/sec\taccuracy=0.540347\tlr=0.000002\n",
      "Epoch[36] Batch [5099]\tSpeed: 129.016029 samples/sec\taccuracy=0.540588\tlr=0.000002\n",
      "Epoch[36] Batch [5149]\tSpeed: 127.106053 samples/sec\taccuracy=0.540534\tlr=0.000002\n",
      "Epoch[36] Batch [5199]\tSpeed: 128.691341 samples/sec\taccuracy=0.540793\tlr=0.000002\n",
      "Epoch[36] Batch [5249]\tSpeed: 126.366754 samples/sec\taccuracy=0.541024\tlr=0.000002\n",
      "Epoch[36] Batch [5299]\tSpeed: 125.904779 samples/sec\taccuracy=0.541061\tlr=0.000002\n",
      "Epoch[36] Batch [5349]\tSpeed: 129.467823 samples/sec\taccuracy=0.540771\tlr=0.000002\n",
      "Epoch[36] Batch [5399]\tSpeed: 128.611048 samples/sec\taccuracy=0.541088\tlr=0.000002\n",
      "Epoch[36] Batch [5449]\tSpeed: 125.833795 samples/sec\taccuracy=0.540803\tlr=0.000002\n",
      "Epoch[36] Batch [5499]\tSpeed: 129.110529 samples/sec\taccuracy=0.540977\tlr=0.000002\n",
      "Epoch[36] Batch [5549]\tSpeed: 126.080123 samples/sec\taccuracy=0.541239\tlr=0.000002\n",
      "Epoch[36] Batch [5599]\tSpeed: 125.487784 samples/sec\taccuracy=0.541027\tlr=0.000002\n",
      "Epoch[36] Batch [5649]\tSpeed: 128.911989 samples/sec\taccuracy=0.540951\tlr=0.000002\n",
      "Epoch[36] Batch [5699]\tSpeed: 129.722892 samples/sec\taccuracy=0.541031\tlr=0.000002\n",
      "Epoch[36] Batch [5749]\tSpeed: 128.311067 samples/sec\taccuracy=0.540891\tlr=0.000002\n",
      "Epoch[36] Batch [5799]\tSpeed: 124.851294 samples/sec\taccuracy=0.540841\tlr=0.000002\n",
      "Epoch[41] Batch [1349]\tSpeed: 127.641184 samples/sec\taccuracy=0.546481\tlr=0.000002\n",
      "Epoch[41] Batch [1399]\tSpeed: 127.334291 samples/sec\taccuracy=0.547143\tlr=0.000002\n",
      "Epoch[41] Batch [1449]\tSpeed: 127.091167 samples/sec\taccuracy=0.547931\tlr=0.000002\n",
      "Epoch[41] Batch [1499]\tSpeed: 129.293188 samples/sec\taccuracy=0.547917\tlr=0.000002\n",
      "Epoch[41] Batch [1549]\tSpeed: 128.424548 samples/sec\taccuracy=0.547419\tlr=0.000002\n",
      "Epoch[41] Batch [1599]\tSpeed: 131.328388 samples/sec\taccuracy=0.547578\tlr=0.000002\n",
      "Epoch[41] Batch [1649]\tSpeed: 128.918299 samples/sec\taccuracy=0.547348\tlr=0.000002\n",
      "Epoch[41] Batch [1699]\tSpeed: 128.082534 samples/sec\taccuracy=0.547279\tlr=0.000002\n",
      "Epoch[41] Batch [1749]\tSpeed: 128.952941 samples/sec\taccuracy=0.546929\tlr=0.000002\n",
      "Epoch[41] Batch [1799]\tSpeed: 126.569803 samples/sec\taccuracy=0.547222\tlr=0.000002\n",
      "Epoch[41] Batch [1849]\tSpeed: 128.815584 samples/sec\taccuracy=0.547095\tlr=0.000002\n",
      "Epoch[41] Batch [1899]\tSpeed: 128.233756 samples/sec\taccuracy=0.547368\tlr=0.000002\n",
      "Epoch[41] Batch [1949]\tSpeed: 125.381492 samples/sec\taccuracy=0.547179\tlr=0.000002\n",
      "Epoch[41] Batch [1999]\tSpeed: 125.943029 samples/sec\taccuracy=0.548125\tlr=0.000002\n",
      "Epoch[41] Batch [2049]\tSpeed: 128.080822 samples/sec\taccuracy=0.547256\tlr=0.000002\n",
      "Epoch[41] Batch [2099]\tSpeed: 127.364963 samples/sec\taccuracy=0.546071\tlr=0.000002\n",
      "Epoch[41] Batch [2149]\tSpeed: 127.742657 samples/sec\taccuracy=0.546802\tlr=0.000002\n",
      "Epoch[41] Batch [2199]\tSpeed: 125.888105 samples/sec\taccuracy=0.547330\tlr=0.000002\n",
      "Epoch[41] Batch [2249]\tSpeed: 125.274184 samples/sec\taccuracy=0.548556\tlr=0.000002\n",
      "Epoch[41] Batch [2299]\tSpeed: 125.745226 samples/sec\taccuracy=0.548804\tlr=0.000002\n",
      "Epoch[41] Batch [2349]\tSpeed: 126.946852 samples/sec\taccuracy=0.549149\tlr=0.000002\n",
      "Epoch[41] Batch [2399]\tSpeed: 127.544352 samples/sec\taccuracy=0.548542\tlr=0.000002\n",
      "Epoch[41] Batch [2449]\tSpeed: 126.613542 samples/sec\taccuracy=0.548622\tlr=0.000002\n",
      "Epoch[41] Batch [2499]\tSpeed: 127.850020 samples/sec\taccuracy=0.547950\tlr=0.000002\n",
      "Epoch[41] Batch [2549]\tSpeed: 127.521066 samples/sec\taccuracy=0.548382\tlr=0.000002\n",
      "Epoch[41] Batch [2599]\tSpeed: 127.296225 samples/sec\taccuracy=0.548317\tlr=0.000002\n",
      "Epoch[41] Batch [2649]\tSpeed: 126.973379 samples/sec\taccuracy=0.548302\tlr=0.000002\n",
      "Epoch[41] Batch [2699]\tSpeed: 129.447225 samples/sec\taccuracy=0.548056\tlr=0.000002\n",
      "Epoch[41] Batch [2749]\tSpeed: 124.151164 samples/sec\taccuracy=0.548227\tlr=0.000002\n",
      "Epoch[41] Batch [2799]\tSpeed: 117.904938 samples/sec\taccuracy=0.547902\tlr=0.000002\n",
      "Epoch[41] Batch [2849]\tSpeed: 127.041451 samples/sec\taccuracy=0.548333\tlr=0.000002\n",
      "Epoch[41] Batch [2899]\tSpeed: 127.211508 samples/sec\taccuracy=0.548448\tlr=0.000002\n",
      "Epoch[41] Batch [2949]\tSpeed: 125.436079 samples/sec\taccuracy=0.548856\tlr=0.000002\n",
      "Epoch[41] Batch [2999]\tSpeed: 129.275774 samples/sec\taccuracy=0.548625\tlr=0.000002\n",
      "Epoch[41] Batch [3049]\tSpeed: 128.213324 samples/sec\taccuracy=0.549262\tlr=0.000002\n",
      "Epoch[41] Batch [3099]\tSpeed: 129.427612 samples/sec\taccuracy=0.548790\tlr=0.000002\n",
      "Epoch[41] Batch [3149]\tSpeed: 125.716582 samples/sec\taccuracy=0.549008\tlr=0.000002\n",
      "Epoch[41] Batch [3199]\tSpeed: 130.244500 samples/sec\taccuracy=0.549102\tlr=0.000002\n",
      "Epoch[41] Batch [3249]\tSpeed: 129.440773 samples/sec\taccuracy=0.549038\tlr=0.000002\n",
      "Epoch[41] Batch [3299]\tSpeed: 128.211707 samples/sec\taccuracy=0.548485\tlr=0.000002\n",
      "Epoch[41] Batch [3349]\tSpeed: 127.489960 samples/sec\taccuracy=0.548396\tlr=0.000002\n",
      "Epoch[41] Batch [3399]\tSpeed: 128.427714 samples/sec\taccuracy=0.548897\tlr=0.000002\n",
      "Epoch[41] Batch [3449]\tSpeed: 125.980971 samples/sec\taccuracy=0.548913\tlr=0.000002\n",
      "Epoch[41] Batch [3499]\tSpeed: 131.972610 samples/sec\taccuracy=0.548714\tlr=0.000002\n",
      "Epoch[41] Batch [3549]\tSpeed: 126.709635 samples/sec\taccuracy=0.548415\tlr=0.000002\n",
      "Epoch[41] Batch [3599]\tSpeed: 128.362675 samples/sec\taccuracy=0.548611\tlr=0.000002\n",
      "Epoch[41] Batch [3649]\tSpeed: 127.278823 samples/sec\taccuracy=0.548733\tlr=0.000002\n",
      "Epoch[41] Batch [3699]\tSpeed: 129.593350 samples/sec\taccuracy=0.548750\tlr=0.000002\n",
      "Epoch[41] Batch [3749]\tSpeed: 127.176215 samples/sec\taccuracy=0.548333\tlr=0.000002\n",
      "Epoch[41] Batch [3799]\tSpeed: 127.512770 samples/sec\taccuracy=0.547566\tlr=0.000002\n",
      "Epoch[41] Batch [3849]\tSpeed: 128.452315 samples/sec\taccuracy=0.546494\tlr=0.000002\n",
      "Epoch[41] Batch [3899]\tSpeed: 128.394405 samples/sec\taccuracy=0.546506\tlr=0.000002\n",
      "Epoch[41] Batch [3949]\tSpeed: 129.579607 samples/sec\taccuracy=0.546772\tlr=0.000002\n",
      "Epoch[41] Batch [3999]\tSpeed: 127.588175 samples/sec\taccuracy=0.546687\tlr=0.000002\n",
      "Epoch[41] Batch [4049]\tSpeed: 129.044033 samples/sec\taccuracy=0.546235\tlr=0.000002\n",
      "Epoch[41] Batch [4099]\tSpeed: 126.849804 samples/sec\taccuracy=0.545854\tlr=0.000002\n",
      "Epoch[41] Batch [4149]\tSpeed: 131.483090 samples/sec\taccuracy=0.545904\tlr=0.000002\n",
      "Epoch[41] Batch [4199]\tSpeed: 126.740055 samples/sec\taccuracy=0.546429\tlr=0.000002\n",
      "Epoch[41] Batch [4249]\tSpeed: 131.431599 samples/sec\taccuracy=0.546176\tlr=0.000002\n",
      "Epoch[41] Batch [4299]\tSpeed: 128.974016 samples/sec\taccuracy=0.545930\tlr=0.000002\n",
      "Epoch[41] Batch [4349]\tSpeed: 126.043410 samples/sec\taccuracy=0.545776\tlr=0.000002\n",
      "Epoch[41] Batch [4399]\tSpeed: 128.666065 samples/sec\taccuracy=0.545540\tlr=0.000002\n",
      "Epoch[41] Batch [4449]\tSpeed: 126.947208 samples/sec\taccuracy=0.545197\tlr=0.000002\n",
      "Epoch[41] Batch [4499]\tSpeed: 126.285874 samples/sec\taccuracy=0.544611\tlr=0.000002\n",
      "Epoch[41] Batch [4549]\tSpeed: 130.029710 samples/sec\taccuracy=0.544835\tlr=0.000002\n",
      "Epoch[41] Batch [4599]\tSpeed: 127.925639 samples/sec\taccuracy=0.544592\tlr=0.000002\n",
      "Epoch[41] Batch [4649]\tSpeed: 127.776436 samples/sec\taccuracy=0.544543\tlr=0.000002\n",
      "Epoch[41] Batch [4699]\tSpeed: 127.435382 samples/sec\taccuracy=0.544255\tlr=0.000002\n",
      "Epoch[41] Batch [4749]\tSpeed: 128.009942 samples/sec\taccuracy=0.544132\tlr=0.000002\n",
      "Epoch[41] Batch [4799]\tSpeed: 131.288750 samples/sec\taccuracy=0.543776\tlr=0.000002\n",
      "Epoch[41] Batch [4849]\tSpeed: 127.186955 samples/sec\taccuracy=0.544072\tlr=0.000002\n",
      "Epoch[41] Batch [4899]\tSpeed: 127.898469 samples/sec\taccuracy=0.544362\tlr=0.000002\n",
      "Epoch[41] Batch [4949]\tSpeed: 126.874352 samples/sec\taccuracy=0.544141\tlr=0.000002\n",
      "Epoch[41] Batch [4999]\tSpeed: 128.147817 samples/sec\taccuracy=0.544075\tlr=0.000002\n",
      "Epoch[41] Batch [5049]\tSpeed: 129.673652 samples/sec\taccuracy=0.544356\tlr=0.000002\n",
      "Epoch[41] Batch [5099]\tSpeed: 129.494924 samples/sec\taccuracy=0.544314\tlr=0.000002\n",
      "Epoch[41] Batch [5149]\tSpeed: 127.057788 samples/sec\taccuracy=0.544248\tlr=0.000002\n",
      "Epoch[41] Batch [5199]\tSpeed: 127.921845 samples/sec\taccuracy=0.544279\tlr=0.000002\n",
      "Epoch[41] Batch [5249]\tSpeed: 129.005979 samples/sec\taccuracy=0.543929\tlr=0.000002\n",
      "Epoch[41] Batch [5299]\tSpeed: 126.966258 samples/sec\taccuracy=0.543821\tlr=0.000002\n",
      "Epoch[41] Batch [5349]\tSpeed: 129.520056 samples/sec\taccuracy=0.543715\tlr=0.000002\n",
      "Epoch[41] Batch [5399]\tSpeed: 126.266200 samples/sec\taccuracy=0.543981\tlr=0.000002\n",
      "Epoch[41] Batch [5449]\tSpeed: 127.486734 samples/sec\taccuracy=0.543624\tlr=0.000002\n",
      "Epoch[41] Batch [5499]\tSpeed: 126.764081 samples/sec\taccuracy=0.543455\tlr=0.000002\n",
      "Epoch[41] Batch [5549]\tSpeed: 129.132282 samples/sec\taccuracy=0.543221\tlr=0.000002\n",
      "Epoch[41] Batch [5599]\tSpeed: 126.116868 samples/sec\taccuracy=0.543080\tlr=0.000002\n",
      "Epoch[41] Batch [5649]\tSpeed: 119.256520 samples/sec\taccuracy=0.542942\tlr=0.000002\n",
      "Epoch[41] Batch [5699]\tSpeed: 122.052377 samples/sec\taccuracy=0.543136\tlr=0.000002\n",
      "Epoch[41] Batch [5749]\tSpeed: 126.797584 samples/sec\taccuracy=0.543630\tlr=0.000002\n",
      "Epoch[41] Batch [5799]\tSpeed: 125.212252 samples/sec\taccuracy=0.543534\tlr=0.000002\n",
      "Epoch[41] Batch [5849]\tSpeed: 125.897306 samples/sec\taccuracy=0.544167\tlr=0.000002\n",
      "Epoch[41] Batch [5899]\tSpeed: 126.877557 samples/sec\taccuracy=0.544322\tlr=0.000002\n",
      "Epoch[41] Batch [5949]\tSpeed: 127.662154 samples/sec\taccuracy=0.544307\tlr=0.000002\n",
      "Epoch[41] Batch [5999]\tSpeed: 128.183838 samples/sec\taccuracy=0.544271\tlr=0.000002\n",
      "Epoch[41] Batch [6049]\tSpeed: 127.955826 samples/sec\taccuracy=0.544360\tlr=0.000002\n",
      "Epoch[41] Batch [6099]\tSpeed: 129.139180 samples/sec\taccuracy=0.544283\tlr=0.000002\n",
      "Epoch[41] Batch [6149]\tSpeed: 130.542111 samples/sec\taccuracy=0.544024\tlr=0.000002\n",
      "Epoch[41] Batch [6199]\tSpeed: 128.399298 samples/sec\taccuracy=0.544032\tlr=0.000002\n",
      "Epoch[41] Batch [6249]\tSpeed: 129.249750 samples/sec\taccuracy=0.544020\tlr=0.000002\n",
      "Epoch[41] Batch [6299]\tSpeed: 124.526396 samples/sec\taccuracy=0.544365\tlr=0.000002\n",
      "Epoch[41] Batch [6349]\tSpeed: 127.084573 samples/sec\taccuracy=0.544370\tlr=0.000002\n",
      "Epoch[41] Batch [6399]\tSpeed: 129.164324 samples/sec\taccuracy=0.544453\tlr=0.000002\n",
      "Epoch[41] Batch [6449]\tSpeed: 126.236843 samples/sec\taccuracy=0.544729\tlr=0.000002\n",
      "Epoch[41] Batch [6499]\tSpeed: 126.062853 samples/sec\taccuracy=0.544635\tlr=0.000002\n",
      "Epoch[41] Batch [6549]\tSpeed: 127.483954 samples/sec\taccuracy=0.544561\tlr=0.000002\n",
      "Epoch[41] Batch [6599]\tSpeed: 127.105725 samples/sec\taccuracy=0.544451\tlr=0.000002\n",
      "Epoch[41] Batch [6649]\tSpeed: 126.666796 samples/sec\taccuracy=0.544568\tlr=0.000002\n",
      "Epoch[41] Batch [6699]\tSpeed: 127.552992 samples/sec\taccuracy=0.544422\tlr=0.000002\n",
      "Epoch[41] Batch [6749]\tSpeed: 127.133773 samples/sec\taccuracy=0.544167\tlr=0.000002\n",
      "Epoch[41] Batch [6799]\tSpeed: 127.022474 samples/sec\taccuracy=0.544228\tlr=0.000002\n",
      "Epoch[41] Batch [6849]\tSpeed: 128.863897 samples/sec\taccuracy=0.544416\tlr=0.000002\n",
      "Epoch[41] Batch [6899]\tSpeed: 128.648711 samples/sec\taccuracy=0.544547\tlr=0.000002\n",
      "Epoch[41] Batch [6949]\tSpeed: 126.324717 samples/sec\taccuracy=0.544460\tlr=0.000002\n",
      "Epoch[41] Batch [6999]\tSpeed: 125.540894 samples/sec\taccuracy=0.544268\tlr=0.000002\n",
      "Epoch[41] Batch [7049]\tSpeed: 129.042742 samples/sec\taccuracy=0.544149\tlr=0.000002\n",
      "Epoch[41] Batch [7099]\tSpeed: 128.892142 samples/sec\taccuracy=0.544085\tlr=0.000002\n",
      "Epoch[41] Batch [7149]\tSpeed: 127.472699 samples/sec\taccuracy=0.543864\tlr=0.000002\n",
      "Epoch[41] Batch [7199]\tSpeed: 128.121737 samples/sec\taccuracy=0.543941\tlr=0.000002\n",
      "Epoch[41] Batch [7249]\tSpeed: 127.496403 samples/sec\taccuracy=0.543828\tlr=0.000002\n",
      "Epoch[41] Batch [7299]\tSpeed: 127.877256 samples/sec\taccuracy=0.543476\tlr=0.000002\n",
      "Epoch[41] Batch [7349]\tSpeed: 128.517080 samples/sec\taccuracy=0.543316\tlr=0.000002\n",
      "Epoch[41] Batch [7399]\tSpeed: 125.490290 samples/sec\taccuracy=0.543345\tlr=0.000002\n",
      "Epoch[41] Batch [7449]\tSpeed: 126.244319 samples/sec\taccuracy=0.543372\tlr=0.000002\n",
      "Epoch[41] Batch [7499]\tSpeed: 127.411856 samples/sec\taccuracy=0.543400\tlr=0.000002\n",
      "Epoch[41] Batch [7549]\tSpeed: 125.229019 samples/sec\taccuracy=0.543377\tlr=0.000002\n",
      "Epoch[41] Batch [7599]\tSpeed: 127.057557 samples/sec\taccuracy=0.543487\tlr=0.000002\n",
      "Epoch[41] Batch [7649]\tSpeed: 127.268125 samples/sec\taccuracy=0.543203\tlr=0.000002\n",
      "Epoch[41] Batch [7699]\tSpeed: 127.807847 samples/sec\taccuracy=0.543036\tlr=0.000002\n",
      "Epoch[41] Batch [7749]\tSpeed: 130.402110 samples/sec\taccuracy=0.542984\tlr=0.000002\n",
      "Epoch[41] Batch [7799]\tSpeed: 129.327065 samples/sec\taccuracy=0.542740\tlr=0.000002\n",
      "Epoch[41] Batch [7849]\tSpeed: 127.003868 samples/sec\taccuracy=0.542611\tlr=0.000002\n",
      "Epoch[41] Batch [7899]\tSpeed: 128.503122 samples/sec\taccuracy=0.542579\tlr=0.000002\n",
      "Epoch[41] Batch [7949]\tSpeed: 126.426155 samples/sec\taccuracy=0.542720\tlr=0.000002\n",
      "Epoch[41] Batch [7999]\tSpeed: 128.325857 samples/sec\taccuracy=0.542922\tlr=0.000002\n",
      "Epoch[41] Batch [8049]\tSpeed: 127.537187 samples/sec\taccuracy=0.542780\tlr=0.000002\n",
      "Epoch[41] Batch [8099]\tSpeed: 127.385552 samples/sec\taccuracy=0.542623\tlr=0.000002\n",
      "Epoch[41] Batch [8149]\tSpeed: 128.144832 samples/sec\taccuracy=0.542469\tlr=0.000002\n",
      "Epoch[41] Batch [8199]\tSpeed: 125.833172 samples/sec\taccuracy=0.542287\tlr=0.000002\n",
      "Epoch[41] Batch [8249]\tSpeed: 123.002327 samples/sec\taccuracy=0.542303\tlr=0.000002\n",
      "Epoch[41] Batch [8299]\tSpeed: 125.216430 samples/sec\taccuracy=0.542590\tlr=0.000002\n",
      "Epoch[41] Batch [8349]\tSpeed: 128.445471 samples/sec\taccuracy=0.542605\tlr=0.000002\n",
      "Epoch[41] Batch [8399]\tSpeed: 128.769708 samples/sec\taccuracy=0.542723\tlr=0.000002\n",
      "Epoch[41] Batch [8449]\tSpeed: 129.563346 samples/sec\taccuracy=0.542825\tlr=0.000002\n",
      "Epoch[41] Batch [8499]\tSpeed: 124.292900 samples/sec\taccuracy=0.542912\tlr=0.000002\n",
      "Epoch[41] Batch [8549]\tSpeed: 120.000192 samples/sec\taccuracy=0.543173\tlr=0.000002\n",
      "Epoch[41] Batch [8599]\tSpeed: 128.751901 samples/sec\taccuracy=0.543372\tlr=0.000002\n",
      "Epoch[41] Batch [8649]\tSpeed: 127.204265 samples/sec\taccuracy=0.543555\tlr=0.000002\n",
      "Epoch[41] Batch [8699]\tSpeed: 128.299812 samples/sec\taccuracy=0.543592\tlr=0.000002\n",
      "Epoch[41] Batch [8749]\tSpeed: 127.973179 samples/sec\taccuracy=0.543657\tlr=0.000002\n",
      "Epoch[41] Batch [8799]\tSpeed: 128.255872 samples/sec\taccuracy=0.543849\tlr=0.000002\n",
      "Epoch[41] Batch [8849]\tSpeed: 127.265933 samples/sec\taccuracy=0.543743\tlr=0.000002\n",
      "Epoch[41] Batch [8899]\tSpeed: 127.930077 samples/sec\taccuracy=0.543680\tlr=0.000002\n",
      "Epoch[41] Batch [8949]\tSpeed: 128.619823 samples/sec\taccuracy=0.543589\tlr=0.000002\n",
      "Epoch[41] Batch [8999]\tSpeed: 129.308106 samples/sec\taccuracy=0.543472\tlr=0.000002\n",
      "Epoch[41] Batch [9049]\tSpeed: 129.515427 samples/sec\taccuracy=0.543481\tlr=0.000002\n",
      "Epoch[41] Batch [9099]\tSpeed: 128.766803 samples/sec\taccuracy=0.543558\tlr=0.000002\n",
      "Epoch[41] Batch [9149]\tSpeed: 126.832246 samples/sec\taccuracy=0.543552\tlr=0.000002\n",
      "Epoch[41] Batch [9199]\tSpeed: 128.614696 samples/sec\taccuracy=0.543179\tlr=0.000002\n",
      "Epoch[41] Batch [9249]\tSpeed: 127.239188 samples/sec\taccuracy=0.542797\tlr=0.000002\n",
      "Epoch[41] Batch [9299]\tSpeed: 129.251523 samples/sec\taccuracy=0.543145\tlr=0.000002\n",
      "Epoch[41] Batch [9349]\tSpeed: 126.701415 samples/sec\taccuracy=0.542995\tlr=0.000002\n",
      "Epoch[41] Batch [9399]\tSpeed: 130.677334 samples/sec\taccuracy=0.543032\tlr=0.000002\n",
      "Epoch[41] Batch [9449]\tSpeed: 128.239500 samples/sec\taccuracy=0.543148\tlr=0.000002\n",
      "Epoch[41] Batch [9499]\tSpeed: 127.699195 samples/sec\taccuracy=0.543263\tlr=0.000002\n",
      "Epoch[41] Batch [9549]\tSpeed: 125.834616 samples/sec\taccuracy=0.543246\tlr=0.000002\n",
      "Epoch[41] Batch [9599]\tSpeed: 126.892124 samples/sec\taccuracy=0.543333\tlr=0.000002\n",
      "Epoch[41] Batch [9649]\tSpeed: 130.457900 samples/sec\taccuracy=0.543251\tlr=0.000002\n",
      "Epoch[41] Batch [9699]\tSpeed: 129.311176 samples/sec\taccuracy=0.543080\tlr=0.000002\n",
      "Epoch[41] Batch [9749]\tSpeed: 126.055219 samples/sec\taccuracy=0.542974\tlr=0.000002\n",
      "Epoch[41] Batch [9799]\tSpeed: 128.512237 samples/sec\taccuracy=0.542895\tlr=0.000002\n",
      "Epoch[41] Batch [9849]\tSpeed: 124.693712 samples/sec\taccuracy=0.542614\tlr=0.000002\n",
      "Epoch[41] Batch [9899]\tSpeed: 128.843540 samples/sec\taccuracy=0.542487\tlr=0.000002\n",
      "Epoch[41] Batch [9949]\tSpeed: 125.147265 samples/sec\taccuracy=0.542362\tlr=0.000002\n",
      "Epoch[41] Batch [9999]\tSpeed: 127.038257 samples/sec\taccuracy=0.542338\tlr=0.000002\n",
      "Epoch[41] Batch [10049]\tSpeed: 127.263433 samples/sec\taccuracy=0.542226\tlr=0.000002\n",
      "Epoch[41] Batch [10099]\tSpeed: 127.510909 samples/sec\taccuracy=0.542240\tlr=0.000002\n",
      "Epoch[41] Batch [10149]\tSpeed: 128.749885 samples/sec\taccuracy=0.542094\tlr=0.000002\n",
      "Epoch[41] Batch [10199]\tSpeed: 128.936361 samples/sec\taccuracy=0.542132\tlr=0.000002\n",
      "Epoch[41] Batch [10249]\tSpeed: 128.949115 samples/sec\taccuracy=0.541988\tlr=0.000002\n",
      "Epoch[41] Batch [10299]\tSpeed: 130.443892 samples/sec\taccuracy=0.542148\tlr=0.000002\n",
      "Epoch[41] Batch [10349]\tSpeed: 132.187112 samples/sec\taccuracy=0.542077\tlr=0.000002\n",
      "Epoch[41] Batch [10399]\tSpeed: 125.078036 samples/sec\taccuracy=0.541767\tlr=0.000002\n",
      "Epoch[41] Batch [10449]\tSpeed: 126.830625 samples/sec\taccuracy=0.541579\tlr=0.000002\n",
      "Epoch[41] Batch [10499]\tSpeed: 127.233205 samples/sec\taccuracy=0.541583\tlr=0.000002\n",
      "Epoch[41] Batch [10549]\tSpeed: 127.559141 samples/sec\taccuracy=0.541671\tlr=0.000002\n",
      "Epoch[41] Batch [10599]\tSpeed: 128.614538 samples/sec\taccuracy=0.541946\tlr=0.000002\n",
      "Epoch[41] Batch [10649]\tSpeed: 126.996033 samples/sec\taccuracy=0.541772\tlr=0.000002\n",
      "Epoch[41] Batch [10699]\tSpeed: 128.442865 samples/sec\taccuracy=0.541776\tlr=0.000002\n",
      "Epoch[41] Batch [10749]\tSpeed: 126.154128 samples/sec\taccuracy=0.541721\tlr=0.000002\n",
      "Epoch[41] Batch [10799]\tSpeed: 129.183966 samples/sec\taccuracy=0.541782\tlr=0.000002\n",
      "Epoch[41] Batch [10849]\tSpeed: 124.100316 samples/sec\taccuracy=0.541728\tlr=0.000002\n",
      "Epoch[41] Batch [10899]\tSpeed: 118.481867 samples/sec\taccuracy=0.541869\tlr=0.000002\n",
      "Epoch[41] Batch [10949]\tSpeed: 122.339707 samples/sec\taccuracy=0.541906\tlr=0.000002\n",
      "Epoch[41] Batch [10999]\tSpeed: 118.832528 samples/sec\taccuracy=0.541864\tlr=0.000002\n",
      "Epoch[41] Batch [11049]\tSpeed: 114.054661 samples/sec\taccuracy=0.541686\tlr=0.000002\n",
      "Epoch[41] Batch [11099]\tSpeed: 123.662747 samples/sec\taccuracy=0.541768\tlr=0.000002\n",
      "Epoch[41] Batch [11149]\tSpeed: 122.184731 samples/sec\taccuracy=0.541704\tlr=0.000002\n",
      "Epoch[41] Batch [11199]\tSpeed: 125.837655 samples/sec\taccuracy=0.541719\tlr=0.000002\n",
      "Epoch[41] Batch [11249]\tSpeed: 129.569380 samples/sec\taccuracy=0.541700\tlr=0.000002\n",
      "Epoch[41] Batch [11299]\tSpeed: 129.822509 samples/sec\taccuracy=0.541704\tlr=0.000002\n",
      "Epoch[41] Batch [11349]\tSpeed: 132.968937 samples/sec\taccuracy=0.541762\tlr=0.000002\n",
      "[Epoch 41] training: accuracy=0.541766\n",
      "[Epoch 41] speed: 127 samples/sec\ttime cost: 714.179701\n",
      "[Epoch 41] validation: top1=0.543058 top5=0.879100\n",
      "Epoch[42] Batch [49]\tSpeed: 107.094291 samples/sec\taccuracy=0.525000\tlr=0.000002\n",
      "Epoch[42] Batch [99]\tSpeed: 128.245088 samples/sec\taccuracy=0.548750\tlr=0.000002\n",
      "Epoch[42] Batch [149]\tSpeed: 89.604969 samples/sec\taccuracy=0.559167\tlr=0.000002\n",
      "Epoch[42] Batch [199]\tSpeed: 129.437238 samples/sec\taccuracy=0.551250\tlr=0.000002\n",
      "Epoch[42] Batch [249]\tSpeed: 130.078666 samples/sec\taccuracy=0.550000\tlr=0.000002\n",
      "Epoch[42] Batch [299]\tSpeed: 128.386908 samples/sec\taccuracy=0.553750\tlr=0.000002\n",
      "Epoch[42] Batch [349]\tSpeed: 129.899102 samples/sec\taccuracy=0.550357\tlr=0.000002\n",
      "Epoch[42] Batch [399]\tSpeed: 127.636912 samples/sec\taccuracy=0.551250\tlr=0.000002\n",
      "Epoch[42] Batch [449]\tSpeed: 125.363973 samples/sec\taccuracy=0.545000\tlr=0.000002\n",
      "Epoch[42] Batch [499]\tSpeed: 127.819094 samples/sec\taccuracy=0.545250\tlr=0.000002\n",
      "Epoch[42] Batch [3099]\tSpeed: 128.239784 samples/sec\taccuracy=0.540645\tlr=0.000002\n",
      "Epoch[42] Batch [3149]\tSpeed: 127.921786 samples/sec\taccuracy=0.540437\tlr=0.000002\n",
      "Epoch[42] Batch [3199]\tSpeed: 128.923579 samples/sec\taccuracy=0.540391\tlr=0.000002\n",
      "Epoch[42] Batch [3249]\tSpeed: 127.482036 samples/sec\taccuracy=0.540269\tlr=0.000002\n",
      "Epoch[42] Batch [3299]\tSpeed: 129.308176 samples/sec\taccuracy=0.540492\tlr=0.000002\n",
      "Epoch[42] Batch [3349]\tSpeed: 130.829567 samples/sec\taccuracy=0.540858\tlr=0.000002\n",
      "Epoch[42] Batch [3399]\tSpeed: 130.402516 samples/sec\taccuracy=0.541691\tlr=0.000002\n",
      "Epoch[42] Batch [3449]\tSpeed: 130.112835 samples/sec\taccuracy=0.542428\tlr=0.000002\n",
      "Epoch[42] Batch [3499]\tSpeed: 126.735191 samples/sec\taccuracy=0.542107\tlr=0.000002\n",
      "Epoch[42] Batch [3549]\tSpeed: 126.700525 samples/sec\taccuracy=0.542254\tlr=0.000002\n",
      "Epoch[42] Batch [3599]\tSpeed: 128.681036 samples/sec\taccuracy=0.541979\tlr=0.000002\n",
      "Epoch[42] Batch [3649]\tSpeed: 128.195366 samples/sec\taccuracy=0.541918\tlr=0.000002\n",
      "Epoch[42] Batch [3699]\tSpeed: 128.346168 samples/sec\taccuracy=0.542635\tlr=0.000002\n",
      "Epoch[42] Batch [3749]\tSpeed: 128.334063 samples/sec\taccuracy=0.542833\tlr=0.000002\n",
      "Epoch[42] Batch [3799]\tSpeed: 130.918355 samples/sec\taccuracy=0.542599\tlr=0.000002\n",
      "Epoch[42] Batch [3849]\tSpeed: 127.244775 samples/sec\taccuracy=0.542305\tlr=0.000002\n",
      "Epoch[42] Batch [3899]\tSpeed: 128.004815 samples/sec\taccuracy=0.542308\tlr=0.000002\n",
      "Epoch[42] Batch [3949]\tSpeed: 126.305136 samples/sec\taccuracy=0.542247\tlr=0.000002\n",
      "Epoch[42] Batch [3999]\tSpeed: 128.056871 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[42] Batch [4049]\tSpeed: 126.980221 samples/sec\taccuracy=0.542901\tlr=0.000002\n",
      "Epoch[42] Batch [4099]\tSpeed: 130.455516 samples/sec\taccuracy=0.542561\tlr=0.000002\n",
      "Epoch[42] Batch [4149]\tSpeed: 130.175255 samples/sec\taccuracy=0.542560\tlr=0.000002\n",
      "Epoch[42] Batch [4199]\tSpeed: 131.278548 samples/sec\taccuracy=0.542113\tlr=0.000002\n",
      "Epoch[42] Batch [4249]\tSpeed: 126.194997 samples/sec\taccuracy=0.542029\tlr=0.000002\n",
      "Epoch[42] Batch [4299]\tSpeed: 130.586626 samples/sec\taccuracy=0.542006\tlr=0.000002\n",
      "Epoch[42] Batch [4349]\tSpeed: 127.810175 samples/sec\taccuracy=0.542098\tlr=0.000002\n",
      "Epoch[42] Batch [4399]\tSpeed: 127.918724 samples/sec\taccuracy=0.542386\tlr=0.000002\n",
      "Epoch[42] Batch [4449]\tSpeed: 131.618112 samples/sec\taccuracy=0.542444\tlr=0.000002\n",
      "Epoch[42] Batch [4499]\tSpeed: 127.284858 samples/sec\taccuracy=0.542889\tlr=0.000002\n",
      "Epoch[42] Batch [4549]\tSpeed: 127.522966 samples/sec\taccuracy=0.542637\tlr=0.000002\n",
      "Epoch[42] Batch [4599]\tSpeed: 126.025827 samples/sec\taccuracy=0.542364\tlr=0.000002\n",
      "Epoch[42] Batch [4649]\tSpeed: 130.978494 samples/sec\taccuracy=0.542715\tlr=0.000002\n",
      "Epoch[42] Batch [4699]\tSpeed: 129.105044 samples/sec\taccuracy=0.542766\tlr=0.000002\n",
      "Epoch[42] Batch [4749]\tSpeed: 124.596024 samples/sec\taccuracy=0.543132\tlr=0.000002\n",
      "Epoch[42] Batch [4799]\tSpeed: 124.507757 samples/sec\taccuracy=0.543151\tlr=0.000002\n",
      "Epoch[42] Batch [4849]\tSpeed: 128.149393 samples/sec\taccuracy=0.543454\tlr=0.000002\n",
      "Epoch[42] Batch [4899]\tSpeed: 128.400743 samples/sec\taccuracy=0.543061\tlr=0.000002\n",
      "Epoch[42] Batch [4949]\tSpeed: 129.991476 samples/sec\taccuracy=0.543157\tlr=0.000002\n",
      "Epoch[42] Batch [4999]\tSpeed: 127.965595 samples/sec\taccuracy=0.543600\tlr=0.000002\n",
      "Epoch[42] Batch [5049]\tSpeed: 128.476661 samples/sec\taccuracy=0.543713\tlr=0.000002\n",
      "Epoch[42] Batch [5099]\tSpeed: 125.825613 samples/sec\taccuracy=0.543407\tlr=0.000002\n",
      "Epoch[42] Batch [5149]\tSpeed: 128.584917 samples/sec\taccuracy=0.543034\tlr=0.000002\n",
      "Epoch[42] Batch [5199]\tSpeed: 130.548694 samples/sec\taccuracy=0.543005\tlr=0.000002\n",
      "Epoch[42] Batch [5249]\tSpeed: 126.819245 samples/sec\taccuracy=0.543310\tlr=0.000002\n",
      "Epoch[42] Batch [5299]\tSpeed: 126.166783 samples/sec\taccuracy=0.543443\tlr=0.000002\n",
      "Epoch[42] Batch [5349]\tSpeed: 129.412866 samples/sec\taccuracy=0.543411\tlr=0.000002\n",
      "Epoch[42] Batch [5399]\tSpeed: 127.300069 samples/sec\taccuracy=0.543380\tlr=0.000002\n",
      "Epoch[42] Batch [5449]\tSpeed: 128.041664 samples/sec\taccuracy=0.543188\tlr=0.000002\n",
      "Epoch[42] Batch [5499]\tSpeed: 127.980198 samples/sec\taccuracy=0.543295\tlr=0.000002\n",
      "Epoch[42] Batch [5549]\tSpeed: 128.727312 samples/sec\taccuracy=0.543491\tlr=0.000002\n",
      "Epoch[42] Batch [5599]\tSpeed: 128.923411 samples/sec\taccuracy=0.543326\tlr=0.000002\n",
      "Epoch[42] Batch [5649]\tSpeed: 127.981487 samples/sec\taccuracy=0.542832\tlr=0.000002\n",
      "Epoch[42] Batch [5699]\tSpeed: 125.401379 samples/sec\taccuracy=0.542851\tlr=0.000002\n",
      "Epoch[42] Batch [5749]\tSpeed: 128.173017 samples/sec\taccuracy=0.542652\tlr=0.000002\n",
      "Epoch[42] Batch [5799]\tSpeed: 125.795988 samples/sec\taccuracy=0.542629\tlr=0.000002\n",
      "Epoch[42] Batch [5849]\tSpeed: 125.958687 samples/sec\taccuracy=0.542906\tlr=0.000002\n",
      "Epoch[42] Batch [5899]\tSpeed: 127.772164 samples/sec\taccuracy=0.543114\tlr=0.000002\n",
      "Epoch[42] Batch [5949]\tSpeed: 126.558527 samples/sec\taccuracy=0.543193\tlr=0.000002\n",
      "Epoch[42] Batch [5999]\tSpeed: 126.933338 samples/sec\taccuracy=0.543125\tlr=0.000002\n",
      "Epoch[42] Batch [6049]\tSpeed: 125.370099 samples/sec\taccuracy=0.543140\tlr=0.000002\n",
      "Epoch[42] Batch [6099]\tSpeed: 128.199774 samples/sec\taccuracy=0.543053\tlr=0.000002\n",
      "Epoch[42] Batch [6149]\tSpeed: 129.464646 samples/sec\taccuracy=0.542907\tlr=0.000002\n",
      "Epoch[42] Batch [6199]\tSpeed: 130.267496 samples/sec\taccuracy=0.542964\tlr=0.000002\n",
      "Epoch[42] Batch [6249]\tSpeed: 126.782636 samples/sec\taccuracy=0.542760\tlr=0.000002\n",
      "Epoch[42] Batch [6299]\tSpeed: 129.119124 samples/sec\taccuracy=0.542758\tlr=0.000002\n",
      "Epoch[42] Batch [6349]\tSpeed: 127.027340 samples/sec\taccuracy=0.542717\tlr=0.000002\n",
      "Epoch[42] Batch [6399]\tSpeed: 124.869201 samples/sec\taccuracy=0.542715\tlr=0.000002\n",
      "Epoch[42] Batch [6449]\tSpeed: 127.418494 samples/sec\taccuracy=0.542829\tlr=0.000002\n",
      "Epoch[42] Batch [6499]\tSpeed: 129.085207 samples/sec\taccuracy=0.542788\tlr=0.000002\n",
      "Epoch[42] Batch [6549]\tSpeed: 125.951217 samples/sec\taccuracy=0.542252\tlr=0.000002\n",
      "Epoch[42] Batch [6599]\tSpeed: 127.046088 samples/sec\taccuracy=0.542689\tlr=0.000002\n",
      "Epoch[42] Batch [6649]\tSpeed: 124.930960 samples/sec\taccuracy=0.542801\tlr=0.000002\n",
      "Epoch[42] Batch [6699]\tSpeed: 125.709781 samples/sec\taccuracy=0.542836\tlr=0.000002\n",
      "Epoch[42] Batch [6749]\tSpeed: 130.857598 samples/sec\taccuracy=0.542759\tlr=0.000002\n",
      "Epoch[43] Batch [2199]\tSpeed: 128.683109 samples/sec\taccuracy=0.539602\tlr=0.000002\n",
      "Epoch[43] Batch [2249]\tSpeed: 130.551518 samples/sec\taccuracy=0.540167\tlr=0.000002\n",
      "Epoch[43] Batch [2299]\tSpeed: 127.884421 samples/sec\taccuracy=0.539620\tlr=0.000002\n",
      "Epoch[43] Batch [2349]\tSpeed: 126.395286 samples/sec\taccuracy=0.539628\tlr=0.000002\n",
      "Epoch[43] Batch [2399]\tSpeed: 130.537023 samples/sec\taccuracy=0.540208\tlr=0.000002\n",
      "Epoch[43] Batch [2449]\tSpeed: 128.479062 samples/sec\taccuracy=0.540867\tlr=0.000002\n",
      "Epoch[43] Batch [2499]\tSpeed: 129.389472 samples/sec\taccuracy=0.541200\tlr=0.000002\n",
      "Epoch[43] Batch [2549]\tSpeed: 127.433747 samples/sec\taccuracy=0.540833\tlr=0.000002\n",
      "Epoch[43] Batch [2599]\tSpeed: 129.128038 samples/sec\taccuracy=0.540481\tlr=0.000002\n",
      "Epoch[43] Batch [2649]\tSpeed: 129.890554 samples/sec\taccuracy=0.540613\tlr=0.000002\n",
      "Epoch[43] Batch [2699]\tSpeed: 129.543888 samples/sec\taccuracy=0.540648\tlr=0.000002\n",
      "Epoch[43] Batch [2749]\tSpeed: 128.966025 samples/sec\taccuracy=0.540409\tlr=0.000002\n",
      "Epoch[43] Batch [2799]\tSpeed: 127.336224 samples/sec\taccuracy=0.539643\tlr=0.000002\n",
      "Epoch[43] Batch [2849]\tSpeed: 129.102998 samples/sec\taccuracy=0.540175\tlr=0.000002\n",
      "Epoch[43] Batch [2899]\tSpeed: 116.168699 samples/sec\taccuracy=0.540302\tlr=0.000002\n",
      "Epoch[43] Batch [2949]\tSpeed: 119.840382 samples/sec\taccuracy=0.540000\tlr=0.000002\n",
      "Epoch[43] Batch [2999]\tSpeed: 120.795963 samples/sec\taccuracy=0.540042\tlr=0.000002\n",
      "Epoch[43] Batch [3049]\tSpeed: 124.010265 samples/sec\taccuracy=0.540246\tlr=0.000002\n",
      "Epoch[43] Batch [3099]\tSpeed: 120.723236 samples/sec\taccuracy=0.540282\tlr=0.000002\n",
      "Epoch[43] Batch [3149]\tSpeed: 121.705206 samples/sec\taccuracy=0.540595\tlr=0.000002\n",
      "Epoch[43] Batch [3199]\tSpeed: 127.709917 samples/sec\taccuracy=0.541328\tlr=0.000002\n",
      "Epoch[43] Batch [3249]\tSpeed: 97.514972 samples/sec\taccuracy=0.541615\tlr=0.000002\n",
      "Epoch[43] Batch [3299]\tSpeed: 127.542374 samples/sec\taccuracy=0.541932\tlr=0.000002\n",
      "Epoch[43] Batch [3349]\tSpeed: 126.447041 samples/sec\taccuracy=0.542090\tlr=0.000002\n",
      "Epoch[43] Batch [3399]\tSpeed: 128.485811 samples/sec\taccuracy=0.542574\tlr=0.000002\n",
      "Epoch[43] Batch [3449]\tSpeed: 132.121790 samples/sec\taccuracy=0.542174\tlr=0.000002\n",
      "Epoch[43] Batch [3499]\tSpeed: 129.106674 samples/sec\taccuracy=0.542357\tlr=0.000002\n",
      "Epoch[43] Batch [3549]\tSpeed: 128.469715 samples/sec\taccuracy=0.542359\tlr=0.000002\n",
      "Epoch[43] Batch [3599]\tSpeed: 130.920153 samples/sec\taccuracy=0.542708\tlr=0.000002\n",
      "Epoch[43] Batch [3649]\tSpeed: 126.875024 samples/sec\taccuracy=0.542808\tlr=0.000002\n",
      "Epoch[43] Batch [3699]\tSpeed: 86.494414 samples/sec\taccuracy=0.542770\tlr=0.000002\n",
      "Epoch[43] Batch [3749]\tSpeed: 125.600461 samples/sec\taccuracy=0.542233\tlr=0.000002\n",
      "Epoch[43] Batch [3799]\tSpeed: 126.362433 samples/sec\taccuracy=0.542533\tlr=0.000002\n",
      "Epoch[43] Batch [3849]\tSpeed: 119.328653 samples/sec\taccuracy=0.542727\tlr=0.000002\n",
      "Epoch[43] Batch [3899]\tSpeed: 129.336516 samples/sec\taccuracy=0.542949\tlr=0.000002\n",
      "Epoch[43] Batch [3949]\tSpeed: 128.193143 samples/sec\taccuracy=0.543101\tlr=0.000002\n",
      "Epoch[43] Batch [3999]\tSpeed: 127.680089 samples/sec\taccuracy=0.543219\tlr=0.000002\n",
      "Epoch[43] Batch [4049]\tSpeed: 128.293405 samples/sec\taccuracy=0.543302\tlr=0.000002\n",
      "Epoch[43] Batch [4099]\tSpeed: 127.885054 samples/sec\taccuracy=0.543293\tlr=0.000002\n",
      "Epoch[43] Batch [4149]\tSpeed: 130.046744 samples/sec\taccuracy=0.543524\tlr=0.000002\n",
      "Epoch[43] Batch [4199]\tSpeed: 131.153007 samples/sec\taccuracy=0.543839\tlr=0.000002\n",
      "Epoch[43] Batch [4249]\tSpeed: 130.109263 samples/sec\taccuracy=0.543324\tlr=0.000002\n",
      "Epoch[43] Batch [4299]\tSpeed: 127.039950 samples/sec\taccuracy=0.543401\tlr=0.000002\n",
      "Epoch[43] Batch [4349]\tSpeed: 129.790150 samples/sec\taccuracy=0.543362\tlr=0.000002\n",
      "Epoch[43] Batch [4399]\tSpeed: 130.443111 samples/sec\taccuracy=0.543722\tlr=0.000002\n",
      "Epoch[43] Batch [4449]\tSpeed: 129.503620 samples/sec\taccuracy=0.542949\tlr=0.000002\n",
      "Epoch[43] Batch [4499]\tSpeed: 128.542967 samples/sec\taccuracy=0.542750\tlr=0.000002\n",
      "Epoch[43] Batch [4549]\tSpeed: 130.827067 samples/sec\taccuracy=0.542747\tlr=0.000002\n",
      "Epoch[43] Batch [4599]\tSpeed: 130.978279 samples/sec\taccuracy=0.542364\tlr=0.000002\n",
      "Epoch[43] Batch [4649]\tSpeed: 132.452779 samples/sec\taccuracy=0.542554\tlr=0.000002\n",
      "Epoch[43] Batch [4699]\tSpeed: 131.900096 samples/sec\taccuracy=0.543085\tlr=0.000002\n",
      "Epoch[43] Batch [4749]\tSpeed: 128.114438 samples/sec\taccuracy=0.543079\tlr=0.000002\n",
      "Epoch[43] Batch [4799]\tSpeed: 128.970982 samples/sec\taccuracy=0.543047\tlr=0.000002\n",
      "Epoch[43] Batch [4849]\tSpeed: 129.347754 samples/sec\taccuracy=0.542809\tlr=0.000002\n",
      "Epoch[43] Batch [4899]\tSpeed: 131.239217 samples/sec\taccuracy=0.542883\tlr=0.000002\n",
      "Epoch[43] Batch [4949]\tSpeed: 131.416527 samples/sec\taccuracy=0.542904\tlr=0.000002\n",
      "Epoch[43] Batch [4999]\tSpeed: 131.345527 samples/sec\taccuracy=0.542675\tlr=0.000002\n",
      "Epoch[43] Batch [5049]\tSpeed: 129.587314 samples/sec\taccuracy=0.542450\tlr=0.000002\n",
      "Epoch[43] Batch [5099]\tSpeed: 131.013627 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[43] Batch [5149]\tSpeed: 129.879634 samples/sec\taccuracy=0.542427\tlr=0.000002\n",
      "Epoch[43] Batch [5199]\tSpeed: 129.201953 samples/sec\taccuracy=0.542885\tlr=0.000002\n",
      "Epoch[43] Batch [5249]\tSpeed: 129.296477 samples/sec\taccuracy=0.543143\tlr=0.000002\n",
      "Epoch[43] Batch [5299]\tSpeed: 129.295191 samples/sec\taccuracy=0.543302\tlr=0.000002\n",
      "Epoch[43] Batch [5349]\tSpeed: 127.381567 samples/sec\taccuracy=0.543294\tlr=0.000002\n",
      "Epoch[43] Batch [5399]\tSpeed: 128.120250 samples/sec\taccuracy=0.543380\tlr=0.000002\n",
      "Epoch[43] Batch [5449]\tSpeed: 130.261721 samples/sec\taccuracy=0.543394\tlr=0.000002\n",
      "Epoch[43] Batch [5499]\tSpeed: 131.101272 samples/sec\taccuracy=0.543341\tlr=0.000002\n",
      "Epoch[43] Batch [5549]\tSpeed: 128.040423 samples/sec\taccuracy=0.543243\tlr=0.000002\n",
      "Epoch[43] Batch [5599]\tSpeed: 128.820767 samples/sec\taccuracy=0.543304\tlr=0.000002\n",
      "Epoch[43] Batch [5649]\tSpeed: 132.477598 samples/sec\taccuracy=0.543274\tlr=0.000002\n",
      "Epoch[43] Batch [5699]\tSpeed: 127.965156 samples/sec\taccuracy=0.543268\tlr=0.000002\n",
      "Epoch[43] Batch [5749]\tSpeed: 129.711840 samples/sec\taccuracy=0.543326\tlr=0.000002\n",
      "Epoch[43] Batch [5799]\tSpeed: 128.972489 samples/sec\taccuracy=0.543491\tlr=0.000002\n",
      "Epoch[43] Batch [5849]\tSpeed: 127.156398 samples/sec\taccuracy=0.543547\tlr=0.000002\n",
      "Epoch[43] Batch [5899]\tSpeed: 129.461569 samples/sec\taccuracy=0.543517\tlr=0.000002\n",
      "Epoch[43] Batch [5949]\tSpeed: 130.678057 samples/sec\taccuracy=0.543445\tlr=0.000002\n",
      "Epoch[43] Batch [5999]\tSpeed: 129.923024 samples/sec\taccuracy=0.543208\tlr=0.000002\n",
      "Epoch[43] Batch [6049]\tSpeed: 129.005305 samples/sec\taccuracy=0.543636\tlr=0.000002\n",
      "Epoch[43] Batch [6099]\tSpeed: 127.829115 samples/sec\taccuracy=0.543504\tlr=0.000002\n",
      "Epoch[43] Batch [6149]\tSpeed: 126.768755 samples/sec\taccuracy=0.543537\tlr=0.000002\n",
      "Epoch[43] Batch [6199]\tSpeed: 130.058842 samples/sec\taccuracy=0.543508\tlr=0.000002\n",
      "Epoch[43] Batch [6249]\tSpeed: 128.476012 samples/sec\taccuracy=0.543480\tlr=0.000002\n",
      "Epoch[43] Batch [6299]\tSpeed: 129.273831 samples/sec\taccuracy=0.543413\tlr=0.000002\n",
      "Epoch[43] Batch [6349]\tSpeed: 128.769560 samples/sec\taccuracy=0.543209\tlr=0.000002\n",
      "Epoch[43] Batch [6399]\tSpeed: 128.985161 samples/sec\taccuracy=0.543145\tlr=0.000002\n",
      "Epoch[43] Batch [6449]\tSpeed: 128.362272 samples/sec\taccuracy=0.543236\tlr=0.000002\n",
      "Epoch[43] Batch [6499]\tSpeed: 128.863650 samples/sec\taccuracy=0.543231\tlr=0.000002\n",
      "Epoch[43] Batch [6549]\tSpeed: 126.921182 samples/sec\taccuracy=0.543263\tlr=0.000002\n",
      "Epoch[43] Batch [6599]\tSpeed: 130.017094 samples/sec\taccuracy=0.543295\tlr=0.000002\n",
      "Epoch[43] Batch [6649]\tSpeed: 129.396288 samples/sec\taccuracy=0.543327\tlr=0.000002\n",
      "Epoch[43] Batch [6699]\tSpeed: 129.705452 samples/sec\taccuracy=0.543340\tlr=0.000002\n",
      "Epoch[43] Batch [6749]\tSpeed: 117.181751 samples/sec\taccuracy=0.543241\tlr=0.000002\n",
      "Epoch[43] Batch [6799]\tSpeed: 126.096081 samples/sec\taccuracy=0.543088\tlr=0.000002\n",
      "Epoch[43] Batch [6849]\tSpeed: 129.662508 samples/sec\taccuracy=0.543066\tlr=0.000002\n",
      "Epoch[43] Batch [6899]\tSpeed: 129.495663 samples/sec\taccuracy=0.542953\tlr=0.000002\n",
      "Epoch[43] Batch [6949]\tSpeed: 128.948005 samples/sec\taccuracy=0.543004\tlr=0.000002\n",
      "Epoch[43] Batch [6999]\tSpeed: 128.960236 samples/sec\taccuracy=0.542946\tlr=0.000002\n",
      "Epoch[43] Batch [7049]\tSpeed: 128.644567 samples/sec\taccuracy=0.542518\tlr=0.000002\n",
      "Epoch[43] Batch [7099]\tSpeed: 129.695204 samples/sec\taccuracy=0.542324\tlr=0.000002\n",
      "Epoch[43] Batch [7149]\tSpeed: 130.240597 samples/sec\taccuracy=0.542570\tlr=0.000002\n",
      "Epoch[43] Batch [7199]\tSpeed: 127.876447 samples/sec\taccuracy=0.542517\tlr=0.000002\n",
      "Epoch[43] Batch [7249]\tSpeed: 127.165505 samples/sec\taccuracy=0.542466\tlr=0.000002\n",
      "Epoch[43] Batch [7299]\tSpeed: 131.596236 samples/sec\taccuracy=0.542295\tlr=0.000002\n",
      "Epoch[43] Batch [7349]\tSpeed: 129.371752 samples/sec\taccuracy=0.542381\tlr=0.000002\n",
      "Epoch[43] Batch [7399]\tSpeed: 128.827770 samples/sec\taccuracy=0.542483\tlr=0.000002\n",
      "Epoch[43] Batch [7449]\tSpeed: 131.721935 samples/sec\taccuracy=0.542550\tlr=0.000002\n",
      "Epoch[43] Batch [7499]\tSpeed: 130.514123 samples/sec\taccuracy=0.542667\tlr=0.000002\n",
      "Epoch[43] Batch [7549]\tSpeed: 126.272349 samples/sec\taccuracy=0.542815\tlr=0.000002\n",
      "Epoch[43] Batch [7599]\tSpeed: 129.491805 samples/sec\taccuracy=0.542993\tlr=0.000002\n",
      "Epoch[43] Batch [7649]\tSpeed: 126.609586 samples/sec\taccuracy=0.542631\tlr=0.000002\n",
      "Epoch[43] Batch [7699]\tSpeed: 130.033993 samples/sec\taccuracy=0.542614\tlr=0.000002\n",
      "Epoch[43] Batch [7749]\tSpeed: 128.072961 samples/sec\taccuracy=0.542613\tlr=0.000002\n",
      "Epoch[43] Batch [7799]\tSpeed: 126.464236 samples/sec\taccuracy=0.542324\tlr=0.000002\n",
      "Epoch[43] Batch [7849]\tSpeed: 128.554412 samples/sec\taccuracy=0.542341\tlr=0.000002\n",
      "Epoch[43] Batch [7899]\tSpeed: 127.812005 samples/sec\taccuracy=0.542358\tlr=0.000002\n",
      "Epoch[43] Batch [7949]\tSpeed: 129.076756 samples/sec\taccuracy=0.542547\tlr=0.000002\n",
      "Epoch[43] Batch [7999]\tSpeed: 127.138783 samples/sec\taccuracy=0.542453\tlr=0.000002\n",
      "Epoch[43] Batch [8049]\tSpeed: 129.257179 samples/sec\taccuracy=0.542407\tlr=0.000002\n",
      "Epoch[43] Batch [8099]\tSpeed: 130.102240 samples/sec\taccuracy=0.542438\tlr=0.000002\n",
      "Epoch[43] Batch [8149]\tSpeed: 129.584241 samples/sec\taccuracy=0.542439\tlr=0.000002\n",
      "Epoch[43] Batch [8199]\tSpeed: 129.107081 samples/sec\taccuracy=0.542195\tlr=0.000002\n",
      "Epoch[43] Batch [8249]\tSpeed: 128.655222 samples/sec\taccuracy=0.542288\tlr=0.000002\n",
      "Epoch[43] Batch [8299]\tSpeed: 126.301912 samples/sec\taccuracy=0.542033\tlr=0.000002\n",
      "Epoch[43] Batch [8349]\tSpeed: 131.374171 samples/sec\taccuracy=0.541901\tlr=0.000002\n",
      "Epoch[43] Batch [8399]\tSpeed: 127.489273 samples/sec\taccuracy=0.542336\tlr=0.000002\n",
      "Epoch[43] Batch [8449]\tSpeed: 128.441695 samples/sec\taccuracy=0.542456\tlr=0.000002\n",
      "Epoch[43] Batch [8499]\tSpeed: 128.367173 samples/sec\taccuracy=0.542368\tlr=0.000002\n",
      "Epoch[43] Batch [8549]\tSpeed: 128.100636 samples/sec\taccuracy=0.542105\tlr=0.000002\n",
      "Epoch[43] Batch [8599]\tSpeed: 127.047416 samples/sec\taccuracy=0.542180\tlr=0.000002\n",
      "Epoch[43] Batch [8649]\tSpeed: 126.757147 samples/sec\taccuracy=0.542283\tlr=0.000002\n",
      "Epoch[43] Batch [8699]\tSpeed: 127.584721 samples/sec\taccuracy=0.542385\tlr=0.000002\n",
      "Epoch[43] Batch [8749]\tSpeed: 126.316024 samples/sec\taccuracy=0.542343\tlr=0.000002\n",
      "Epoch[43] Batch [8799]\tSpeed: 124.472340 samples/sec\taccuracy=0.542386\tlr=0.000002\n",
      "Epoch[43] Batch [8849]\tSpeed: 127.430291 samples/sec\taccuracy=0.542401\tlr=0.000002\n",
      "Epoch[43] Batch [8899]\tSpeed: 127.888232 samples/sec\taccuracy=0.542388\tlr=0.000002\n",
      "Epoch[43] Batch [8949]\tSpeed: 128.300823 samples/sec\taccuracy=0.542528\tlr=0.000002\n",
      "Epoch[43] Batch [8999]\tSpeed: 125.996506 samples/sec\taccuracy=0.542597\tlr=0.000002\n",
      "Epoch[43] Batch [9049]\tSpeed: 127.860465 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[43] Batch [9099]\tSpeed: 126.743300 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[43] Batch [9149]\tSpeed: 126.642768 samples/sec\taccuracy=0.542486\tlr=0.000002\n",
      "Epoch[43] Batch [9199]\tSpeed: 129.954976 samples/sec\taccuracy=0.542133\tlr=0.000002\n",
      "Epoch[43] Batch [9249]\tSpeed: 126.890051 samples/sec\taccuracy=0.542041\tlr=0.000002\n",
      "Epoch[43] Batch [9299]\tSpeed: 127.213457 samples/sec\taccuracy=0.542204\tlr=0.000002\n",
      "Epoch[43] Batch [9349]\tSpeed: 129.970319 samples/sec\taccuracy=0.542059\tlr=0.000002\n",
      "Epoch[43] Batch [9399]\tSpeed: 125.845168 samples/sec\taccuracy=0.542194\tlr=0.000002\n",
      "Epoch[43] Batch [9449]\tSpeed: 127.923317 samples/sec\taccuracy=0.541839\tlr=0.000002\n",
      "Epoch[43] Batch [9499]\tSpeed: 126.906013 samples/sec\taccuracy=0.541921\tlr=0.000002\n",
      "Epoch[43] Batch [9549]\tSpeed: 128.596399 samples/sec\taccuracy=0.541846\tlr=0.000002\n",
      "Epoch[43] Batch [9599]\tSpeed: 120.771685 samples/sec\taccuracy=0.542005\tlr=0.000002\n",
      "Epoch[43] Batch [9649]\tSpeed: 119.127741 samples/sec\taccuracy=0.541956\tlr=0.000002\n",
      "Epoch[43] Batch [9699]\tSpeed: 128.847182 samples/sec\taccuracy=0.541830\tlr=0.000002\n",
      "Epoch[43] Batch [9749]\tSpeed: 125.900433 samples/sec\taccuracy=0.541846\tlr=0.000002\n",
      "Epoch[43] Batch [9799]\tSpeed: 125.722281 samples/sec\taccuracy=0.541760\tlr=0.000002\n",
      "Epoch[43] Batch [9849]\tSpeed: 129.074134 samples/sec\taccuracy=0.541891\tlr=0.000002\n",
      "Epoch[43] Batch [9899]\tSpeed: 125.477178 samples/sec\taccuracy=0.541566\tlr=0.000002\n",
      "Epoch[43] Batch [9949]\tSpeed: 129.361767 samples/sec\taccuracy=0.541457\tlr=0.000002\n",
      "Epoch[43] Batch [9999]\tSpeed: 129.427682 samples/sec\taccuracy=0.541475\tlr=0.000002\n",
      "Epoch[43] Batch [10049]\tSpeed: 124.692609 samples/sec\taccuracy=0.541493\tlr=0.000002\n",
      "Epoch[43] Batch [10099]\tSpeed: 127.509940 samples/sec\taccuracy=0.541436\tlr=0.000002\n",
      "Epoch[43] Batch [10149]\tSpeed: 127.815608 samples/sec\taccuracy=0.541478\tlr=0.000002\n",
      "Epoch[43] Batch [10199]\tSpeed: 129.041492 samples/sec\taccuracy=0.541581\tlr=0.000002\n",
      "Epoch[43] Batch [10249]\tSpeed: 126.500366 samples/sec\taccuracy=0.541439\tlr=0.000002\n",
      "Epoch[43] Batch [10299]\tSpeed: 127.966874 samples/sec\taccuracy=0.541408\tlr=0.000002\n",
      "Epoch[43] Batch [10349]\tSpeed: 126.415590 samples/sec\taccuracy=0.541413\tlr=0.000002\n",
      "Epoch[43] Batch [10399]\tSpeed: 126.438427 samples/sec\taccuracy=0.541454\tlr=0.000002\n",
      "Epoch[43] Batch [10449]\tSpeed: 127.851130 samples/sec\taccuracy=0.541603\tlr=0.000002\n",
      "Epoch[43] Batch [10499]\tSpeed: 126.125411 samples/sec\taccuracy=0.541500\tlr=0.000002\n",
      "Epoch[43] Batch [10549]\tSpeed: 128.140525 samples/sec\taccuracy=0.541505\tlr=0.000002\n",
      "Epoch[43] Batch [10599]\tSpeed: 129.026328 samples/sec\taccuracy=0.541521\tlr=0.000002\n",
      "Epoch[43] Batch [10649]\tSpeed: 128.000605 samples/sec\taccuracy=0.541326\tlr=0.000002\n",
      "Epoch[43] Batch [10699]\tSpeed: 126.823905 samples/sec\taccuracy=0.541343\tlr=0.000002\n",
      "Epoch[43] Batch [10749]\tSpeed: 129.437028 samples/sec\taccuracy=0.541256\tlr=0.000002\n",
      "Epoch[43] Batch [10799]\tSpeed: 128.747761 samples/sec\taccuracy=0.541169\tlr=0.000002\n",
      "Epoch[43] Batch [10849]\tSpeed: 128.739512 samples/sec\taccuracy=0.541037\tlr=0.000002\n",
      "Epoch[43] Batch [10899]\tSpeed: 132.098488 samples/sec\taccuracy=0.541112\tlr=0.000002\n",
      "Epoch[43] Batch [10949]\tSpeed: 127.051418 samples/sec\taccuracy=0.541130\tlr=0.000002\n",
      "Epoch[43] Batch [10999]\tSpeed: 128.316827 samples/sec\taccuracy=0.541227\tlr=0.000002\n",
      "Epoch[43] Batch [11049]\tSpeed: 128.143001 samples/sec\taccuracy=0.541380\tlr=0.000002\n",
      "Epoch[43] Batch [11099]\tSpeed: 127.412165 samples/sec\taccuracy=0.541340\tlr=0.000002\n",
      "Epoch[43] Batch [11149]\tSpeed: 130.460274 samples/sec\taccuracy=0.541300\tlr=0.000002\n",
      "Epoch[43] Batch [11199]\tSpeed: 129.473368 samples/sec\taccuracy=0.541328\tlr=0.000002\n",
      "Epoch[43] Batch [11249]\tSpeed: 130.759149 samples/sec\taccuracy=0.541278\tlr=0.000002\n",
      "Epoch[43] Batch [11299]\tSpeed: 129.265295 samples/sec\taccuracy=0.541062\tlr=0.000002\n",
      "Epoch[43] Batch [11349]\tSpeed: 130.151857 samples/sec\taccuracy=0.540969\tlr=0.000002\n",
      "[Epoch 43] training: accuracy=0.541006\n",
      "[Epoch 43] speed: 127 samples/sec\ttime cost: 713.062690\n",
      "[Epoch 43] validation: top1=0.542761 top5=0.877019\n",
      "Epoch[44] Batch [49]\tSpeed: 106.538200 samples/sec\taccuracy=0.530000\tlr=0.000002\n",
      "Epoch[44] Batch [99]\tSpeed: 118.041342 samples/sec\taccuracy=0.516250\tlr=0.000002\n",
      "Epoch[44] Batch [149]\tSpeed: 123.651099 samples/sec\taccuracy=0.521667\tlr=0.000002\n",
      "Epoch[44] Batch [199]\tSpeed: 125.010635 samples/sec\taccuracy=0.531250\tlr=0.000002\n",
      "Epoch[44] Batch [249]\tSpeed: 130.114701 samples/sec\taccuracy=0.532000\tlr=0.000002\n",
      "Epoch[44] Batch [299]\tSpeed: 132.999959 samples/sec\taccuracy=0.528750\tlr=0.000002\n",
      "Epoch[44] Batch [349]\tSpeed: 127.600800 samples/sec\taccuracy=0.529643\tlr=0.000002\n",
      "Epoch[44] Batch [399]\tSpeed: 129.054922 samples/sec\taccuracy=0.533125\tlr=0.000002\n",
      "Epoch[44] Batch [449]\tSpeed: 133.957475 samples/sec\taccuracy=0.536667\tlr=0.000002\n",
      "Epoch[44] Batch [499]\tSpeed: 128.869282 samples/sec\taccuracy=0.535750\tlr=0.000002\n",
      "Epoch[44] Batch [549]\tSpeed: 127.777341 samples/sec\taccuracy=0.535227\tlr=0.000002\n",
      "Epoch[44] Batch [599]\tSpeed: 129.838273 samples/sec\taccuracy=0.539375\tlr=0.000002\n",
      "Epoch[44] Batch [649]\tSpeed: 129.286782 samples/sec\taccuracy=0.539808\tlr=0.000002\n",
      "Epoch[44] Batch [699]\tSpeed: 129.016902 samples/sec\taccuracy=0.540893\tlr=0.000002\n",
      "Epoch[44] Batch [749]\tSpeed: 130.444501 samples/sec\taccuracy=0.545167\tlr=0.000002\n",
      "Epoch[44] Batch [799]\tSpeed: 127.982219 samples/sec\taccuracy=0.543906\tlr=0.000002\n",
      "Epoch[44] Batch [849]\tSpeed: 129.470461 samples/sec\taccuracy=0.544412\tlr=0.000002\n",
      "Epoch[44] Batch [899]\tSpeed: 129.953688 samples/sec\taccuracy=0.545694\tlr=0.000002\n",
      "Epoch[44] Batch [949]\tSpeed: 131.330413 samples/sec\taccuracy=0.546184\tlr=0.000002\n",
      "Epoch[44] Batch [999]\tSpeed: 130.964067 samples/sec\taccuracy=0.545625\tlr=0.000002\n",
      "Epoch[44] Batch [1049]\tSpeed: 129.641497 samples/sec\taccuracy=0.543690\tlr=0.000002\n",
      "Epoch[44] Batch [1099]\tSpeed: 129.392476 samples/sec\taccuracy=0.544091\tlr=0.000002\n",
      "Epoch[44] Batch [1149]\tSpeed: 129.195874 samples/sec\taccuracy=0.544130\tlr=0.000002\n",
      "Epoch[44] Batch [1199]\tSpeed: 128.791871 samples/sec\taccuracy=0.544896\tlr=0.000002\n",
      "Epoch[44] Batch [1249]\tSpeed: 129.841317 samples/sec\taccuracy=0.543900\tlr=0.000002\n",
      "Epoch[44] Batch [1299]\tSpeed: 127.199915 samples/sec\taccuracy=0.544712\tlr=0.000002\n",
      "Epoch[44] Batch [1349]\tSpeed: 125.811025 samples/sec\taccuracy=0.542870\tlr=0.000002\n",
      "Epoch[44] Batch [1399]\tSpeed: 127.143543 samples/sec\taccuracy=0.542143\tlr=0.000002\n",
      "Epoch[44] Batch [1449]\tSpeed: 130.581808 samples/sec\taccuracy=0.540776\tlr=0.000002\n",
      "Epoch[44] Batch [1499]\tSpeed: 129.892978 samples/sec\taccuracy=0.542167\tlr=0.000002\n",
      "Epoch[44] Batch [1549]\tSpeed: 129.531606 samples/sec\taccuracy=0.543145\tlr=0.000002\n",
      "Epoch[44] Batch [1599]\tSpeed: 131.961960 samples/sec\taccuracy=0.543203\tlr=0.000002\n",
      "Epoch[44] Batch [1649]\tSpeed: 128.517750 samples/sec\taccuracy=0.543636\tlr=0.000002\n",
      "Epoch[44] Batch [1699]\tSpeed: 131.114858 samples/sec\taccuracy=0.542426\tlr=0.000002\n",
      "Epoch[44] Batch [1749]\tSpeed: 128.473739 samples/sec\taccuracy=0.542786\tlr=0.000002\n",
      "Epoch[44] Batch [1799]\tSpeed: 129.614725 samples/sec\taccuracy=0.542986\tlr=0.000002\n",
      "Epoch[44] Batch [1849]\tSpeed: 131.068897 samples/sec\taccuracy=0.543514\tlr=0.000002\n",
      "Epoch[44] Batch [1899]\tSpeed: 128.517327 samples/sec\taccuracy=0.544211\tlr=0.000002\n",
      "Epoch[44] Batch [1949]\tSpeed: 131.653528 samples/sec\taccuracy=0.544936\tlr=0.000002\n",
      "Epoch[44] Batch [1999]\tSpeed: 127.932106 samples/sec\taccuracy=0.544250\tlr=0.000002\n",
      "Epoch[44] Batch [2049]\tSpeed: 129.569430 samples/sec\taccuracy=0.544512\tlr=0.000002\n",
      "Epoch[44] Batch [2099]\tSpeed: 126.812478 samples/sec\taccuracy=0.544524\tlr=0.000002\n",
      "Epoch[44] Batch [2149]\tSpeed: 106.159709 samples/sec\taccuracy=0.544826\tlr=0.000002\n",
      "Epoch[44] Batch [2199]\tSpeed: 129.510438 samples/sec\taccuracy=0.544830\tlr=0.000002\n",
      "Epoch[44] Batch [2249]\tSpeed: 128.644676 samples/sec\taccuracy=0.545111\tlr=0.000002\n",
      "Epoch[44] Batch [2299]\tSpeed: 128.695724 samples/sec\taccuracy=0.544511\tlr=0.000002\n",
      "Epoch[44] Batch [2349]\tSpeed: 128.731668 samples/sec\taccuracy=0.544415\tlr=0.000002\n",
      "Epoch[44] Batch [2399]\tSpeed: 128.613660 samples/sec\taccuracy=0.544583\tlr=0.000002\n",
      "Epoch[44] Batch [2449]\tSpeed: 127.339346 samples/sec\taccuracy=0.544745\tlr=0.000002\n",
      "Epoch[44] Batch [2499]\tSpeed: 127.958636 samples/sec\taccuracy=0.544200\tlr=0.000002\n",
      "Epoch[44] Batch [2549]\tSpeed: 129.035884 samples/sec\taccuracy=0.544069\tlr=0.000002\n",
      "Epoch[44] Batch [2599]\tSpeed: 129.709563 samples/sec\taccuracy=0.544327\tlr=0.000002\n",
      "Epoch[44] Batch [2649]\tSpeed: 128.563594 samples/sec\taccuracy=0.544104\tlr=0.000002\n",
      "Epoch[44] Batch [2699]\tSpeed: 127.747744 samples/sec\taccuracy=0.543935\tlr=0.000002\n",
      "Epoch[44] Batch [2749]\tSpeed: 130.096379 samples/sec\taccuracy=0.543500\tlr=0.000002\n",
      "Epoch[44] Batch [2799]\tSpeed: 128.818126 samples/sec\taccuracy=0.543973\tlr=0.000002\n",
      "Epoch[44] Batch [2849]\tSpeed: 128.705301 samples/sec\taccuracy=0.543246\tlr=0.000002\n",
      "Epoch[44] Batch [2899]\tSpeed: 127.531225 samples/sec\taccuracy=0.542543\tlr=0.000002\n",
      "Epoch[44] Batch [2949]\tSpeed: 127.136692 samples/sec\taccuracy=0.543263\tlr=0.000002\n",
      "Epoch[44] Batch [2999]\tSpeed: 123.729897 samples/sec\taccuracy=0.543958\tlr=0.000002\n",
      "Epoch[44] Batch [3049]\tSpeed: 123.133153 samples/sec\taccuracy=0.544180\tlr=0.000002\n",
      "Epoch[44] Batch [3099]\tSpeed: 128.244156 samples/sec\taccuracy=0.544032\tlr=0.000002\n",
      "Epoch[44] Batch [3149]\tSpeed: 126.621224 samples/sec\taccuracy=0.544405\tlr=0.000002\n",
      "Epoch[44] Batch [3199]\tSpeed: 130.001942 samples/sec\taccuracy=0.544648\tlr=0.000002\n",
      "Epoch[44] Batch [3249]\tSpeed: 128.383912 samples/sec\taccuracy=0.544923\tlr=0.000002\n",
      "Epoch[44] Batch [3299]\tSpeed: 130.035394 samples/sec\taccuracy=0.545114\tlr=0.000002\n",
      "Epoch[44] Batch [3349]\tSpeed: 127.532001 samples/sec\taccuracy=0.544776\tlr=0.000002\n",
      "Epoch[44] Batch [3399]\tSpeed: 124.160260 samples/sec\taccuracy=0.544596\tlr=0.000002\n",
      "Epoch[44] Batch [3449]\tSpeed: 127.708303 samples/sec\taccuracy=0.544203\tlr=0.000002\n",
      "Epoch[44] Batch [3499]\tSpeed: 126.323842 samples/sec\taccuracy=0.544464\tlr=0.000002\n",
      "Epoch[44] Batch [3549]\tSpeed: 126.429661 samples/sec\taccuracy=0.544190\tlr=0.000002\n",
      "Epoch[44] Batch [3599]\tSpeed: 125.759327 samples/sec\taccuracy=0.544306\tlr=0.000002\n",
      "Epoch[44] Batch [3649]\tSpeed: 129.558294 samples/sec\taccuracy=0.544589\tlr=0.000002\n",
      "Epoch[44] Batch [3699]\tSpeed: 127.158518 samples/sec\taccuracy=0.544459\tlr=0.000002\n",
      "Epoch[44] Batch [3749]\tSpeed: 127.734760 samples/sec\taccuracy=0.543933\tlr=0.000002\n",
      "Epoch[44] Batch [3799]\tSpeed: 127.937326 samples/sec\taccuracy=0.544309\tlr=0.000002\n",
      "Epoch[44] Batch [3849]\tSpeed: 127.021156 samples/sec\taccuracy=0.544513\tlr=0.000002\n",
      "Epoch[44] Batch [3899]\tSpeed: 127.404280 samples/sec\taccuracy=0.544712\tlr=0.000002\n",
      "Epoch[44] Batch [3949]\tSpeed: 128.485693 samples/sec\taccuracy=0.544715\tlr=0.000002\n",
      "Epoch[44] Batch [3999]\tSpeed: 129.850935 samples/sec\taccuracy=0.544687\tlr=0.000002\n",
      "Epoch[44] Batch [4049]\tSpeed: 127.417903 samples/sec\taccuracy=0.544599\tlr=0.000002\n",
      "Epoch[44] Batch [4099]\tSpeed: 125.584365 samples/sec\taccuracy=0.544695\tlr=0.000002\n",
      "Epoch[44] Batch [4149]\tSpeed: 126.442324 samples/sec\taccuracy=0.544639\tlr=0.000002\n",
      "Epoch[44] Batch [4199]\tSpeed: 128.860730 samples/sec\taccuracy=0.544345\tlr=0.000002\n",
      "Epoch[44] Batch [4249]\tSpeed: 128.135925 samples/sec\taccuracy=0.544471\tlr=0.000002\n",
      "Epoch[44] Batch [4299]\tSpeed: 127.202857 samples/sec\taccuracy=0.544128\tlr=0.000002\n",
      "Epoch[44] Batch [4349]\tSpeed: 128.653180 samples/sec\taccuracy=0.544138\tlr=0.000002\n",
      "Epoch[44] Batch [4399]\tSpeed: 125.829878 samples/sec\taccuracy=0.544233\tlr=0.000002\n",
      "Epoch[44] Batch [4449]\tSpeed: 128.114702 samples/sec\taccuracy=0.544831\tlr=0.000002\n",
      "Epoch[44] Batch [4499]\tSpeed: 125.046787 samples/sec\taccuracy=0.544833\tlr=0.000002\n",
      "Epoch[44] Batch [4549]\tSpeed: 129.283713 samples/sec\taccuracy=0.544890\tlr=0.000002\n",
      "Epoch[44] Batch [4599]\tSpeed: 126.071559 samples/sec\taccuracy=0.544755\tlr=0.000002\n",
      "Epoch[44] Batch [4649]\tSpeed: 126.984642 samples/sec\taccuracy=0.544570\tlr=0.000002\n",
      "Epoch[44] Batch [4699]\tSpeed: 127.117436 samples/sec\taccuracy=0.544043\tlr=0.000002\n",
      "Epoch[44] Batch [4749]\tSpeed: 126.955536 samples/sec\taccuracy=0.544237\tlr=0.000002\n",
      "Epoch[44] Batch [4799]\tSpeed: 126.568037 samples/sec\taccuracy=0.544401\tlr=0.000002\n",
      "Epoch[44] Batch [4849]\tSpeed: 126.181586 samples/sec\taccuracy=0.544485\tlr=0.000002\n",
      "Epoch[44] Batch [4899]\tSpeed: 125.813544 samples/sec\taccuracy=0.544592\tlr=0.000002\n",
      "Epoch[44] Batch [4949]\tSpeed: 127.745828 samples/sec\taccuracy=0.544192\tlr=0.000002\n",
      "Epoch[44] Batch [4999]\tSpeed: 126.293935 samples/sec\taccuracy=0.544100\tlr=0.000002\n",
      "Epoch[44] Batch [5049]\tSpeed: 127.854453 samples/sec\taccuracy=0.544158\tlr=0.000002\n",
      "Epoch[44] Batch [5099]\tSpeed: 128.776676 samples/sec\taccuracy=0.544044\tlr=0.000002\n",
      "Epoch[44] Batch [5149]\tSpeed: 130.517697 samples/sec\taccuracy=0.543641\tlr=0.000002\n",
      "Epoch[44] Batch [5199]\tSpeed: 128.435549 samples/sec\taccuracy=0.543750\tlr=0.000002\n",
      "Epoch[44] Batch [5249]\tSpeed: 131.228439 samples/sec\taccuracy=0.543619\tlr=0.000002\n",
      "Epoch[44] Batch [5299]\tSpeed: 127.035843 samples/sec\taccuracy=0.543750\tlr=0.000002\n",
      "Epoch[44] Batch [5349]\tSpeed: 127.739350 samples/sec\taccuracy=0.543808\tlr=0.000002\n",
      "Epoch[44] Batch [5399]\tSpeed: 128.550019 samples/sec\taccuracy=0.543565\tlr=0.000002\n",
      "Epoch[44] Batch [5449]\tSpeed: 127.534075 samples/sec\taccuracy=0.543807\tlr=0.000002\n",
      "Epoch[44] Batch [5499]\tSpeed: 128.275161 samples/sec\taccuracy=0.543409\tlr=0.000002\n",
      "Epoch[44] Batch [5549]\tSpeed: 126.687839 samples/sec\taccuracy=0.543356\tlr=0.000002\n",
      "Epoch[44] Batch [5599]\tSpeed: 126.434882 samples/sec\taccuracy=0.543259\tlr=0.000002\n",
      "Epoch[44] Batch [5649]\tSpeed: 128.588366 samples/sec\taccuracy=0.543296\tlr=0.000002\n",
      "Epoch[44] Batch [5699]\tSpeed: 128.748265 samples/sec\taccuracy=0.543684\tlr=0.000002\n",
      "Epoch[44] Batch [5749]\tSpeed: 127.620348 samples/sec\taccuracy=0.544043\tlr=0.000002\n",
      "Epoch[44] Batch [5799]\tSpeed: 126.971784 samples/sec\taccuracy=0.543836\tlr=0.000002\n",
      "Epoch[44] Batch [5849]\tSpeed: 119.366051 samples/sec\taccuracy=0.543547\tlr=0.000002\n",
      "Epoch[44] Batch [5899]\tSpeed: 120.609425 samples/sec\taccuracy=0.543432\tlr=0.000002\n",
      "Epoch[44] Batch [5949]\tSpeed: 128.077244 samples/sec\taccuracy=0.543634\tlr=0.000002\n",
      "Epoch[44] Batch [5999]\tSpeed: 128.495367 samples/sec\taccuracy=0.543625\tlr=0.000002\n",
      "Epoch[44] Batch [6049]\tSpeed: 129.690933 samples/sec\taccuracy=0.543905\tlr=0.000002\n",
      "Epoch[44] Batch [6099]\tSpeed: 127.485698 samples/sec\taccuracy=0.544016\tlr=0.000002\n",
      "Epoch[44] Batch [6149]\tSpeed: 123.769274 samples/sec\taccuracy=0.544106\tlr=0.000002\n",
      "Epoch[44] Batch [6199]\tSpeed: 125.347788 samples/sec\taccuracy=0.544133\tlr=0.000002\n",
      "Epoch[44] Batch [6249]\tSpeed: 128.577240 samples/sec\taccuracy=0.543880\tlr=0.000002\n",
      "Epoch[44] Batch [6299]\tSpeed: 128.464039 samples/sec\taccuracy=0.543552\tlr=0.000002\n",
      "Epoch[44] Batch [6349]\tSpeed: 127.461978 samples/sec\taccuracy=0.543406\tlr=0.000002\n",
      "Epoch[44] Batch [6399]\tSpeed: 128.699633 samples/sec\taccuracy=0.543496\tlr=0.000002\n",
      "Epoch[44] Batch [6449]\tSpeed: 126.459012 samples/sec\taccuracy=0.543450\tlr=0.000002\n",
      "Epoch[44] Batch [6499]\tSpeed: 129.071145 samples/sec\taccuracy=0.543077\tlr=0.000002\n",
      "Epoch[44] Batch [6549]\tSpeed: 127.950293 samples/sec\taccuracy=0.543340\tlr=0.000002\n",
      "Epoch[44] Batch [6599]\tSpeed: 127.223075 samples/sec\taccuracy=0.543542\tlr=0.000002\n",
      "Epoch[44] Batch [6649]\tSpeed: 129.596393 samples/sec\taccuracy=0.543609\tlr=0.000002\n",
      "Epoch[44] Batch [6699]\tSpeed: 127.537614 samples/sec\taccuracy=0.543825\tlr=0.000002\n",
      "Epoch[44] Batch [6749]\tSpeed: 127.159279 samples/sec\taccuracy=0.543926\tlr=0.000002\n",
      "Epoch[44] Batch [6799]\tSpeed: 128.441901 samples/sec\taccuracy=0.543805\tlr=0.000002\n",
      "Epoch[44] Batch [6849]\tSpeed: 126.907520 samples/sec\taccuracy=0.543668\tlr=0.000002\n",
      "Epoch[44] Batch [6899]\tSpeed: 125.923613 samples/sec\taccuracy=0.544076\tlr=0.000002\n",
      "Epoch[44] Batch [6949]\tSpeed: 129.177212 samples/sec\taccuracy=0.544011\tlr=0.000002\n",
      "Epoch[44] Batch [6999]\tSpeed: 126.871378 samples/sec\taccuracy=0.543911\tlr=0.000002\n",
      "Epoch[44] Batch [7049]\tSpeed: 128.374845 samples/sec\taccuracy=0.543635\tlr=0.000002\n",
      "Epoch[44] Batch [7099]\tSpeed: 127.050379 samples/sec\taccuracy=0.543345\tlr=0.000002\n",
      "Epoch[44] Batch [7149]\tSpeed: 126.185904 samples/sec\taccuracy=0.543479\tlr=0.000002\n",
      "Epoch[44] Batch [7199]\tSpeed: 126.534493 samples/sec\taccuracy=0.543507\tlr=0.000002\n",
      "Epoch[44] Batch [7249]\tSpeed: 126.539732 samples/sec\taccuracy=0.543155\tlr=0.000002\n",
      "Epoch[44] Batch [7299]\tSpeed: 127.156985 samples/sec\taccuracy=0.543339\tlr=0.000002\n",
      "Epoch[44] Batch [7349]\tSpeed: 127.934331 samples/sec\taccuracy=0.543350\tlr=0.000002\n",
      "Epoch[44] Batch [7399]\tSpeed: 127.896724 samples/sec\taccuracy=0.543007\tlr=0.000002\n",
      "Epoch[44] Batch [7449]\tSpeed: 127.918782 samples/sec\taccuracy=0.543171\tlr=0.000002\n",
      "Epoch[44] Batch [7499]\tSpeed: 128.470984 samples/sec\taccuracy=0.543083\tlr=0.000002\n",
      "Epoch[44] Batch [7549]\tSpeed: 127.725142 samples/sec\taccuracy=0.543046\tlr=0.000002\n",
      "Epoch[44] Batch [7599]\tSpeed: 125.970774 samples/sec\taccuracy=0.542928\tlr=0.000002\n",
      "Epoch[44] Batch [7649]\tSpeed: 127.867608 samples/sec\taccuracy=0.542908\tlr=0.000002\n",
      "Epoch[44] Batch [7699]\tSpeed: 125.206534 samples/sec\taccuracy=0.542630\tlr=0.000002\n",
      "Epoch[44] Batch [7749]\tSpeed: 128.092616 samples/sec\taccuracy=0.542597\tlr=0.000002\n",
      "Epoch[44] Batch [7799]\tSpeed: 128.165800 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[44] Batch [7849]\tSpeed: 126.886385 samples/sec\taccuracy=0.542580\tlr=0.000002\n",
      "Epoch[44] Batch [7899]\tSpeed: 128.414099 samples/sec\taccuracy=0.542611\tlr=0.000002\n",
      "Epoch[44] Batch [7949]\tSpeed: 126.936565 samples/sec\taccuracy=0.542626\tlr=0.000002\n",
      "Epoch[44] Batch [7999]\tSpeed: 129.966704 samples/sec\taccuracy=0.542766\tlr=0.000002\n",
      "Epoch[44] Batch [8049]\tSpeed: 129.836655 samples/sec\taccuracy=0.542516\tlr=0.000002\n",
      "Epoch[44] Batch [8099]\tSpeed: 125.248680 samples/sec\taccuracy=0.542284\tlr=0.000002\n",
      "Epoch[44] Batch [8149]\tSpeed: 127.145624 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[44] Batch [8199]\tSpeed: 128.585646 samples/sec\taccuracy=0.542744\tlr=0.000002\n",
      "Epoch[44] Batch [8249]\tSpeed: 127.935160 samples/sec\taccuracy=0.543045\tlr=0.000002\n",
      "Epoch[44] Batch [8299]\tSpeed: 126.380090 samples/sec\taccuracy=0.543102\tlr=0.000002\n",
      "Epoch[44] Batch [8349]\tSpeed: 127.706641 samples/sec\taccuracy=0.542919\tlr=0.000002\n",
      "Epoch[44] Batch [8399]\tSpeed: 129.470950 samples/sec\taccuracy=0.542738\tlr=0.000002\n",
      "Epoch[44] Batch [8449]\tSpeed: 128.087912 samples/sec\taccuracy=0.542530\tlr=0.000002\n",
      "Epoch[44] Batch [8499]\tSpeed: 128.504894 samples/sec\taccuracy=0.542559\tlr=0.000002\n",
      "Epoch[44] Batch [8549]\tSpeed: 126.795390 samples/sec\taccuracy=0.542544\tlr=0.000002\n",
      "Epoch[44] Batch [8599]\tSpeed: 128.176748 samples/sec\taccuracy=0.542529\tlr=0.000002\n",
      "Epoch[44] Batch [8649]\tSpeed: 125.967104 samples/sec\taccuracy=0.542514\tlr=0.000002\n",
      "Epoch[44] Batch [8699]\tSpeed: 122.276784 samples/sec\taccuracy=0.542687\tlr=0.000002\n",
      "Epoch[44] Batch [8749]\tSpeed: 120.502543 samples/sec\taccuracy=0.542557\tlr=0.000002\n",
      "Epoch[44] Batch [8799]\tSpeed: 128.525823 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[44] Batch [8849]\tSpeed: 127.123918 samples/sec\taccuracy=0.542655\tlr=0.000002\n",
      "Epoch[44] Batch [8899]\tSpeed: 130.251406 samples/sec\taccuracy=0.542669\tlr=0.000002\n",
      "Epoch[44] Batch [8949]\tSpeed: 127.066333 samples/sec\taccuracy=0.542612\tlr=0.000002\n",
      "Epoch[44] Batch [8999]\tSpeed: 128.150215 samples/sec\taccuracy=0.542264\tlr=0.000002\n",
      "Epoch[44] Batch [9049]\tSpeed: 129.354236 samples/sec\taccuracy=0.542320\tlr=0.000002\n",
      "Epoch[44] Batch [9099]\tSpeed: 125.666373 samples/sec\taccuracy=0.542280\tlr=0.000002\n",
      "Epoch[44] Batch [9149]\tSpeed: 128.897935 samples/sec\taccuracy=0.542077\tlr=0.000002\n",
      "Epoch[44] Batch [9199]\tSpeed: 129.872626 samples/sec\taccuracy=0.541848\tlr=0.000002\n",
      "Epoch[44] Batch [9249]\tSpeed: 128.803370 samples/sec\taccuracy=0.542095\tlr=0.000002\n",
      "Epoch[44] Batch [9299]\tSpeed: 127.872773 samples/sec\taccuracy=0.542030\tlr=0.000002\n",
      "Epoch[44] Batch [9349]\tSpeed: 125.326280 samples/sec\taccuracy=0.542019\tlr=0.000002\n",
      "Epoch[44] Batch [9399]\tSpeed: 127.823408 samples/sec\taccuracy=0.542088\tlr=0.000002\n",
      "Epoch[44] Batch [9449]\tSpeed: 126.436140 samples/sec\taccuracy=0.542024\tlr=0.000002\n",
      "Epoch[44] Batch [9499]\tSpeed: 126.954422 samples/sec\taccuracy=0.542066\tlr=0.000002\n",
      "Epoch[44] Batch [9549]\tSpeed: 127.550054 samples/sec\taccuracy=0.542055\tlr=0.000002\n",
      "Epoch[44] Batch [9599]\tSpeed: 127.562875 samples/sec\taccuracy=0.542122\tlr=0.000002\n",
      "Epoch[44] Batch [9649]\tSpeed: 129.475576 samples/sec\taccuracy=0.542228\tlr=0.000002\n",
      "Epoch[44] Batch [9699]\tSpeed: 128.289942 samples/sec\taccuracy=0.542113\tlr=0.000002\n",
      "Epoch[44] Batch [9749]\tSpeed: 127.431811 samples/sec\taccuracy=0.542295\tlr=0.000002\n",
      "Epoch[44] Batch [9799]\tSpeed: 128.902075 samples/sec\taccuracy=0.542232\tlr=0.000002\n",
      "Epoch[44] Batch [9849]\tSpeed: 124.421607 samples/sec\taccuracy=0.542170\tlr=0.000002\n",
      "Epoch[44] Batch [9899]\tSpeed: 126.667791 samples/sec\taccuracy=0.542374\tlr=0.000002\n",
      "Epoch[44] Batch [9949]\tSpeed: 127.512353 samples/sec\taccuracy=0.542450\tlr=0.000002\n",
      "Epoch[44] Batch [9999]\tSpeed: 128.167465 samples/sec\taccuracy=0.542250\tlr=0.000002\n",
      "Epoch[44] Batch [10049]\tSpeed: 129.489597 samples/sec\taccuracy=0.542425\tlr=0.000002\n",
      "Epoch[44] Batch [10099]\tSpeed: 128.129281 samples/sec\taccuracy=0.542475\tlr=0.000002\n",
      "Epoch[44] Batch [10149]\tSpeed: 128.061524 samples/sec\taccuracy=0.542685\tlr=0.000002\n",
      "Epoch[44] Batch [10199]\tSpeed: 130.012923 samples/sec\taccuracy=0.542525\tlr=0.000002\n",
      "Epoch[44] Batch [10249]\tSpeed: 126.095854 samples/sec\taccuracy=0.542524\tlr=0.000002\n",
      "Epoch[44] Batch [10299]\tSpeed: 126.977684 samples/sec\taccuracy=0.542512\tlr=0.000002\n",
      "Epoch[44] Batch [10349]\tSpeed: 128.606523 samples/sec\taccuracy=0.542560\tlr=0.000002\n",
      "Epoch[44] Batch [10399]\tSpeed: 128.724478 samples/sec\taccuracy=0.542320\tlr=0.000002\n",
      "Epoch[44] Batch [10449]\tSpeed: 129.874285 samples/sec\taccuracy=0.542261\tlr=0.000002\n",
      "Epoch[44] Batch [10499]\tSpeed: 126.775691 samples/sec\taccuracy=0.542345\tlr=0.000002\n",
      "Epoch[44] Batch [10549]\tSpeed: 127.482220 samples/sec\taccuracy=0.542299\tlr=0.000002\n",
      "Epoch[44] Batch [10599]\tSpeed: 127.639563 samples/sec\taccuracy=0.542193\tlr=0.000002\n",
      "Epoch[44] Batch [10649]\tSpeed: 129.000672 samples/sec\taccuracy=0.542113\tlr=0.000002\n",
      "Epoch[44] Batch [10699]\tSpeed: 125.948314 samples/sec\taccuracy=0.541998\tlr=0.000002\n",
      "Epoch[44] Batch [10749]\tSpeed: 124.608628 samples/sec\taccuracy=0.541977\tlr=0.000002\n",
      "Epoch[44] Batch [10799]\tSpeed: 127.181681 samples/sec\taccuracy=0.541771\tlr=0.000002\n",
      "Epoch[44] Batch [10849]\tSpeed: 127.799738 samples/sec\taccuracy=0.541809\tlr=0.000002\n",
      "Epoch[44] Batch [10899]\tSpeed: 126.842966 samples/sec\taccuracy=0.541732\tlr=0.000002\n",
      "Epoch[44] Batch [10949]\tSpeed: 129.961147 samples/sec\taccuracy=0.541849\tlr=0.000002\n",
      "Epoch[44] Batch [10999]\tSpeed: 129.678032 samples/sec\taccuracy=0.541807\tlr=0.000002\n",
      "Epoch[44] Batch [11049]\tSpeed: 128.231502 samples/sec\taccuracy=0.541708\tlr=0.000002\n",
      "Epoch[44] Batch [11099]\tSpeed: 128.300577 samples/sec\taccuracy=0.541599\tlr=0.000002\n",
      "Epoch[44] Batch [11149]\tSpeed: 129.635587 samples/sec\taccuracy=0.541738\tlr=0.000002\n",
      "Epoch[44] Batch [11199]\tSpeed: 127.360922 samples/sec\taccuracy=0.541797\tlr=0.000002\n",
      "Epoch[44] Batch [11249]\tSpeed: 127.922937 samples/sec\taccuracy=0.541644\tlr=0.000002\n",
      "Epoch[44] Batch [11299]\tSpeed: 128.474526 samples/sec\taccuracy=0.541759\tlr=0.000002\n",
      "Epoch[44] Batch [11349]\tSpeed: 138.883204 samples/sec\taccuracy=0.541685\tlr=0.000002\n",
      "[Epoch 44] training: accuracy=0.541689\n",
      "[Epoch 44] speed: 127 samples/sec\ttime cost: 712.015935\n",
      "[Epoch 44] validation: top1=0.543554 top5=0.877812\n",
      "Epoch[45] Batch [49]\tSpeed: 107.265955 samples/sec\taccuracy=0.537500\tlr=0.000002\n",
      "Epoch[45] Batch [99]\tSpeed: 129.283554 samples/sec\taccuracy=0.545000\tlr=0.000002\n",
      "Epoch[45] Batch [149]\tSpeed: 128.399829 samples/sec\taccuracy=0.545833\tlr=0.000002\n",
      "Epoch[45] Batch [199]\tSpeed: 127.864031 samples/sec\taccuracy=0.546250\tlr=0.000002\n",
      "Epoch[45] Batch [249]\tSpeed: 127.159270 samples/sec\taccuracy=0.548500\tlr=0.000002\n",
      "Epoch[45] Batch [299]\tSpeed: 126.402180 samples/sec\taccuracy=0.548333\tlr=0.000002\n",
      "Epoch[45] Batch [349]\tSpeed: 128.439404 samples/sec\taccuracy=0.550714\tlr=0.000002\n",
      "Epoch[45] Batch [399]\tSpeed: 127.834404 samples/sec\taccuracy=0.544687\tlr=0.000002\n",
      "Epoch[45] Batch [449]\tSpeed: 128.882528 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[45] Batch [499]\tSpeed: 129.316189 samples/sec\taccuracy=0.540000\tlr=0.000002\n",
      "Epoch[45] Batch [549]\tSpeed: 127.136827 samples/sec\taccuracy=0.543864\tlr=0.000002\n",
      "Epoch[45] Batch [599]\tSpeed: 127.370997 samples/sec\taccuracy=0.547083\tlr=0.000002\n",
      "Epoch[45] Batch [649]\tSpeed: 128.921548 samples/sec\taccuracy=0.549423\tlr=0.000002\n",
      "Epoch[45] Batch [699]\tSpeed: 126.690068 samples/sec\taccuracy=0.550357\tlr=0.000002\n",
      "Epoch[45] Batch [749]\tSpeed: 126.125041 samples/sec\taccuracy=0.550500\tlr=0.000002\n",
      "Epoch[45] Batch [799]\tSpeed: 128.451971 samples/sec\taccuracy=0.549375\tlr=0.000002\n",
      "Epoch[45] Batch [849]\tSpeed: 127.687814 samples/sec\taccuracy=0.549265\tlr=0.000002\n",
      "Epoch[45] Batch [899]\tSpeed: 130.744567 samples/sec\taccuracy=0.550833\tlr=0.000002\n",
      "Epoch[45] Batch [949]\tSpeed: 125.228646 samples/sec\taccuracy=0.550789\tlr=0.000002\n",
      "Epoch[45] Batch [999]\tSpeed: 128.877627 samples/sec\taccuracy=0.550000\tlr=0.000002\n",
      "Epoch[45] Batch [1049]\tSpeed: 126.513892 samples/sec\taccuracy=0.551190\tlr=0.000002\n",
      "Epoch[45] Batch [1099]\tSpeed: 127.855749 samples/sec\taccuracy=0.551477\tlr=0.000002\n",
      "Epoch[45] Batch [1149]\tSpeed: 127.333151 samples/sec\taccuracy=0.552065\tlr=0.000002\n",
      "Epoch[45] Batch [1199]\tSpeed: 127.877422 samples/sec\taccuracy=0.552500\tlr=0.000002\n",
      "Epoch[45] Batch [1249]\tSpeed: 128.782805 samples/sec\taccuracy=0.553000\tlr=0.000002\n",
      "Epoch[45] Batch [1299]\tSpeed: 127.541230 samples/sec\taccuracy=0.551154\tlr=0.000002\n",
      "Epoch[45] Batch [1349]\tSpeed: 127.419858 samples/sec\taccuracy=0.551111\tlr=0.000002\n",
      "Epoch[45] Batch [1399]\tSpeed: 127.566861 samples/sec\taccuracy=0.550804\tlr=0.000002\n",
      "Epoch[45] Batch [1449]\tSpeed: 128.784328 samples/sec\taccuracy=0.550948\tlr=0.000002\n",
      "Epoch[45] Batch [1499]\tSpeed: 125.633182 samples/sec\taccuracy=0.550250\tlr=0.000002\n",
      "Epoch[45] Batch [1549]\tSpeed: 127.150057 samples/sec\taccuracy=0.551129\tlr=0.000002\n",
      "Epoch[45] Batch [1599]\tSpeed: 128.703928 samples/sec\taccuracy=0.549844\tlr=0.000002\n",
      "Epoch[45] Batch [1649]\tSpeed: 128.267521 samples/sec\taccuracy=0.549242\tlr=0.000002\n",
      "Epoch[45] Batch [1699]\tSpeed: 128.637535 samples/sec\taccuracy=0.548603\tlr=0.000002\n",
      "Epoch[45] Batch [1749]\tSpeed: 127.207650 samples/sec\taccuracy=0.548214\tlr=0.000002\n",
      "Epoch[45] Batch [1799]\tSpeed: 127.959144 samples/sec\taccuracy=0.547778\tlr=0.000002\n",
      "Epoch[45] Batch [1849]\tSpeed: 128.803884 samples/sec\taccuracy=0.547973\tlr=0.000002\n",
      "Epoch[45] Batch [1899]\tSpeed: 127.791940 samples/sec\taccuracy=0.546579\tlr=0.000002\n",
      "Epoch[45] Batch [1949]\tSpeed: 129.080023 samples/sec\taccuracy=0.546667\tlr=0.000002\n",
      "Epoch[45] Batch [1999]\tSpeed: 127.951161 samples/sec\taccuracy=0.546250\tlr=0.000002\n",
      "Epoch[45] Batch [2049]\tSpeed: 127.722147 samples/sec\taccuracy=0.546524\tlr=0.000002\n",
      "Epoch[45] Batch [2099]\tSpeed: 119.257563 samples/sec\taccuracy=0.546250\tlr=0.000002\n",
      "Epoch[45] Batch [2149]\tSpeed: 122.347879 samples/sec\taccuracy=0.546279\tlr=0.000002\n",
      "Epoch[45] Batch [2199]\tSpeed: 126.586754 samples/sec\taccuracy=0.546136\tlr=0.000002\n",
      "Epoch[45] Batch [2249]\tSpeed: 127.323864 samples/sec\taccuracy=0.546278\tlr=0.000002\n",
      "Epoch[45] Batch [2299]\tSpeed: 128.364404 samples/sec\taccuracy=0.546630\tlr=0.000002\n",
      "Epoch[45] Batch [2349]\tSpeed: 128.530697 samples/sec\taccuracy=0.546436\tlr=0.000002\n",
      "Epoch[45] Batch [2399]\tSpeed: 124.507997 samples/sec\taccuracy=0.546302\tlr=0.000002\n",
      "Epoch[45] Batch [2449]\tSpeed: 128.398954 samples/sec\taccuracy=0.545765\tlr=0.000002\n",
      "Epoch[45] Batch [2499]\tSpeed: 128.156049 samples/sec\taccuracy=0.544550\tlr=0.000002\n",
      "Epoch[45] Batch [2549]\tSpeed: 126.968680 samples/sec\taccuracy=0.543725\tlr=0.000002\n",
      "Epoch[45] Batch [2599]\tSpeed: 127.024147 samples/sec\taccuracy=0.543221\tlr=0.000002\n",
      "Epoch[45] Batch [2649]\tSpeed: 128.466538 samples/sec\taccuracy=0.543019\tlr=0.000002\n",
      "Epoch[45] Batch [2699]\tSpeed: 129.317814 samples/sec\taccuracy=0.543102\tlr=0.000002\n",
      "Epoch[45] Batch [2749]\tSpeed: 127.687241 samples/sec\taccuracy=0.542864\tlr=0.000002\n",
      "Epoch[45] Batch [2799]\tSpeed: 129.503910 samples/sec\taccuracy=0.542946\tlr=0.000002\n",
      "Epoch[45] Batch [2849]\tSpeed: 126.364669 samples/sec\taccuracy=0.543158\tlr=0.000002\n",
      "Epoch[45] Batch [2899]\tSpeed: 126.299250 samples/sec\taccuracy=0.543405\tlr=0.000002\n",
      "Epoch[45] Batch [2949]\tSpeed: 129.440464 samples/sec\taccuracy=0.543347\tlr=0.000002\n",
      "Epoch[45] Batch [2999]\tSpeed: 127.818539 samples/sec\taccuracy=0.543375\tlr=0.000002\n",
      "Epoch[45] Batch [3049]\tSpeed: 125.586565 samples/sec\taccuracy=0.543770\tlr=0.000002\n",
      "Epoch[45] Batch [3099]\tSpeed: 127.336910 samples/sec\taccuracy=0.543266\tlr=0.000002\n",
      "Epoch[45] Batch [3149]\tSpeed: 128.274945 samples/sec\taccuracy=0.543016\tlr=0.000002\n",
      "Epoch[45] Batch [3199]\tSpeed: 126.243844 samples/sec\taccuracy=0.543125\tlr=0.000002\n",
      "Epoch[45] Batch [3249]\tSpeed: 126.041383 samples/sec\taccuracy=0.543192\tlr=0.000002\n",
      "Epoch[45] Batch [3299]\tSpeed: 127.744369 samples/sec\taccuracy=0.543258\tlr=0.000002\n",
      "Epoch[45] Batch [3349]\tSpeed: 127.300513 samples/sec\taccuracy=0.543358\tlr=0.000002\n",
      "Epoch[45] Batch [3399]\tSpeed: 128.024565 samples/sec\taccuracy=0.542684\tlr=0.000002\n",
      "Epoch[45] Batch [3449]\tSpeed: 125.622976 samples/sec\taccuracy=0.543007\tlr=0.000002\n",
      "Epoch[45] Batch [3499]\tSpeed: 127.423003 samples/sec\taccuracy=0.543071\tlr=0.000002\n",
      "Epoch[45] Batch [3549]\tSpeed: 130.504286 samples/sec\taccuracy=0.543345\tlr=0.000002\n",
      "Epoch[45] Batch [3599]\tSpeed: 124.882297 samples/sec\taccuracy=0.543438\tlr=0.000002\n",
      "Epoch[45] Batch [3649]\tSpeed: 124.351925 samples/sec\taccuracy=0.543664\tlr=0.000002\n",
      "Epoch[45] Batch [3699]\tSpeed: 128.526453 samples/sec\taccuracy=0.543885\tlr=0.000002\n",
      "Epoch[45] Batch [3749]\tSpeed: 129.041998 samples/sec\taccuracy=0.543600\tlr=0.000002\n",
      "Epoch[45] Batch [3799]\tSpeed: 129.798685 samples/sec\taccuracy=0.543224\tlr=0.000002\n",
      "Epoch[45] Batch [3849]\tSpeed: 124.642131 samples/sec\taccuracy=0.543799\tlr=0.000002\n",
      "Epoch[45] Batch [3899]\tSpeed: 127.446563 samples/sec\taccuracy=0.543718\tlr=0.000002\n",
      "Epoch[45] Batch [3949]\tSpeed: 127.954167 samples/sec\taccuracy=0.543734\tlr=0.000002\n",
      "Epoch[45] Batch [3999]\tSpeed: 125.900376 samples/sec\taccuracy=0.543500\tlr=0.000002\n",
      "Epoch[45] Batch [4049]\tSpeed: 128.742920 samples/sec\taccuracy=0.543673\tlr=0.000002\n",
      "Epoch[45] Batch [4099]\tSpeed: 126.470299 samples/sec\taccuracy=0.543720\tlr=0.000002\n",
      "Epoch[45] Batch [4149]\tSpeed: 128.147377 samples/sec\taccuracy=0.543765\tlr=0.000002\n",
      "Epoch[45] Batch [4199]\tSpeed: 126.421706 samples/sec\taccuracy=0.543601\tlr=0.000002\n",
      "Epoch[45] Batch [4249]\tSpeed: 127.703987 samples/sec\taccuracy=0.543618\tlr=0.000002\n",
      "Epoch[45] Batch [4299]\tSpeed: 127.407937 samples/sec\taccuracy=0.543459\tlr=0.000002\n",
      "Epoch[45] Batch [4349]\tSpeed: 127.891089 samples/sec\taccuracy=0.543420\tlr=0.000002\n",
      "Epoch[45] Batch [4399]\tSpeed: 130.320043 samples/sec\taccuracy=0.543608\tlr=0.000002\n",
      "Epoch[45] Batch [4449]\tSpeed: 128.698715 samples/sec\taccuracy=0.543680\tlr=0.000002\n",
      "Epoch[45] Batch [4499]\tSpeed: 127.896188 samples/sec\taccuracy=0.543444\tlr=0.000002\n",
      "Epoch[45] Batch [4549]\tSpeed: 129.430568 samples/sec\taccuracy=0.542995\tlr=0.000002\n",
      "Epoch[45] Batch [4599]\tSpeed: 126.046705 samples/sec\taccuracy=0.542799\tlr=0.000002\n",
      "Epoch[45] Batch [4649]\tSpeed: 128.357794 samples/sec\taccuracy=0.542715\tlr=0.000002\n",
      "Epoch[45] Batch [4699]\tSpeed: 125.892904 samples/sec\taccuracy=0.542793\tlr=0.000002\n",
      "Epoch[45] Batch [4749]\tSpeed: 124.665786 samples/sec\taccuracy=0.542684\tlr=0.000002\n",
      "Epoch[45] Batch [4799]\tSpeed: 127.694228 samples/sec\taccuracy=0.542839\tlr=0.000002\n",
      "Epoch[45] Batch [4849]\tSpeed: 128.458895 samples/sec\taccuracy=0.543170\tlr=0.000002\n",
      "Epoch[45] Batch [4899]\tSpeed: 125.443629 samples/sec\taccuracy=0.543036\tlr=0.000002\n",
      "Epoch[45] Batch [4949]\tSpeed: 123.100077 samples/sec\taccuracy=0.542955\tlr=0.000002\n",
      "Epoch[45] Batch [4999]\tSpeed: 121.004344 samples/sec\taccuracy=0.543150\tlr=0.000002\n",
      "Epoch[45] Batch [5049]\tSpeed: 125.612827 samples/sec\taccuracy=0.543218\tlr=0.000002\n",
      "Epoch[45] Batch [5099]\tSpeed: 128.769718 samples/sec\taccuracy=0.542990\tlr=0.000002\n",
      "Epoch[45] Batch [5149]\tSpeed: 124.883236 samples/sec\taccuracy=0.543131\tlr=0.000002\n",
      "Epoch[45] Batch [5199]\tSpeed: 127.454280 samples/sec\taccuracy=0.543269\tlr=0.000002\n",
      "Epoch[45] Batch [5249]\tSpeed: 127.304078 samples/sec\taccuracy=0.543048\tlr=0.000002\n",
      "Epoch[45] Batch [5299]\tSpeed: 127.800750 samples/sec\taccuracy=0.542877\tlr=0.000002\n",
      "Epoch[45] Batch [5349]\tSpeed: 130.682006 samples/sec\taccuracy=0.543154\tlr=0.000002\n",
      "Epoch[45] Batch [5399]\tSpeed: 129.511388 samples/sec\taccuracy=0.543079\tlr=0.000002\n",
      "Epoch[45] Batch [5449]\tSpeed: 126.143201 samples/sec\taccuracy=0.542638\tlr=0.000002\n",
      "Epoch[45] Batch [5499]\tSpeed: 128.796231 samples/sec\taccuracy=0.542682\tlr=0.000002\n",
      "Epoch[45] Batch [5549]\tSpeed: 127.063090 samples/sec\taccuracy=0.542680\tlr=0.000002\n",
      "Epoch[45] Batch [5599]\tSpeed: 127.479828 samples/sec\taccuracy=0.542790\tlr=0.000002\n",
      "Epoch[45] Batch [5649]\tSpeed: 127.256328 samples/sec\taccuracy=0.542721\tlr=0.000002\n",
      "Epoch[45] Batch [5699]\tSpeed: 128.954982 samples/sec\taccuracy=0.542675\tlr=0.000002\n",
      "Epoch[45] Batch [5749]\tSpeed: 126.607503 samples/sec\taccuracy=0.542783\tlr=0.000002\n",
      "Epoch[45] Batch [5799]\tSpeed: 128.386103 samples/sec\taccuracy=0.542866\tlr=0.000002\n",
      "Epoch[45] Batch [5849]\tSpeed: 128.127040 samples/sec\taccuracy=0.542906\tlr=0.000002\n",
      "Epoch[45] Batch [5899]\tSpeed: 126.675844 samples/sec\taccuracy=0.542987\tlr=0.000002\n",
      "Epoch[45] Batch [5949]\tSpeed: 125.816214 samples/sec\taccuracy=0.543109\tlr=0.000002\n",
      "Epoch[45] Batch [5999]\tSpeed: 127.087682 samples/sec\taccuracy=0.543375\tlr=0.000002\n",
      "Epoch[45] Batch [6049]\tSpeed: 128.596152 samples/sec\taccuracy=0.543409\tlr=0.000002\n",
      "Epoch[45] Batch [6099]\tSpeed: 126.573250 samples/sec\taccuracy=0.543238\tlr=0.000002\n",
      "Epoch[45] Batch [6149]\tSpeed: 125.451921 samples/sec\taccuracy=0.543171\tlr=0.000002\n",
      "Epoch[45] Batch [6199]\tSpeed: 125.087044 samples/sec\taccuracy=0.543206\tlr=0.000002\n",
      "Epoch[45] Batch [6249]\tSpeed: 127.167279 samples/sec\taccuracy=0.543020\tlr=0.000002\n",
      "Epoch[45] Batch [6299]\tSpeed: 130.006032 samples/sec\taccuracy=0.542937\tlr=0.000002\n",
      "Epoch[45] Batch [6349]\tSpeed: 126.066367 samples/sec\taccuracy=0.543071\tlr=0.000002\n",
      "Epoch[45] Batch [6399]\tSpeed: 125.507958 samples/sec\taccuracy=0.543105\tlr=0.000002\n",
      "Epoch[45] Batch [6449]\tSpeed: 126.732434 samples/sec\taccuracy=0.542868\tlr=0.000002\n",
      "Epoch[45] Batch [6499]\tSpeed: 127.515203 samples/sec\taccuracy=0.542519\tlr=0.000002\n",
      "Epoch[45] Batch [6549]\tSpeed: 127.836751 samples/sec\taccuracy=0.542634\tlr=0.000002\n",
      "Epoch[45] Batch [6599]\tSpeed: 127.167491 samples/sec\taccuracy=0.542765\tlr=0.000002\n",
      "Epoch[45] Batch [6649]\tSpeed: 128.852703 samples/sec\taccuracy=0.542970\tlr=0.000002\n",
      "Epoch[45] Batch [6699]\tSpeed: 126.256954 samples/sec\taccuracy=0.542985\tlr=0.000002\n",
      "Epoch[45] Batch [6749]\tSpeed: 125.903598 samples/sec\taccuracy=0.542981\tlr=0.000002\n",
      "Epoch[45] Batch [6799]\tSpeed: 126.881692 samples/sec\taccuracy=0.543199\tlr=0.000002\n",
      "Epoch[45] Batch [6849]\tSpeed: 127.045828 samples/sec\taccuracy=0.543047\tlr=0.000002\n",
      "Epoch[45] Batch [6899]\tSpeed: 125.302645 samples/sec\taccuracy=0.542699\tlr=0.000002\n",
      "Epoch[45] Batch [6949]\tSpeed: 127.791668 samples/sec\taccuracy=0.542590\tlr=0.000002\n",
      "Epoch[45] Batch [6999]\tSpeed: 127.019781 samples/sec\taccuracy=0.542714\tlr=0.000002\n",
      "Epoch[45] Batch [7049]\tSpeed: 127.583178 samples/sec\taccuracy=0.542677\tlr=0.000002\n",
      "Epoch[45] Batch [7099]\tSpeed: 128.894974 samples/sec\taccuracy=0.542852\tlr=0.000002\n",
      "Epoch[45] Batch [7149]\tSpeed: 126.849699 samples/sec\taccuracy=0.542815\tlr=0.000002\n",
      "Epoch[45] Batch [7199]\tSpeed: 127.646875 samples/sec\taccuracy=0.542691\tlr=0.000002\n",
      "Epoch[45] Batch [7249]\tSpeed: 127.744972 samples/sec\taccuracy=0.542672\tlr=0.000002\n",
      "Epoch[45] Batch [7299]\tSpeed: 128.357441 samples/sec\taccuracy=0.542637\tlr=0.000002\n",
      "Epoch[45] Batch [7349]\tSpeed: 127.928360 samples/sec\taccuracy=0.542908\tlr=0.000002\n",
      "Epoch[45] Batch [7399]\tSpeed: 128.222838 samples/sec\taccuracy=0.542821\tlr=0.000002\n",
      "Epoch[45] Batch [7449]\tSpeed: 126.413038 samples/sec\taccuracy=0.542936\tlr=0.000002\n",
      "Epoch[45] Batch [7499]\tSpeed: 129.983893 samples/sec\taccuracy=0.543033\tlr=0.000002\n",
      "Epoch[45] Batch [7549]\tSpeed: 127.942799 samples/sec\taccuracy=0.542914\tlr=0.000002\n",
      "Epoch[45] Batch [7599]\tSpeed: 127.437619 samples/sec\taccuracy=0.543158\tlr=0.000002\n",
      "Epoch[45] Batch [7649]\tSpeed: 127.204352 samples/sec\taccuracy=0.543039\tlr=0.000002\n",
      "Epoch[45] Batch [7699]\tSpeed: 127.409040 samples/sec\taccuracy=0.542776\tlr=0.000002\n",
      "Epoch[45] Batch [7749]\tSpeed: 126.738551 samples/sec\taccuracy=0.542710\tlr=0.000002\n",
      "Epoch[45] Batch [7799]\tSpeed: 126.546681 samples/sec\taccuracy=0.542564\tlr=0.000002\n",
      "Epoch[45] Batch [7849]\tSpeed: 117.948680 samples/sec\taccuracy=0.542341\tlr=0.000002\n",
      "Epoch[45] Batch [7899]\tSpeed: 128.200313 samples/sec\taccuracy=0.542358\tlr=0.000002\n",
      "Epoch[45] Batch [7949]\tSpeed: 127.259716 samples/sec\taccuracy=0.542280\tlr=0.000002\n",
      "Epoch[45] Batch [7999]\tSpeed: 127.046011 samples/sec\taccuracy=0.542188\tlr=0.000002\n",
      "Epoch[45] Batch [8049]\tSpeed: 128.970992 samples/sec\taccuracy=0.542360\tlr=0.000002\n",
      "Epoch[45] Batch [8099]\tSpeed: 127.649828 samples/sec\taccuracy=0.542346\tlr=0.000002\n",
      "Epoch[45] Batch [8149]\tSpeed: 126.134201 samples/sec\taccuracy=0.542362\tlr=0.000002\n",
      "Epoch[45] Batch [8199]\tSpeed: 127.467943 samples/sec\taccuracy=0.542302\tlr=0.000002\n",
      "Epoch[45] Batch [8249]\tSpeed: 127.508990 samples/sec\taccuracy=0.542545\tlr=0.000002\n",
      "Epoch[45] Batch [8299]\tSpeed: 127.290488 samples/sec\taccuracy=0.542154\tlr=0.000002\n",
      "Epoch[45] Batch [8349]\tSpeed: 126.517050 samples/sec\taccuracy=0.542066\tlr=0.000002\n",
      "Epoch[45] Batch [8399]\tSpeed: 128.661053 samples/sec\taccuracy=0.541875\tlr=0.000002\n",
      "Epoch[45] Batch [8449]\tSpeed: 127.432546 samples/sec\taccuracy=0.541760\tlr=0.000002\n",
      "Epoch[45] Batch [8499]\tSpeed: 127.335190 samples/sec\taccuracy=0.541662\tlr=0.000002\n",
      "Epoch[45] Batch [8549]\tSpeed: 129.276700 samples/sec\taccuracy=0.541535\tlr=0.000002\n",
      "Epoch[45] Batch [8599]\tSpeed: 126.389849 samples/sec\taccuracy=0.541483\tlr=0.000002\n",
      "Epoch[45] Batch [8649]\tSpeed: 125.832577 samples/sec\taccuracy=0.541647\tlr=0.000002\n",
      "Epoch[45] Batch [8699]\tSpeed: 129.417838 samples/sec\taccuracy=0.541408\tlr=0.000002\n",
      "Epoch[45] Batch [8749]\tSpeed: 124.588937 samples/sec\taccuracy=0.541343\tlr=0.000002\n",
      "Epoch[45] Batch [8799]\tSpeed: 127.589592 samples/sec\taccuracy=0.541491\tlr=0.000002\n",
      "Epoch[45] Batch [8849]\tSpeed: 126.113825 samples/sec\taccuracy=0.541653\tlr=0.000002\n",
      "Epoch[45] Batch [8899]\tSpeed: 131.785164 samples/sec\taccuracy=0.541924\tlr=0.000002\n",
      "Epoch[45] Batch [8949]\tSpeed: 126.351669 samples/sec\taccuracy=0.542137\tlr=0.000002\n",
      "Epoch[45] Batch [8999]\tSpeed: 127.210226 samples/sec\taccuracy=0.542042\tlr=0.000002\n",
      "Epoch[45] Batch [9049]\tSpeed: 128.204604 samples/sec\taccuracy=0.542058\tlr=0.000002\n",
      "Epoch[45] Batch [9099]\tSpeed: 126.731496 samples/sec\taccuracy=0.542129\tlr=0.000002\n",
      "Epoch[45] Batch [9149]\tSpeed: 127.412059 samples/sec\taccuracy=0.542077\tlr=0.000002\n",
      "Epoch[45] Batch [9199]\tSpeed: 125.864504 samples/sec\taccuracy=0.541916\tlr=0.000002\n",
      "Epoch[45] Batch [9249]\tSpeed: 124.245883 samples/sec\taccuracy=0.541851\tlr=0.000002\n",
      "Epoch[45] Batch [9299]\tSpeed: 126.018917 samples/sec\taccuracy=0.541909\tlr=0.000002\n",
      "Epoch[45] Batch [9349]\tSpeed: 126.971005 samples/sec\taccuracy=0.541979\tlr=0.000002\n",
      "Epoch[45] Batch [9399]\tSpeed: 127.205229 samples/sec\taccuracy=0.541995\tlr=0.000002\n",
      "Epoch[45] Batch [9449]\tSpeed: 128.001836 samples/sec\taccuracy=0.542024\tlr=0.000002\n",
      "Epoch[45] Batch [9499]\tSpeed: 128.745153 samples/sec\taccuracy=0.541908\tlr=0.000002\n",
      "Epoch[45] Batch [9549]\tSpeed: 129.435231 samples/sec\taccuracy=0.541780\tlr=0.000002\n",
      "Epoch[45] Batch [9599]\tSpeed: 127.743377 samples/sec\taccuracy=0.541745\tlr=0.000002\n",
      "Epoch[45] Batch [9649]\tSpeed: 125.808695 samples/sec\taccuracy=0.541762\tlr=0.000002\n",
      "Epoch[45] Batch [9699]\tSpeed: 126.641095 samples/sec\taccuracy=0.541714\tlr=0.000002\n",
      "Epoch[45] Batch [9749]\tSpeed: 125.959671 samples/sec\taccuracy=0.541782\tlr=0.000002\n",
      "Epoch[45] Batch [9799]\tSpeed: 127.568704 samples/sec\taccuracy=0.541824\tlr=0.000002\n",
      "Epoch[45] Batch [9849]\tSpeed: 128.855356 samples/sec\taccuracy=0.541726\tlr=0.000002\n",
      "Epoch[46] Batch [5599]\tSpeed: 129.361777 samples/sec\taccuracy=0.539286\tlr=0.000002\n",
      "Epoch[46] Batch [5649]\tSpeed: 128.355673 samples/sec\taccuracy=0.539292\tlr=0.000002\n",
      "Epoch[46] Batch [5699]\tSpeed: 128.857464 samples/sec\taccuracy=0.539254\tlr=0.000002\n",
      "Epoch[46] Batch [5749]\tSpeed: 131.346514 samples/sec\taccuracy=0.539239\tlr=0.000002\n",
      "Epoch[46] Batch [5799]\tSpeed: 128.778416 samples/sec\taccuracy=0.539095\tlr=0.000002\n",
      "Epoch[46] Batch [5849]\tSpeed: 126.089125 samples/sec\taccuracy=0.538932\tlr=0.000002\n",
      "Epoch[46] Batch [5899]\tSpeed: 127.287243 samples/sec\taccuracy=0.538665\tlr=0.000002\n",
      "Epoch[46] Batch [5949]\tSpeed: 129.040251 samples/sec\taccuracy=0.538992\tlr=0.000002\n",
      "Epoch[46] Batch [5999]\tSpeed: 127.713728 samples/sec\taccuracy=0.539188\tlr=0.000002\n",
      "Epoch[46] Batch [6049]\tSpeed: 127.025696 samples/sec\taccuracy=0.539091\tlr=0.000002\n",
      "Epoch[46] Batch [6099]\tSpeed: 127.718978 samples/sec\taccuracy=0.539406\tlr=0.000002\n",
      "Epoch[46] Batch [6149]\tSpeed: 126.528729 samples/sec\taccuracy=0.539512\tlr=0.000002\n",
      "Epoch[46] Batch [6199]\tSpeed: 127.749291 samples/sec\taccuracy=0.539819\tlr=0.000002\n",
      "Epoch[46] Batch [6249]\tSpeed: 123.490394 samples/sec\taccuracy=0.539760\tlr=0.000002\n",
      "Epoch[46] Batch [6299]\tSpeed: 118.686020 samples/sec\taccuracy=0.539464\tlr=0.000002\n",
      "Epoch[46] Batch [6349]\tSpeed: 119.598581 samples/sec\taccuracy=0.539331\tlr=0.000002\n",
      "Epoch[46] Batch [6399]\tSpeed: 121.306601 samples/sec\taccuracy=0.539375\tlr=0.000002\n",
      "Epoch[46] Batch [6449]\tSpeed: 118.726242 samples/sec\taccuracy=0.539477\tlr=0.000002\n",
      "Epoch[46] Batch [6499]\tSpeed: 117.909148 samples/sec\taccuracy=0.539538\tlr=0.000002\n",
      "Epoch[46] Batch [6549]\tSpeed: 122.882100 samples/sec\taccuracy=0.539752\tlr=0.000002\n",
      "Epoch[46] Batch [6599]\tSpeed: 125.725183 samples/sec\taccuracy=0.539697\tlr=0.000002\n",
      "Epoch[46] Batch [6649]\tSpeed: 129.027022 samples/sec\taccuracy=0.539793\tlr=0.000002\n",
      "Epoch[46] Batch [6699]\tSpeed: 128.817028 samples/sec\taccuracy=0.539888\tlr=0.000002\n",
      "Epoch[46] Batch [6749]\tSpeed: 128.906571 samples/sec\taccuracy=0.539796\tlr=0.000002\n",
      "Epoch[46] Batch [6799]\tSpeed: 126.303633 samples/sec\taccuracy=0.539853\tlr=0.000002\n",
      "Epoch[46] Batch [6849]\tSpeed: 124.141078 samples/sec\taccuracy=0.539745\tlr=0.000002\n",
      "Epoch[46] Batch [6899]\tSpeed: 118.647351 samples/sec\taccuracy=0.540054\tlr=0.000002\n",
      "Epoch[46] Batch [6949]\tSpeed: 127.069913 samples/sec\taccuracy=0.539694\tlr=0.000002\n",
      "Epoch[46] Batch [6999]\tSpeed: 127.068027 samples/sec\taccuracy=0.539732\tlr=0.000002\n",
      "Epoch[46] Batch [7049]\tSpeed: 128.188990 samples/sec\taccuracy=0.539876\tlr=0.000002\n",
      "Epoch[46] Batch [7099]\tSpeed: 128.781204 samples/sec\taccuracy=0.539683\tlr=0.000002\n",
      "Epoch[46] Batch [7149]\tSpeed: 126.795208 samples/sec\taccuracy=0.539720\tlr=0.000002\n",
      "Epoch[46] Batch [7199]\tSpeed: 128.552117 samples/sec\taccuracy=0.539740\tlr=0.000002\n",
      "Epoch[46] Batch [7249]\tSpeed: 127.379275 samples/sec\taccuracy=0.539690\tlr=0.000002\n",
      "Epoch[46] Batch [7299]\tSpeed: 127.403767 samples/sec\taccuracy=0.539555\tlr=0.000002\n",
      "Epoch[46] Batch [7349]\tSpeed: 127.348750 samples/sec\taccuracy=0.539609\tlr=0.000002\n",
      "Epoch[46] Batch [7399]\tSpeed: 125.994273 samples/sec\taccuracy=0.539932\tlr=0.000002\n",
      "Epoch[46] Batch [7449]\tSpeed: 127.771999 samples/sec\taccuracy=0.539933\tlr=0.000002\n",
      "Epoch[46] Batch [7499]\tSpeed: 126.824230 samples/sec\taccuracy=0.539917\tlr=0.000002\n",
      "Epoch[46] Batch [7549]\tSpeed: 127.374739 samples/sec\taccuracy=0.539967\tlr=0.000002\n",
      "Epoch[46] Batch [7599]\tSpeed: 128.825198 samples/sec\taccuracy=0.539934\tlr=0.000002\n",
      "Epoch[46] Batch [7649]\tSpeed: 126.134874 samples/sec\taccuracy=0.539739\tlr=0.000002\n",
      "Epoch[46] Batch [7699]\tSpeed: 129.395499 samples/sec\taccuracy=0.539643\tlr=0.000002\n",
      "Epoch[46] Batch [7749]\tSpeed: 127.464719 samples/sec\taccuracy=0.539629\tlr=0.000002\n",
      "Epoch[46] Batch [7799]\tSpeed: 130.896241 samples/sec\taccuracy=0.539503\tlr=0.000002\n",
      "Epoch[46] Batch [7849]\tSpeed: 126.737412 samples/sec\taccuracy=0.539347\tlr=0.000002\n",
      "Epoch[46] Batch [7899]\tSpeed: 127.220933 samples/sec\taccuracy=0.539842\tlr=0.000002\n",
      "Epoch[46] Batch [7949]\tSpeed: 127.606244 samples/sec\taccuracy=0.539890\tlr=0.000002\n",
      "Epoch[46] Batch [7999]\tSpeed: 126.800162 samples/sec\taccuracy=0.540078\tlr=0.000002\n",
      "Epoch[46] Batch [8049]\tSpeed: 126.097882 samples/sec\taccuracy=0.540093\tlr=0.000002\n",
      "Epoch[46] Batch [8099]\tSpeed: 127.305266 samples/sec\taccuracy=0.540231\tlr=0.000002\n",
      "Epoch[46] Batch [8149]\tSpeed: 127.740264 samples/sec\taccuracy=0.540414\tlr=0.000002\n",
      "Epoch[46] Batch [8199]\tSpeed: 128.721703 samples/sec\taccuracy=0.540716\tlr=0.000002\n",
      "Epoch[46] Batch [8249]\tSpeed: 126.793243 samples/sec\taccuracy=0.540864\tlr=0.000002\n",
      "Epoch[46] Batch [8299]\tSpeed: 119.973358 samples/sec\taccuracy=0.540964\tlr=0.000002\n",
      "Epoch[46] Batch [8349]\tSpeed: 128.356410 samples/sec\taccuracy=0.540808\tlr=0.000002\n",
      "Epoch[46] Batch [8399]\tSpeed: 128.253754 samples/sec\taccuracy=0.540937\tlr=0.000002\n",
      "Epoch[46] Batch [8449]\tSpeed: 127.624328 samples/sec\taccuracy=0.541080\tlr=0.000002\n",
      "Epoch[46] Batch [8499]\tSpeed: 125.183496 samples/sec\taccuracy=0.540971\tlr=0.000002\n",
      "Epoch[46] Batch [8549]\tSpeed: 128.516628 samples/sec\taccuracy=0.541287\tlr=0.000002\n",
      "Epoch[46] Batch [8599]\tSpeed: 129.190671 samples/sec\taccuracy=0.541265\tlr=0.000002\n",
      "Epoch[46] Batch [8649]\tSpeed: 125.794470 samples/sec\taccuracy=0.541445\tlr=0.000002\n",
      "Epoch[46] Batch [8699]\tSpeed: 128.131551 samples/sec\taccuracy=0.541322\tlr=0.000002\n",
      "Epoch[46] Batch [8749]\tSpeed: 128.140281 samples/sec\taccuracy=0.541371\tlr=0.000002\n",
      "Epoch[46] Batch [8799]\tSpeed: 130.442462 samples/sec\taccuracy=0.541250\tlr=0.000002\n",
      "Epoch[46] Batch [8849]\tSpeed: 128.219369 samples/sec\taccuracy=0.541201\tlr=0.000002\n",
      "Epoch[46] Batch [8899]\tSpeed: 125.251261 samples/sec\taccuracy=0.541306\tlr=0.000002\n",
      "Epoch[46] Batch [8949]\tSpeed: 128.701233 samples/sec\taccuracy=0.541299\tlr=0.000002\n",
      "Epoch[46] Batch [8999]\tSpeed: 127.647915 samples/sec\taccuracy=0.541403\tlr=0.000002\n",
      "Epoch[46] Batch [9049]\tSpeed: 119.159817 samples/sec\taccuracy=0.541312\tlr=0.000002\n",
      "Epoch[46] Batch [9099]\tSpeed: 123.071678 samples/sec\taccuracy=0.541195\tlr=0.000002\n",
      "Epoch[46] Batch [9149]\tSpeed: 125.905223 samples/sec\taccuracy=0.541202\tlr=0.000002\n",
      "Epoch[46] Batch [9199]\tSpeed: 128.088499 samples/sec\taccuracy=0.541236\tlr=0.000002\n",
      "Epoch[46] Batch [9249]\tSpeed: 128.648040 samples/sec\taccuracy=0.541068\tlr=0.000002\n",
      "Epoch[46] Batch [9299]\tSpeed: 128.312774 samples/sec\taccuracy=0.540874\tlr=0.000002\n",
      "Epoch[46] Batch [9349]\tSpeed: 126.441867 samples/sec\taccuracy=0.540762\tlr=0.000002\n",
      "Epoch[46] Batch [9399]\tSpeed: 126.404037 samples/sec\taccuracy=0.540691\tlr=0.000002\n",
      "Epoch[46] Batch [9449]\tSpeed: 128.802104 samples/sec\taccuracy=0.540450\tlr=0.000002\n",
      "Epoch[46] Batch [9499]\tSpeed: 127.978783 samples/sec\taccuracy=0.540434\tlr=0.000002\n",
      "Epoch[46] Batch [9549]\tSpeed: 125.296020 samples/sec\taccuracy=0.540615\tlr=0.000002\n",
      "Epoch[46] Batch [9599]\tSpeed: 127.493225 samples/sec\taccuracy=0.540495\tlr=0.000002\n",
      "Epoch[46] Batch [9649]\tSpeed: 128.559919 samples/sec\taccuracy=0.540389\tlr=0.000002\n",
      "Epoch[46] Batch [9699]\tSpeed: 127.424339 samples/sec\taccuracy=0.540464\tlr=0.000002\n",
      "Epoch[46] Batch [9749]\tSpeed: 120.509874 samples/sec\taccuracy=0.540179\tlr=0.000002\n",
      "Epoch[46] Batch [9799]\tSpeed: 123.547876 samples/sec\taccuracy=0.540000\tlr=0.000002\n",
      "Epoch[46] Batch [9849]\tSpeed: 129.962385 samples/sec\taccuracy=0.539962\tlr=0.000002\n",
      "Epoch[46] Batch [9899]\tSpeed: 128.824535 samples/sec\taccuracy=0.539912\tlr=0.000002\n",
      "Epoch[46] Batch [9949]\tSpeed: 126.413933 samples/sec\taccuracy=0.540063\tlr=0.000002\n",
      "Epoch[46] Batch [9999]\tSpeed: 128.541667 samples/sec\taccuracy=0.540075\tlr=0.000002\n",
      "Epoch[46] Batch [10049]\tSpeed: 126.532222 samples/sec\taccuracy=0.540124\tlr=0.000002\n",
      "Epoch[46] Batch [10099]\tSpeed: 126.720172 samples/sec\taccuracy=0.539926\tlr=0.000002\n",
      "Epoch[46] Batch [10149]\tSpeed: 125.916903 samples/sec\taccuracy=0.539901\tlr=0.000002\n",
      "Epoch[46] Batch [10199]\tSpeed: 128.257774 samples/sec\taccuracy=0.539939\tlr=0.000002\n",
      "Epoch[46] Batch [10249]\tSpeed: 128.452414 samples/sec\taccuracy=0.539915\tlr=0.000002\n",
      "Epoch[46] Batch [10299]\tSpeed: 126.168491 samples/sec\taccuracy=0.540133\tlr=0.000002\n",
      "Epoch[46] Batch [10349]\tSpeed: 128.978359 samples/sec\taccuracy=0.540109\tlr=0.000002\n",
      "Epoch[46] Batch [10399]\tSpeed: 128.481413 samples/sec\taccuracy=0.540288\tlr=0.000002\n",
      "Epoch[46] Batch [10449]\tSpeed: 128.653959 samples/sec\taccuracy=0.540263\tlr=0.000002\n",
      "Epoch[46] Batch [10499]\tSpeed: 128.395633 samples/sec\taccuracy=0.540464\tlr=0.000002\n",
      "Epoch[46] Batch [10549]\tSpeed: 126.162884 samples/sec\taccuracy=0.540592\tlr=0.000002\n",
      "Epoch[46] Batch [10599]\tSpeed: 126.308996 samples/sec\taccuracy=0.540767\tlr=0.000002\n",
      "Epoch[46] Batch [10649]\tSpeed: 124.608425 samples/sec\taccuracy=0.540833\tlr=0.000002\n",
      "Epoch[46] Batch [10699]\tSpeed: 126.700822 samples/sec\taccuracy=0.540864\tlr=0.000002\n",
      "Epoch[46] Batch [10749]\tSpeed: 126.956314 samples/sec\taccuracy=0.540895\tlr=0.000002\n",
      "Epoch[46] Batch [10799]\tSpeed: 128.557033 samples/sec\taccuracy=0.540914\tlr=0.000002\n",
      "Epoch[46] Batch [10849]\tSpeed: 127.506083 samples/sec\taccuracy=0.540968\tlr=0.000002\n",
      "Epoch[46] Batch [10899]\tSpeed: 131.391034 samples/sec\taccuracy=0.540963\tlr=0.000002\n",
      "Epoch[46] Batch [10949]\tSpeed: 127.416539 samples/sec\taccuracy=0.541050\tlr=0.000002\n",
      "Epoch[46] Batch [10999]\tSpeed: 127.023830 samples/sec\taccuracy=0.541148\tlr=0.000002\n",
      "Epoch[46] Batch [11049]\tSpeed: 127.232173 samples/sec\taccuracy=0.541154\tlr=0.000002\n",
      "Epoch[46] Batch [11099]\tSpeed: 124.280443 samples/sec\taccuracy=0.541126\tlr=0.000002\n",
      "Epoch[46] Batch [11149]\tSpeed: 115.040882 samples/sec\taccuracy=0.541457\tlr=0.000002\n",
      "Epoch[46] Batch [11199]\tSpeed: 116.038256 samples/sec\taccuracy=0.541339\tlr=0.000002\n",
      "Epoch[46] Batch [11249]\tSpeed: 115.096443 samples/sec\taccuracy=0.541278\tlr=0.000002\n",
      "Epoch[46] Batch [11299]\tSpeed: 128.040833 samples/sec\taccuracy=0.541350\tlr=0.000002\n",
      "Epoch[46] Batch [11349]\tSpeed: 134.161091 samples/sec\taccuracy=0.541520\tlr=0.000002\n",
      "[Epoch 46] training: accuracy=0.541524\n",
      "[Epoch 46] speed: 126 samples/sec\ttime cost: 717.772362\n",
      "[Epoch 46] validation: top1=0.543058 top5=0.877614\n",
      "Epoch[47] Batch [49]\tSpeed: 106.574380 samples/sec\taccuracy=0.507500\tlr=0.000002\n",
      "Epoch[47] Batch [99]\tSpeed: 129.434082 samples/sec\taccuracy=0.501250\tlr=0.000002\n",
      "Epoch[47] Batch [149]\tSpeed: 125.120478 samples/sec\taccuracy=0.516667\tlr=0.000002\n",
      "Epoch[47] Batch [199]\tSpeed: 127.125392 samples/sec\taccuracy=0.511250\tlr=0.000002\n",
      "Epoch[47] Batch [249]\tSpeed: 118.061926 samples/sec\taccuracy=0.511500\tlr=0.000002\n",
      "Epoch[47] Batch [299]\tSpeed: 128.961039 samples/sec\taccuracy=0.520833\tlr=0.000002\n",
      "Epoch[47] Batch [349]\tSpeed: 125.459829 samples/sec\taccuracy=0.521429\tlr=0.000002\n",
      "Epoch[47] Batch [399]\tSpeed: 126.911043 samples/sec\taccuracy=0.518437\tlr=0.000002\n",
      "Epoch[47] Batch [449]\tSpeed: 126.372569 samples/sec\taccuracy=0.521667\tlr=0.000002\n",
      "Epoch[47] Batch [499]\tSpeed: 129.009243 samples/sec\taccuracy=0.523250\tlr=0.000002\n",
      "Epoch[47] Batch [549]\tSpeed: 127.442130 samples/sec\taccuracy=0.525455\tlr=0.000002\n",
      "Epoch[47] Batch [599]\tSpeed: 125.790480 samples/sec\taccuracy=0.527500\tlr=0.000002\n",
      "Epoch[47] Batch [649]\tSpeed: 129.003936 samples/sec\taccuracy=0.528077\tlr=0.000002\n",
      "Epoch[47] Batch [699]\tSpeed: 127.686687 samples/sec\taccuracy=0.530000\tlr=0.000002\n",
      "Epoch[47] Batch [749]\tSpeed: 128.622515 samples/sec\taccuracy=0.529833\tlr=0.000002\n",
      "Epoch[47] Batch [799]\tSpeed: 129.146506 samples/sec\taccuracy=0.533125\tlr=0.000002\n",
      "Epoch[47] Batch [849]\tSpeed: 128.847013 samples/sec\taccuracy=0.532941\tlr=0.000002\n",
      "Epoch[47] Batch [899]\tSpeed: 130.570609 samples/sec\taccuracy=0.533472\tlr=0.000002\n",
      "Epoch[47] Batch [949]\tSpeed: 128.203614 samples/sec\taccuracy=0.533816\tlr=0.000002\n",
      "Epoch[47] Batch [999]\tSpeed: 125.820857 samples/sec\taccuracy=0.534375\tlr=0.000002\n",
      "Epoch[47] Batch [1049]\tSpeed: 129.033205 samples/sec\taccuracy=0.533452\tlr=0.000002\n",
      "Epoch[47] Batch [1099]\tSpeed: 127.720815 samples/sec\taccuracy=0.532500\tlr=0.000002\n",
      "Epoch[47] Batch [1149]\tSpeed: 124.321406 samples/sec\taccuracy=0.533261\tlr=0.000002\n",
      "Epoch[47] Batch [1199]\tSpeed: 127.744272 samples/sec\taccuracy=0.533333\tlr=0.000002\n",
      "Epoch[47] Batch [1249]\tSpeed: 127.007512 samples/sec\taccuracy=0.536300\tlr=0.000002\n",
      "Epoch[47] Batch [1299]\tSpeed: 125.251466 samples/sec\taccuracy=0.536538\tlr=0.000002\n",
      "Epoch[47] Batch [1349]\tSpeed: 127.021079 samples/sec\taccuracy=0.538241\tlr=0.000002\n",
      "Epoch[47] Batch [1399]\tSpeed: 126.923937 samples/sec\taccuracy=0.538304\tlr=0.000002\n",
      "Epoch[47] Batch [1449]\tSpeed: 129.158397 samples/sec\taccuracy=0.537414\tlr=0.000002\n",
      "Epoch[47] Batch [1499]\tSpeed: 129.237793 samples/sec\taccuracy=0.536833\tlr=0.000002\n",
      "Epoch[47] Batch [1549]\tSpeed: 125.459960 samples/sec\taccuracy=0.536935\tlr=0.000002\n",
      "Epoch[47] Batch [1599]\tSpeed: 130.280910 samples/sec\taccuracy=0.538359\tlr=0.000002\n",
      "Epoch[47] Batch [1649]\tSpeed: 127.549850 samples/sec\taccuracy=0.538258\tlr=0.000002\n",
      "Epoch[47] Batch [1699]\tSpeed: 126.832773 samples/sec\taccuracy=0.539485\tlr=0.000002\n",
      "Epoch[47] Batch [1749]\tSpeed: 130.441935 samples/sec\taccuracy=0.539429\tlr=0.000002\n",
      "Epoch[47] Batch [1799]\tSpeed: 127.004377 samples/sec\taccuracy=0.540694\tlr=0.000002\n",
      "Epoch[47] Batch [1849]\tSpeed: 128.744343 samples/sec\taccuracy=0.542095\tlr=0.000002\n",
      "Epoch[47] Batch [1899]\tSpeed: 128.492592 samples/sec\taccuracy=0.541316\tlr=0.000002\n",
      "Epoch[47] Batch [1949]\tSpeed: 127.295723 samples/sec\taccuracy=0.541923\tlr=0.000002\n",
      "Epoch[47] Batch [1999]\tSpeed: 125.790641 samples/sec\taccuracy=0.542312\tlr=0.000002\n",
      "Epoch[47] Batch [2049]\tSpeed: 126.919809 samples/sec\taccuracy=0.542134\tlr=0.000002\n",
      "Epoch[47] Batch [2099]\tSpeed: 128.494845 samples/sec\taccuracy=0.540536\tlr=0.000002\n",
      "Epoch[47] Batch [2149]\tSpeed: 130.097509 samples/sec\taccuracy=0.541628\tlr=0.000002\n",
      "Epoch[47] Batch [2199]\tSpeed: 125.805978 samples/sec\taccuracy=0.542557\tlr=0.000002\n",
      "Epoch[47] Batch [2249]\tSpeed: 126.041573 samples/sec\taccuracy=0.543278\tlr=0.000002\n",
      "Epoch[47] Batch [2299]\tSpeed: 126.692737 samples/sec\taccuracy=0.543641\tlr=0.000002\n",
      "Epoch[47] Batch [2349]\tSpeed: 127.291618 samples/sec\taccuracy=0.544309\tlr=0.000002\n",
      "Epoch[47] Batch [2399]\tSpeed: 127.590727 samples/sec\taccuracy=0.543854\tlr=0.000002\n",
      "Epoch[47] Batch [2449]\tSpeed: 127.904242 samples/sec\taccuracy=0.543776\tlr=0.000002\n",
      "Epoch[47] Batch [2499]\tSpeed: 125.900376 samples/sec\taccuracy=0.543900\tlr=0.000002\n",
      "Epoch[47] Batch [2549]\tSpeed: 128.697185 samples/sec\taccuracy=0.545049\tlr=0.000002\n",
      "Epoch[47] Batch [2599]\tSpeed: 127.112158 samples/sec\taccuracy=0.544856\tlr=0.000002\n",
      "Epoch[47] Batch [2649]\tSpeed: 128.311047 samples/sec\taccuracy=0.544623\tlr=0.000002\n",
      "Epoch[47] Batch [2699]\tSpeed: 129.551781 samples/sec\taccuracy=0.543380\tlr=0.000002\n",
      "Epoch[47] Batch [2749]\tSpeed: 127.392071 samples/sec\taccuracy=0.543000\tlr=0.000002\n",
      "Epoch[47] Batch [2799]\tSpeed: 128.213696 samples/sec\taccuracy=0.542902\tlr=0.000002\n",
      "Epoch[47] Batch [2849]\tSpeed: 124.508413 samples/sec\taccuracy=0.542105\tlr=0.000002\n",
      "Epoch[47] Batch [2899]\tSpeed: 127.212357 samples/sec\taccuracy=0.542586\tlr=0.000002\n",
      "Epoch[47] Batch [2949]\tSpeed: 128.201293 samples/sec\taccuracy=0.542203\tlr=0.000002\n",
      "Epoch[47] Batch [2999]\tSpeed: 127.515503 samples/sec\taccuracy=0.541917\tlr=0.000002\n",
      "Epoch[47] Batch [3049]\tSpeed: 130.998815 samples/sec\taccuracy=0.542131\tlr=0.000002\n",
      "Epoch[47] Batch [3099]\tSpeed: 119.434702 samples/sec\taccuracy=0.542056\tlr=0.000002\n",
      "Epoch[47] Batch [3149]\tSpeed: 123.163381 samples/sec\taccuracy=0.541865\tlr=0.000002\n",
      "Epoch[47] Batch [3199]\tSpeed: 127.116723 samples/sec\taccuracy=0.541289\tlr=0.000002\n",
      "Epoch[47] Batch [3249]\tSpeed: 128.616638 samples/sec\taccuracy=0.541500\tlr=0.000002\n",
      "Epoch[47] Batch [3299]\tSpeed: 129.343306 samples/sec\taccuracy=0.541364\tlr=0.000002\n",
      "Epoch[47] Batch [3349]\tSpeed: 128.084460 samples/sec\taccuracy=0.541082\tlr=0.000002\n",
      "Epoch[47] Batch [3399]\tSpeed: 124.701201 samples/sec\taccuracy=0.542022\tlr=0.000002\n",
      "Epoch[47] Batch [3449]\tSpeed: 126.327095 samples/sec\taccuracy=0.542065\tlr=0.000002\n",
      "Epoch[47] Batch [3499]\tSpeed: 127.320753 samples/sec\taccuracy=0.541714\tlr=0.000002\n",
      "Epoch[47] Batch [3549]\tSpeed: 128.407268 samples/sec\taccuracy=0.541479\tlr=0.000002\n",
      "Epoch[47] Batch [3599]\tSpeed: 128.947311 samples/sec\taccuracy=0.541389\tlr=0.000002\n",
      "Epoch[47] Batch [3649]\tSpeed: 128.420233 samples/sec\taccuracy=0.540856\tlr=0.000002\n",
      "Epoch[47] Batch [3699]\tSpeed: 129.412098 samples/sec\taccuracy=0.540743\tlr=0.000002\n",
      "Epoch[47] Batch [3749]\tSpeed: 128.365199 samples/sec\taccuracy=0.540467\tlr=0.000002\n",
      "Epoch[47] Batch [3799]\tSpeed: 128.496597 samples/sec\taccuracy=0.540625\tlr=0.000002\n",
      "Epoch[47] Batch [3849]\tSpeed: 127.738942 samples/sec\taccuracy=0.540714\tlr=0.000002\n",
      "Epoch[47] Batch [3899]\tSpeed: 128.637623 samples/sec\taccuracy=0.540481\tlr=0.000002\n",
      "Epoch[47] Batch [3949]\tSpeed: 125.270199 samples/sec\taccuracy=0.540886\tlr=0.000002\n",
      "Epoch[47] Batch [3999]\tSpeed: 126.869545 samples/sec\taccuracy=0.540719\tlr=0.000002\n",
      "Epoch[47] Batch [4049]\tSpeed: 127.435005 samples/sec\taccuracy=0.541173\tlr=0.000002\n",
      "Epoch[47] Batch [4099]\tSpeed: 127.589582 samples/sec\taccuracy=0.541189\tlr=0.000002\n",
      "Epoch[47] Batch [4149]\tSpeed: 128.534409 samples/sec\taccuracy=0.541235\tlr=0.000002\n",
      "Epoch[47] Batch [4199]\tSpeed: 126.538425 samples/sec\taccuracy=0.541756\tlr=0.000002\n",
      "Epoch[47] Batch [4249]\tSpeed: 125.625196 samples/sec\taccuracy=0.541912\tlr=0.000002\n",
      "Epoch[47] Batch [4299]\tSpeed: 127.633368 samples/sec\taccuracy=0.541831\tlr=0.000002\n",
      "Epoch[47] Batch [4349]\tSpeed: 127.738037 samples/sec\taccuracy=0.541494\tlr=0.000002\n",
      "Epoch[47] Batch [4399]\tSpeed: 129.924976 samples/sec\taccuracy=0.541619\tlr=0.000002\n",
      "Epoch[47] Batch [4449]\tSpeed: 130.541959 samples/sec\taccuracy=0.541433\tlr=0.000002\n",
      "Epoch[47] Batch [4499]\tSpeed: 129.011435 samples/sec\taccuracy=0.541389\tlr=0.000002\n",
      "Epoch[47] Batch [4549]\tSpeed: 128.512995 samples/sec\taccuracy=0.541429\tlr=0.000002\n",
      "Epoch[47] Batch [4599]\tSpeed: 129.561435 samples/sec\taccuracy=0.540978\tlr=0.000002\n",
      "Epoch[47] Batch [4649]\tSpeed: 126.850447 samples/sec\taccuracy=0.540887\tlr=0.000002\n",
      "Epoch[47] Batch [4699]\tSpeed: 129.618130 samples/sec\taccuracy=0.540638\tlr=0.000002\n",
      "Epoch[47] Batch [4749]\tSpeed: 128.268325 samples/sec\taccuracy=0.540711\tlr=0.000002\n",
      "Epoch[47] Batch [4799]\tSpeed: 130.147960 samples/sec\taccuracy=0.540937\tlr=0.000002\n",
      "Epoch[47] Batch [4849]\tSpeed: 126.961973 samples/sec\taccuracy=0.540954\tlr=0.000002\n",
      "Epoch[47] Batch [4899]\tSpeed: 127.839956 samples/sec\taccuracy=0.540944\tlr=0.000002\n",
      "Epoch[47] Batch [4949]\tSpeed: 127.354831 samples/sec\taccuracy=0.540657\tlr=0.000002\n",
      "Epoch[47] Batch [4999]\tSpeed: 128.279957 samples/sec\taccuracy=0.540625\tlr=0.000002\n",
      "Epoch[47] Batch [5049]\tSpeed: 129.588866 samples/sec\taccuracy=0.540718\tlr=0.000002\n",
      "Epoch[47] Batch [5099]\tSpeed: 125.762758 samples/sec\taccuracy=0.540907\tlr=0.000002\n",
      "Epoch[47] Batch [5149]\tSpeed: 127.875239 samples/sec\taccuracy=0.540898\tlr=0.000002\n",
      "Epoch[47] Batch [5199]\tSpeed: 129.507169 samples/sec\taccuracy=0.540986\tlr=0.000002\n",
      "Epoch[47] Batch [5249]\tSpeed: 127.390697 samples/sec\taccuracy=0.541476\tlr=0.000002\n",
      "Epoch[47] Batch [5299]\tSpeed: 129.829300 samples/sec\taccuracy=0.541509\tlr=0.000002\n",
      "Epoch[47] Batch [5349]\tSpeed: 128.274386 samples/sec\taccuracy=0.541332\tlr=0.000002\n",
      "Epoch[47] Batch [5399]\tSpeed: 128.051134 samples/sec\taccuracy=0.541458\tlr=0.000002\n",
      "Epoch[47] Batch [5449]\tSpeed: 126.084719 samples/sec\taccuracy=0.541560\tlr=0.000002\n",
      "Epoch[47] Batch [5499]\tSpeed: 128.937926 samples/sec\taccuracy=0.541432\tlr=0.000002\n",
      "Epoch[47] Batch [5549]\tSpeed: 128.737032 samples/sec\taccuracy=0.541757\tlr=0.000002\n",
      "Epoch[47] Batch [5599]\tSpeed: 127.779395 samples/sec\taccuracy=0.541808\tlr=0.000002\n",
      "Epoch[47] Batch [5649]\tSpeed: 131.437427 samples/sec\taccuracy=0.541748\tlr=0.000002\n",
      "Epoch[47] Batch [5699]\tSpeed: 125.194864 samples/sec\taccuracy=0.541798\tlr=0.000002\n",
      "Epoch[47] Batch [5749]\tSpeed: 127.581063 samples/sec\taccuracy=0.541870\tlr=0.000002\n",
      "Epoch[47] Batch [5799]\tSpeed: 127.623329 samples/sec\taccuracy=0.542091\tlr=0.000002\n",
      "Epoch[47] Batch [5849]\tSpeed: 126.298470 samples/sec\taccuracy=0.542457\tlr=0.000002\n",
      "Epoch[47] Batch [5899]\tSpeed: 128.556865 samples/sec\taccuracy=0.542140\tlr=0.000002\n",
      "Epoch[47] Batch [5949]\tSpeed: 123.919484 samples/sec\taccuracy=0.542290\tlr=0.000002\n",
      "Epoch[47] Batch [5999]\tSpeed: 118.503458 samples/sec\taccuracy=0.542063\tlr=0.000002\n",
      "Epoch[47] Batch [6049]\tSpeed: 129.884380 samples/sec\taccuracy=0.541901\tlr=0.000002\n",
      "Epoch[47] Batch [6099]\tSpeed: 128.352452 samples/sec\taccuracy=0.541803\tlr=0.000002\n",
      "Epoch[47] Batch [6149]\tSpeed: 127.690399 samples/sec\taccuracy=0.541687\tlr=0.000002\n",
      "Epoch[47] Batch [6199]\tSpeed: 128.464738 samples/sec\taccuracy=0.541411\tlr=0.000002\n",
      "Epoch[47] Batch [6249]\tSpeed: 124.572285 samples/sec\taccuracy=0.541480\tlr=0.000002\n",
      "Epoch[47] Batch [6299]\tSpeed: 127.952957 samples/sec\taccuracy=0.541548\tlr=0.000002\n",
      "Epoch[47] Batch [6349]\tSpeed: 126.062957 samples/sec\taccuracy=0.541713\tlr=0.000002\n",
      "Epoch[47] Batch [6399]\tSpeed: 125.456114 samples/sec\taccuracy=0.541523\tlr=0.000002\n",
      "Epoch[47] Batch [6449]\tSpeed: 126.750846 samples/sec\taccuracy=0.541492\tlr=0.000002\n",
      "Epoch[47] Batch [6499]\tSpeed: 127.605555 samples/sec\taccuracy=0.541673\tlr=0.000002\n",
      "Epoch[47] Batch [6549]\tSpeed: 126.696985 samples/sec\taccuracy=0.541336\tlr=0.000002\n",
      "Epoch[47] Batch [6599]\tSpeed: 128.848211 samples/sec\taccuracy=0.541307\tlr=0.000002\n",
      "Epoch[47] Batch [6649]\tSpeed: 125.897863 samples/sec\taccuracy=0.541335\tlr=0.000002\n",
      "Epoch[47] Batch [6699]\tSpeed: 129.211167 samples/sec\taccuracy=0.541642\tlr=0.000002\n",
      "Epoch[47] Batch [6749]\tSpeed: 127.935813 samples/sec\taccuracy=0.541889\tlr=0.000002\n",
      "Epoch[47] Batch [7599]\tSpeed: 130.261246 samples/sec\taccuracy=0.542122\tlr=0.000002\n",
      "Epoch[47] Batch [7649]\tSpeed: 130.026526 samples/sec\taccuracy=0.542222\tlr=0.000002\n",
      "Epoch[47] Batch [7699]\tSpeed: 125.940042 samples/sec\taccuracy=0.542256\tlr=0.000002\n",
      "Epoch[47] Batch [7749]\tSpeed: 128.545538 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[47] Batch [7799]\tSpeed: 123.780872 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[47] Batch [7849]\tSpeed: 124.793326 samples/sec\taccuracy=0.542516\tlr=0.000002\n",
      "Epoch[47] Batch [7899]\tSpeed: 128.003359 samples/sec\taccuracy=0.542516\tlr=0.000002\n",
      "Epoch[47] Batch [7949]\tSpeed: 127.408150 samples/sec\taccuracy=0.542374\tlr=0.000002\n",
      "Epoch[47] Batch [7999]\tSpeed: 125.789924 samples/sec\taccuracy=0.542063\tlr=0.000002\n",
      "Epoch[47] Batch [8049]\tSpeed: 127.392400 samples/sec\taccuracy=0.541988\tlr=0.000002\n",
      "Epoch[47] Batch [8099]\tSpeed: 127.713407 samples/sec\taccuracy=0.541944\tlr=0.000002\n",
      "Epoch[47] Batch [8149]\tSpeed: 128.561013 samples/sec\taccuracy=0.541917\tlr=0.000002\n",
      "Epoch[47] Batch [8199]\tSpeed: 125.915854 samples/sec\taccuracy=0.542073\tlr=0.000002\n",
      "Epoch[47] Batch [8249]\tSpeed: 129.635196 samples/sec\taccuracy=0.542121\tlr=0.000002\n",
      "Epoch[47] Batch [8299]\tSpeed: 126.818488 samples/sec\taccuracy=0.542108\tlr=0.000002\n",
      "Epoch[47] Batch [8349]\tSpeed: 127.948731 samples/sec\taccuracy=0.542186\tlr=0.000002\n",
      "Epoch[47] Batch [8399]\tSpeed: 130.597462 samples/sec\taccuracy=0.542009\tlr=0.000002\n",
      "Epoch[47] Batch [8449]\tSpeed: 124.814456 samples/sec\taccuracy=0.541997\tlr=0.000002\n",
      "Epoch[47] Batch [8499]\tSpeed: 127.801519 samples/sec\taccuracy=0.542044\tlr=0.000002\n",
      "Epoch[47] Batch [8549]\tSpeed: 126.267036 samples/sec\taccuracy=0.541871\tlr=0.000002\n",
      "Epoch[47] Batch [8599]\tSpeed: 124.985704 samples/sec\taccuracy=0.541599\tlr=0.000002\n",
      "Epoch[47] Batch [8649]\tSpeed: 125.757140 samples/sec\taccuracy=0.541633\tlr=0.000002\n",
      "Epoch[47] Batch [8699]\tSpeed: 125.761702 samples/sec\taccuracy=0.541652\tlr=0.000002\n",
      "Epoch[47] Batch [8749]\tSpeed: 126.460032 samples/sec\taccuracy=0.541500\tlr=0.000002\n",
      "Epoch[47] Batch [8799]\tSpeed: 124.986663 samples/sec\taccuracy=0.541435\tlr=0.000002\n",
      "Epoch[47] Batch [8849]\tSpeed: 118.806332 samples/sec\taccuracy=0.541427\tlr=0.000002\n",
      "Epoch[47] Batch [8899]\tSpeed: 127.460458 samples/sec\taccuracy=0.541433\tlr=0.000002\n",
      "Epoch[47] Batch [8949]\tSpeed: 128.596359 samples/sec\taccuracy=0.541564\tlr=0.000002\n",
      "Epoch[47] Batch [8999]\tSpeed: 130.019664 samples/sec\taccuracy=0.541542\tlr=0.000002\n",
      "Epoch[47] Batch [9049]\tSpeed: 126.622190 samples/sec\taccuracy=0.541685\tlr=0.000002\n",
      "Epoch[47] Batch [9099]\tSpeed: 131.427481 samples/sec\taccuracy=0.541896\tlr=0.000002\n",
      "Epoch[47] Batch [9149]\tSpeed: 127.918460 samples/sec\taccuracy=0.541721\tlr=0.000002\n",
      "Epoch[47] Batch [9199]\tSpeed: 128.301274 samples/sec\taccuracy=0.541671\tlr=0.000002\n",
      "Epoch[47] Batch [9249]\tSpeed: 128.821528 samples/sec\taccuracy=0.541459\tlr=0.000002\n",
      "Epoch[47] Batch [9299]\tSpeed: 127.680419 samples/sec\taccuracy=0.541263\tlr=0.000002\n",
      "Epoch[47] Batch [9349]\tSpeed: 128.124584 samples/sec\taccuracy=0.541297\tlr=0.000002\n",
      "Epoch[47] Batch [9399]\tSpeed: 127.222071 samples/sec\taccuracy=0.541277\tlr=0.000002\n",
      "Epoch[47] Batch [9449]\tSpeed: 126.262579 samples/sec\taccuracy=0.541270\tlr=0.000002\n",
      "Epoch[47] Batch [9499]\tSpeed: 125.098974 samples/sec\taccuracy=0.541421\tlr=0.000002\n",
      "Epoch[47] Batch [9549]\tSpeed: 127.226258 samples/sec\taccuracy=0.541571\tlr=0.000002\n",
      "Epoch[47] Batch [9599]\tSpeed: 127.421987 samples/sec\taccuracy=0.541432\tlr=0.000002\n",
      "Epoch[47] Batch [9649]\tSpeed: 127.321516 samples/sec\taccuracy=0.541412\tlr=0.000002\n",
      "Epoch[47] Batch [9699]\tSpeed: 129.151298 samples/sec\taccuracy=0.541649\tlr=0.000002\n",
      "Epoch[47] Batch [9749]\tSpeed: 125.727765 samples/sec\taccuracy=0.541603\tlr=0.000002\n",
      "Epoch[47] Batch [9799]\tSpeed: 124.780638 samples/sec\taccuracy=0.541760\tlr=0.000002\n",
      "Epoch[47] Batch [9849]\tSpeed: 128.212902 samples/sec\taccuracy=0.541701\tlr=0.000002\n",
      "Epoch[47] Batch [9899]\tSpeed: 128.328821 samples/sec\taccuracy=0.541376\tlr=0.000002\n",
      "Epoch[47] Batch [9949]\tSpeed: 128.713970 samples/sec\taccuracy=0.541369\tlr=0.000002\n",
      "Epoch[47] Batch [9999]\tSpeed: 128.919082 samples/sec\taccuracy=0.541375\tlr=0.000002\n",
      "Epoch[47] Batch [10049]\tSpeed: 126.401313 samples/sec\taccuracy=0.541356\tlr=0.000002\n",
      "Epoch[47] Batch [10099]\tSpeed: 127.357103 samples/sec\taccuracy=0.541473\tlr=0.000002\n",
      "Epoch[47] Batch [10149]\tSpeed: 127.760585 samples/sec\taccuracy=0.541268\tlr=0.000002\n",
      "Epoch[47] Batch [10199]\tSpeed: 126.348501 samples/sec\taccuracy=0.541275\tlr=0.000002\n",
      "Epoch[47] Batch [10249]\tSpeed: 128.417844 samples/sec\taccuracy=0.541220\tlr=0.000002\n",
      "Epoch[47] Batch [10299]\tSpeed: 128.575841 samples/sec\taccuracy=0.541214\tlr=0.000002\n",
      "Epoch[47] Batch [10349]\tSpeed: 127.637223 samples/sec\taccuracy=0.541051\tlr=0.000002\n",
      "Epoch[47] Batch [10399]\tSpeed: 128.100929 samples/sec\taccuracy=0.541034\tlr=0.000002\n",
      "Epoch[47] Batch [10449]\tSpeed: 127.513642 samples/sec\taccuracy=0.540957\tlr=0.000002\n",
      "Epoch[47] Batch [10499]\tSpeed: 129.793052 samples/sec\taccuracy=0.541048\tlr=0.000002\n",
      "Epoch[47] Batch [10549]\tSpeed: 128.022641 samples/sec\taccuracy=0.541090\tlr=0.000002\n",
      "Epoch[47] Batch [10599]\tSpeed: 125.482340 samples/sec\taccuracy=0.541203\tlr=0.000002\n",
      "Epoch[47] Batch [10649]\tSpeed: 130.240759 samples/sec\taccuracy=0.541338\tlr=0.000002\n",
      "Epoch[47] Batch [10699]\tSpeed: 127.497459 samples/sec\taccuracy=0.541273\tlr=0.000002\n",
      "Epoch[47] Batch [10749]\tSpeed: 125.180908 samples/sec\taccuracy=0.541174\tlr=0.000002\n",
      "Epoch[47] Batch [10799]\tSpeed: 126.399552 samples/sec\taccuracy=0.541088\tlr=0.000002\n",
      "Epoch[47] Batch [10849]\tSpeed: 125.625506 samples/sec\taccuracy=0.540979\tlr=0.000002\n",
      "Epoch[47] Batch [10899]\tSpeed: 128.577880 samples/sec\taccuracy=0.540986\tlr=0.000002\n",
      "Epoch[47] Batch [10949]\tSpeed: 125.684758 samples/sec\taccuracy=0.540868\tlr=0.000002\n",
      "Epoch[47] Batch [10999]\tSpeed: 127.042327 samples/sec\taccuracy=0.540795\tlr=0.000002\n",
      "Epoch[47] Batch [11049]\tSpeed: 127.092544 samples/sec\taccuracy=0.540837\tlr=0.000002\n",
      "Epoch[47] Batch [11099]\tSpeed: 128.029372 samples/sec\taccuracy=0.540721\tlr=0.000002\n",
      "Epoch[47] Batch [11149]\tSpeed: 128.222574 samples/sec\taccuracy=0.540729\tlr=0.000002\n",
      "Epoch[47] Batch [11199]\tSpeed: 129.192651 samples/sec\taccuracy=0.540725\tlr=0.000002\n",
      "Epoch[47] Batch [11249]\tSpeed: 131.019388 samples/sec\taccuracy=0.540656\tlr=0.000002\n",
      "Epoch[47] Batch [11299]\tSpeed: 129.505320 samples/sec\taccuracy=0.540542\tlr=0.000002\n",
      "Epoch[47] Batch [11349]\tSpeed: 138.024692 samples/sec\taccuracy=0.540683\tlr=0.000002\n",
      "[Epoch 47] training: accuracy=0.540643\n",
      "[Epoch 47] speed: 127 samples/sec\ttime cost: 713.098389\n",
      "[Epoch 47] validation: top1=0.544049 top5=0.876524\n",
      "Epoch[48] Batch [49]\tSpeed: 106.665141 samples/sec\taccuracy=0.530000\tlr=0.000002\n",
      "Epoch[48] Batch [99]\tSpeed: 129.475326 samples/sec\taccuracy=0.550000\tlr=0.000002\n",
      "Epoch[48] Batch [2399]\tSpeed: 126.227735 samples/sec\taccuracy=0.542865\tlr=0.000002\n",
      "Epoch[48] Batch [2449]\tSpeed: 127.526805 samples/sec\taccuracy=0.543520\tlr=0.000002\n",
      "Epoch[48] Batch [2499]\tSpeed: 126.164677 samples/sec\taccuracy=0.543550\tlr=0.000002\n",
      "Epoch[48] Batch [2549]\tSpeed: 130.570131 samples/sec\taccuracy=0.543578\tlr=0.000002\n",
      "Epoch[48] Batch [2599]\tSpeed: 126.270334 samples/sec\taccuracy=0.544279\tlr=0.000002\n",
      "Epoch[48] Batch [2649]\tSpeed: 127.807672 samples/sec\taccuracy=0.543774\tlr=0.000002\n",
      "Epoch[48] Batch [2699]\tSpeed: 126.197066 samples/sec\taccuracy=0.543657\tlr=0.000002\n",
      "Epoch[48] Batch [2749]\tSpeed: 128.306926 samples/sec\taccuracy=0.543682\tlr=0.000002\n",
      "Epoch[48] Batch [2799]\tSpeed: 124.927779 samples/sec\taccuracy=0.544598\tlr=0.000002\n",
      "Epoch[48] Batch [2849]\tSpeed: 128.094934 samples/sec\taccuracy=0.545307\tlr=0.000002\n",
      "Epoch[48] Batch [2899]\tSpeed: 126.669962 samples/sec\taccuracy=0.545216\tlr=0.000002\n",
      "Epoch[48] Batch [2949]\tSpeed: 127.716012 samples/sec\taccuracy=0.544449\tlr=0.000002\n",
      "Epoch[48] Batch [2999]\tSpeed: 126.893650 samples/sec\taccuracy=0.543875\tlr=0.000002\n",
      "Epoch[48] Batch [3049]\tSpeed: 128.883102 samples/sec\taccuracy=0.544098\tlr=0.000002\n",
      "Epoch[48] Batch [3099]\tSpeed: 127.243482 samples/sec\taccuracy=0.543871\tlr=0.000002\n",
      "Epoch[48] Batch [3149]\tSpeed: 126.905965 samples/sec\taccuracy=0.543810\tlr=0.000002\n",
      "Epoch[48] Batch [3199]\tSpeed: 127.850887 samples/sec\taccuracy=0.544297\tlr=0.000002\n",
      "Epoch[48] Batch [3249]\tSpeed: 128.094014 samples/sec\taccuracy=0.543846\tlr=0.000002\n",
      "Epoch[48] Batch [3299]\tSpeed: 125.500372 samples/sec\taccuracy=0.543068\tlr=0.000002\n",
      "Epoch[48] Batch [3349]\tSpeed: 127.020627 samples/sec\taccuracy=0.542873\tlr=0.000002\n",
      "Epoch[48] Batch [3399]\tSpeed: 129.712110 samples/sec\taccuracy=0.543199\tlr=0.000002\n",
      "Epoch[48] Batch [3449]\tSpeed: 126.933396 samples/sec\taccuracy=0.543080\tlr=0.000002\n",
      "Epoch[48] Batch [3499]\tSpeed: 125.331963 samples/sec\taccuracy=0.543250\tlr=0.000002\n",
      "Epoch[48] Batch [3549]\tSpeed: 125.346036 samples/sec\taccuracy=0.543345\tlr=0.000002\n",
      "Epoch[48] Batch [3599]\tSpeed: 128.335771 samples/sec\taccuracy=0.543125\tlr=0.000002\n",
      "Epoch[48] Batch [3649]\tSpeed: 126.772204 samples/sec\taccuracy=0.543630\tlr=0.000002\n",
      "Epoch[48] Batch [3699]\tSpeed: 128.399082 samples/sec\taccuracy=0.543851\tlr=0.000002\n",
      "Epoch[48] Batch [3749]\tSpeed: 126.396523 samples/sec\taccuracy=0.543867\tlr=0.000002\n",
      "Epoch[48] Batch [3799]\tSpeed: 131.239669 samples/sec\taccuracy=0.544145\tlr=0.000002\n",
      "Epoch[48] Batch [3849]\tSpeed: 126.921306 samples/sec\taccuracy=0.543994\tlr=0.000002\n",
      "Epoch[48] Batch [3899]\tSpeed: 127.726076 samples/sec\taccuracy=0.544263\tlr=0.000002\n",
      "Epoch[48] Batch [3949]\tSpeed: 129.540637 samples/sec\taccuracy=0.544430\tlr=0.000002\n",
      "Epoch[48] Batch [3999]\tSpeed: 126.194437 samples/sec\taccuracy=0.544219\tlr=0.000002\n",
      "Epoch[48] Batch [4049]\tSpeed: 128.885834 samples/sec\taccuracy=0.544506\tlr=0.000002\n",
      "Epoch[48] Batch [4099]\tSpeed: 129.273861 samples/sec\taccuracy=0.544360\tlr=0.000002\n",
      "Epoch[48] Batch [4149]\tSpeed: 127.911955 samples/sec\taccuracy=0.543735\tlr=0.000002\n",
      "Epoch[48] Batch [4199]\tSpeed: 127.924059 samples/sec\taccuracy=0.543571\tlr=0.000002\n",
      "Epoch[48] Batch [4249]\tSpeed: 127.037786 samples/sec\taccuracy=0.544118\tlr=0.000002\n",
      "Epoch[48] Batch [4299]\tSpeed: 130.513524 samples/sec\taccuracy=0.544302\tlr=0.000002\n",
      "Epoch[48] Batch [4349]\tSpeed: 127.912257 samples/sec\taccuracy=0.544023\tlr=0.000002\n",
      "Epoch[48] Batch [4399]\tSpeed: 127.655587 samples/sec\taccuracy=0.543636\tlr=0.000002\n",
      "Epoch[48] Batch [4449]\tSpeed: 128.240372 samples/sec\taccuracy=0.543933\tlr=0.000002\n",
      "Epoch[48] Batch [4499]\tSpeed: 127.270567 samples/sec\taccuracy=0.543917\tlr=0.000002\n",
      "Epoch[48] Batch [4549]\tSpeed: 126.612510 samples/sec\taccuracy=0.544066\tlr=0.000002\n",
      "Epoch[48] Batch [4599]\tSpeed: 127.460187 samples/sec\taccuracy=0.543995\tlr=0.000002\n",
      "Epoch[48] Batch [4649]\tSpeed: 128.478924 samples/sec\taccuracy=0.544086\tlr=0.000002\n",
      "Epoch[48] Batch [4699]\tSpeed: 127.907108 samples/sec\taccuracy=0.543830\tlr=0.000002\n",
      "Epoch[48] Batch [4749]\tSpeed: 125.328620 samples/sec\taccuracy=0.543921\tlr=0.000002\n",
      "Epoch[48] Batch [4799]\tSpeed: 130.859986 samples/sec\taccuracy=0.543333\tlr=0.000002\n",
      "Epoch[48] Batch [4849]\tSpeed: 127.315806 samples/sec\taccuracy=0.543557\tlr=0.000002\n",
      "Epoch[48] Batch [4899]\tSpeed: 128.728231 samples/sec\taccuracy=0.544082\tlr=0.000002\n",
      "Epoch[48] Batch [4949]\tSpeed: 127.299963 samples/sec\taccuracy=0.543510\tlr=0.000002\n",
      "Epoch[48] Batch [4999]\tSpeed: 128.057839 samples/sec\taccuracy=0.543650\tlr=0.000002\n",
      "Epoch[48] Batch [5049]\tSpeed: 126.548456 samples/sec\taccuracy=0.543564\tlr=0.000002\n",
      "Epoch[48] Batch [5099]\tSpeed: 121.024246 samples/sec\taccuracy=0.543652\tlr=0.000002\n",
      "Epoch[48] Batch [5149]\tSpeed: 128.349428 samples/sec\taccuracy=0.543883\tlr=0.000002\n",
      "Epoch[48] Batch [5199]\tSpeed: 130.882936 samples/sec\taccuracy=0.543870\tlr=0.000002\n",
      "Epoch[48] Batch [5249]\tSpeed: 128.336763 samples/sec\taccuracy=0.544000\tlr=0.000002\n",
      "Epoch[48] Batch [5299]\tSpeed: 126.036403 samples/sec\taccuracy=0.544175\tlr=0.000002\n",
      "Epoch[48] Batch [5349]\tSpeed: 126.727217 samples/sec\taccuracy=0.544112\tlr=0.000002\n",
      "Epoch[48] Batch [5399]\tSpeed: 129.037929 samples/sec\taccuracy=0.543704\tlr=0.000002\n",
      "Epoch[48] Batch [5449]\tSpeed: 128.885795 samples/sec\taccuracy=0.543440\tlr=0.000002\n",
      "Epoch[48] Batch [5499]\tSpeed: 127.986954 samples/sec\taccuracy=0.543318\tlr=0.000002\n",
      "Epoch[48] Batch [5549]\tSpeed: 129.357817 samples/sec\taccuracy=0.543491\tlr=0.000002\n",
      "Epoch[48] Batch [5599]\tSpeed: 125.811837 samples/sec\taccuracy=0.543571\tlr=0.000002\n",
      "Epoch[48] Batch [5649]\tSpeed: 128.011290 samples/sec\taccuracy=0.543429\tlr=0.000002\n",
      "Epoch[48] Batch [5699]\tSpeed: 126.237128 samples/sec\taccuracy=0.543136\tlr=0.000002\n",
      "Epoch[48] Batch [5749]\tSpeed: 125.820432 samples/sec\taccuracy=0.543239\tlr=0.000002\n",
      "Epoch[48] Batch [5799]\tSpeed: 128.116776 samples/sec\taccuracy=0.543125\tlr=0.000002\n",
      "Epoch[48] Batch [5849]\tSpeed: 126.329930 samples/sec\taccuracy=0.542970\tlr=0.000002\n",
      "Epoch[48] Batch [5899]\tSpeed: 128.197247 samples/sec\taccuracy=0.543157\tlr=0.000002\n",
      "Epoch[48] Batch [5949]\tSpeed: 127.795513 samples/sec\taccuracy=0.543172\tlr=0.000002\n",
      "Epoch[48] Batch [5999]\tSpeed: 126.814711 samples/sec\taccuracy=0.543021\tlr=0.000002\n",
      "Epoch[48] Batch [6049]\tSpeed: 127.412020 samples/sec\taccuracy=0.542789\tlr=0.000002\n",
      "Epoch[48] Batch [6099]\tSpeed: 129.335778 samples/sec\taccuracy=0.542480\tlr=0.000002\n",
      "Epoch[48] Batch [6149]\tSpeed: 126.990034 samples/sec\taccuracy=0.542622\tlr=0.000002\n",
      "Epoch[48] Batch [6199]\tSpeed: 127.968348 samples/sec\taccuracy=0.542823\tlr=0.000002\n",
      "Epoch[48] Batch [6249]\tSpeed: 127.146713 samples/sec\taccuracy=0.542860\tlr=0.000002\n",
      "Epoch[48] Batch [6299]\tSpeed: 125.016196 samples/sec\taccuracy=0.543234\tlr=0.000002\n",
      "Epoch[48] Batch [6349]\tSpeed: 125.119443 samples/sec\taccuracy=0.543110\tlr=0.000002\n",
      "Epoch[48] Batch [6399]\tSpeed: 128.723194 samples/sec\taccuracy=0.543184\tlr=0.000002\n",
      "Epoch[48] Batch [6449]\tSpeed: 128.386702 samples/sec\taccuracy=0.543314\tlr=0.000002\n",
      "Epoch[48] Batch [6499]\tSpeed: 128.503782 samples/sec\taccuracy=0.543250\tlr=0.000002\n",
      "Epoch[48] Batch [6549]\tSpeed: 127.139881 samples/sec\taccuracy=0.542920\tlr=0.000002\n",
      "Epoch[48] Batch [6599]\tSpeed: 124.948759 samples/sec\taccuracy=0.542481\tlr=0.000002\n",
      "Epoch[48] Batch [6649]\tSpeed: 127.993897 samples/sec\taccuracy=0.542350\tlr=0.000002\n",
      "Epoch[48] Batch [6699]\tSpeed: 127.272836 samples/sec\taccuracy=0.542575\tlr=0.000002\n",
      "Epoch[48] Batch [6749]\tSpeed: 125.856724 samples/sec\taccuracy=0.542537\tlr=0.000002\n",
      "Epoch[48] Batch [6799]\tSpeed: 127.078672 samples/sec\taccuracy=0.542831\tlr=0.000002\n",
      "Epoch[48] Batch [6849]\tSpeed: 127.508118 samples/sec\taccuracy=0.542464\tlr=0.000002\n",
      "Epoch[48] Batch [6899]\tSpeed: 126.779475 samples/sec\taccuracy=0.542138\tlr=0.000002\n",
      "Epoch[48] Batch [6949]\tSpeed: 128.117314 samples/sec\taccuracy=0.542086\tlr=0.000002\n",
      "Epoch[48] Batch [6999]\tSpeed: 126.619734 samples/sec\taccuracy=0.542125\tlr=0.000002\n",
      "Epoch[48] Batch [7049]\tSpeed: 128.358874 samples/sec\taccuracy=0.542252\tlr=0.000002\n",
      "Epoch[48] Batch [7099]\tSpeed: 126.253011 samples/sec\taccuracy=0.542412\tlr=0.000002\n",
      "Epoch[48] Batch [7149]\tSpeed: 127.039373 samples/sec\taccuracy=0.542465\tlr=0.000002\n",
      "Epoch[48] Batch [7199]\tSpeed: 129.189736 samples/sec\taccuracy=0.542274\tlr=0.000002\n",
      "Epoch[48] Batch [7249]\tSpeed: 127.803096 samples/sec\taccuracy=0.542517\tlr=0.000002\n",
      "Epoch[48] Batch [7299]\tSpeed: 125.401220 samples/sec\taccuracy=0.542551\tlr=0.000002\n",
      "Epoch[48] Batch [7349]\tSpeed: 129.120584 samples/sec\taccuracy=0.542415\tlr=0.000002\n",
      "Epoch[48] Batch [7399]\tSpeed: 128.953069 samples/sec\taccuracy=0.542264\tlr=0.000002\n",
      "Epoch[48] Batch [7449]\tSpeed: 126.190194 samples/sec\taccuracy=0.542148\tlr=0.000002\n",
      "Epoch[48] Batch [7499]\tSpeed: 129.252658 samples/sec\taccuracy=0.542417\tlr=0.000002\n",
      "Epoch[48] Batch [7699]\tSpeed: 129.519896 samples/sec\taccuracy=0.543003\tlr=0.000002\n",
      "Epoch[48] Batch [7749]\tSpeed: 127.248443 samples/sec\taccuracy=0.543129\tlr=0.000002\n",
      "Epoch[48] Batch [7799]\tSpeed: 127.634484 samples/sec\taccuracy=0.542997\tlr=0.000002\n",
      "Epoch[48] Batch [7849]\tSpeed: 129.998698 samples/sec\taccuracy=0.542866\tlr=0.000002\n",
      "Epoch[48] Batch [7899]\tSpeed: 126.773564 samples/sec\taccuracy=0.542832\tlr=0.000002\n",
      "Epoch[48] Batch [7949]\tSpeed: 119.124417 samples/sec\taccuracy=0.542752\tlr=0.000002\n",
      "Epoch[48] Batch [7999]\tSpeed: 124.449987 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[48] Batch [8049]\tSpeed: 127.275945 samples/sec\taccuracy=0.542780\tlr=0.000002\n",
      "Epoch[48] Batch [8099]\tSpeed: 126.959360 samples/sec\taccuracy=0.542670\tlr=0.000002\n",
      "Epoch[48] Batch [8149]\tSpeed: 126.355685 samples/sec\taccuracy=0.542500\tlr=0.000002\n",
      "Epoch[48] Batch [8199]\tSpeed: 125.973347 samples/sec\taccuracy=0.542637\tlr=0.000002\n",
      "Epoch[48] Batch [8249]\tSpeed: 127.846805 samples/sec\taccuracy=0.542606\tlr=0.000002\n",
      "Epoch[48] Batch [8299]\tSpeed: 127.187216 samples/sec\taccuracy=0.542620\tlr=0.000002\n",
      "Epoch[48] Batch [8349]\tSpeed: 127.375561 samples/sec\taccuracy=0.542560\tlr=0.000002\n",
      "Epoch[48] Batch [8399]\tSpeed: 127.114614 samples/sec\taccuracy=0.542455\tlr=0.000002\n",
      "Epoch[48] Batch [8449]\tSpeed: 128.861710 samples/sec\taccuracy=0.542456\tlr=0.000002\n",
      "Epoch[48] Batch [8499]\tSpeed: 126.112583 samples/sec\taccuracy=0.542588\tlr=0.000002\n",
      "Epoch[48] Batch [8549]\tSpeed: 127.633173 samples/sec\taccuracy=0.542295\tlr=0.000002\n",
      "Epoch[48] Batch [8599]\tSpeed: 129.320107 samples/sec\taccuracy=0.542049\tlr=0.000002\n",
      "Epoch[48] Batch [8649]\tSpeed: 127.163443 samples/sec\taccuracy=0.542009\tlr=0.000002\n",
      "Epoch[48] Batch [8699]\tSpeed: 125.841412 samples/sec\taccuracy=0.541997\tlr=0.000002\n",
      "Epoch[48] Batch [8749]\tSpeed: 127.827099 samples/sec\taccuracy=0.541814\tlr=0.000002\n",
      "Epoch[48] Batch [8799]\tSpeed: 127.329575 samples/sec\taccuracy=0.541804\tlr=0.000002\n",
      "Epoch[48] Batch [8849]\tSpeed: 127.189009 samples/sec\taccuracy=0.541638\tlr=0.000002\n",
      "Epoch[48] Batch [8899]\tSpeed: 127.003435 samples/sec\taccuracy=0.541489\tlr=0.000002\n",
      "Epoch[48] Batch [8949]\tSpeed: 126.143703 samples/sec\taccuracy=0.541578\tlr=0.000002\n",
      "Epoch[48] Batch [8999]\tSpeed: 125.927734 samples/sec\taccuracy=0.541667\tlr=0.000002\n",
      "Epoch[48] Batch [9049]\tSpeed: 128.944447 samples/sec\taccuracy=0.541740\tlr=0.000002\n",
      "Epoch[48] Batch [9099]\tSpeed: 128.793828 samples/sec\taccuracy=0.541566\tlr=0.000002\n",
      "Epoch[48] Batch [9149]\tSpeed: 126.524092 samples/sec\taccuracy=0.541626\tlr=0.000002\n",
      "Epoch[48] Batch [9199]\tSpeed: 128.522023 samples/sec\taccuracy=0.541562\tlr=0.000002\n",
      "Epoch[48] Batch [9249]\tSpeed: 128.904986 samples/sec\taccuracy=0.541595\tlr=0.000002\n",
      "Epoch[48] Batch [9299]\tSpeed: 128.250264 samples/sec\taccuracy=0.541505\tlr=0.000002\n",
      "Epoch[48] Batch [9349]\tSpeed: 129.371862 samples/sec\taccuracy=0.541658\tlr=0.000002\n",
      "Epoch[48] Batch [9399]\tSpeed: 128.677532 samples/sec\taccuracy=0.541649\tlr=0.000002\n",
      "Epoch[48] Batch [9449]\tSpeed: 125.868810 samples/sec\taccuracy=0.541667\tlr=0.000002\n",
      "Epoch[48] Batch [9499]\tSpeed: 129.294394 samples/sec\taccuracy=0.541763\tlr=0.000002\n",
      "Epoch[48] Batch [9549]\tSpeed: 126.685505 samples/sec\taccuracy=0.541767\tlr=0.000002\n",
      "Epoch[48] Batch [9599]\tSpeed: 127.558918 samples/sec\taccuracy=0.541888\tlr=0.000002\n",
      "Epoch[48] Batch [9649]\tSpeed: 126.974907 samples/sec\taccuracy=0.541943\tlr=0.000002\n",
      "Epoch[48] Batch [9699]\tSpeed: 127.801032 samples/sec\taccuracy=0.541985\tlr=0.000002\n",
      "Epoch[48] Batch [9749]\tSpeed: 128.785969 samples/sec\taccuracy=0.542269\tlr=0.000002\n",
      "Epoch[48] Batch [9799]\tSpeed: 127.749408 samples/sec\taccuracy=0.542066\tlr=0.000002\n",
      "Epoch[48] Batch [9849]\tSpeed: 129.598425 samples/sec\taccuracy=0.541802\tlr=0.000002\n",
      "Epoch[48] Batch [9899]\tSpeed: 126.016248 samples/sec\taccuracy=0.541982\tlr=0.000002\n",
      "Epoch[48] Batch [9949]\tSpeed: 127.399239 samples/sec\taccuracy=0.541922\tlr=0.000002\n",
      "Epoch[48] Batch [9999]\tSpeed: 127.945297 samples/sec\taccuracy=0.541850\tlr=0.000002\n",
      "Epoch[48] Batch [10049]\tSpeed: 127.292516 samples/sec\taccuracy=0.541704\tlr=0.000002\n",
      "Epoch[48] Batch [10099]\tSpeed: 126.097522 samples/sec\taccuracy=0.541448\tlr=0.000002\n",
      "Epoch[48] Batch [10149]\tSpeed: 126.884331 samples/sec\taccuracy=0.541404\tlr=0.000002\n",
      "Epoch[48] Batch [10199]\tSpeed: 128.601347 samples/sec\taccuracy=0.541348\tlr=0.000002\n",
      "Epoch[48] Batch [10249]\tSpeed: 128.010919 samples/sec\taccuracy=0.541293\tlr=0.000002\n",
      "Epoch[48] Batch [10299]\tSpeed: 128.773425 samples/sec\taccuracy=0.541481\tlr=0.000002\n",
      "Epoch[48] Batch [10349]\tSpeed: 126.160104 samples/sec\taccuracy=0.541425\tlr=0.000002\n",
      "Epoch[48] Batch [10399]\tSpeed: 127.967430 samples/sec\taccuracy=0.541430\tlr=0.000002\n",
      "Epoch[48] Batch [10449]\tSpeed: 127.397575 samples/sec\taccuracy=0.541435\tlr=0.000002\n",
      "Epoch[48] Batch [10499]\tSpeed: 126.158226 samples/sec\taccuracy=0.541643\tlr=0.000002\n",
      "Epoch[48] Batch [10549]\tSpeed: 126.532689 samples/sec\taccuracy=0.541789\tlr=0.000002\n",
      "Epoch[48] Batch [10599]\tSpeed: 127.132713 samples/sec\taccuracy=0.541781\tlr=0.000002\n",
      "Epoch[48] Batch [10649]\tSpeed: 128.579782 samples/sec\taccuracy=0.541690\tlr=0.000002\n",
      "Epoch[48] Batch [10699]\tSpeed: 127.949502 samples/sec\taccuracy=0.541682\tlr=0.000002\n",
      "Epoch[48] Batch [10749]\tSpeed: 129.594231 samples/sec\taccuracy=0.541802\tlr=0.000002\n",
      "Epoch[48] Batch [10799]\tSpeed: 121.668163 samples/sec\taccuracy=0.541725\tlr=0.000002\n",
      "Epoch[48] Batch [10849]\tSpeed: 122.188549 samples/sec\taccuracy=0.541544\tlr=0.000002\n",
      "Epoch[48] Batch [10899]\tSpeed: 128.061455 samples/sec\taccuracy=0.541594\tlr=0.000002\n",
      "Epoch[48] Batch [10949]\tSpeed: 126.691177 samples/sec\taccuracy=0.541381\tlr=0.000002\n",
      "Epoch[48] Batch [10999]\tSpeed: 128.223123 samples/sec\taccuracy=0.541386\tlr=0.000002\n",
      "Epoch[48] Batch [11049]\tSpeed: 126.934558 samples/sec\taccuracy=0.541437\tlr=0.000002\n",
      "Epoch[48] Batch [11099]\tSpeed: 129.281781 samples/sec\taccuracy=0.541486\tlr=0.000002\n",
      "Epoch[48] Batch [11149]\tSpeed: 126.710305 samples/sec\taccuracy=0.541536\tlr=0.000002\n",
      "Epoch[48] Batch [11199]\tSpeed: 128.114106 samples/sec\taccuracy=0.541540\tlr=0.000002\n",
      "Epoch[48] Batch [11249]\tSpeed: 126.518987 samples/sec\taccuracy=0.541311\tlr=0.000002\n",
      "Epoch[48] Batch [11299]\tSpeed: 126.617440 samples/sec\taccuracy=0.541515\tlr=0.000002\n",
      "Epoch[48] Batch [11349]\tSpeed: 135.588199 samples/sec\taccuracy=0.541388\tlr=0.000002\n",
      "[Epoch 48] training: accuracy=0.541413\n",
      "[Epoch 48] speed: 127 samples/sec\ttime cost: 713.341109\n",
      "[Epoch 48] validation: top1=0.543554 top5=0.877614\n",
      "`time_limit=35999.969309568405` reached, exit early...\n",
      "Applying the state from the best checkpoint...\n",
      "\tTime limit exceeded\n",
      "Saving Training Curve in /home/ec2-user/SageMaker/462d3d04/plot_training_curves.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Finished, total runtime is 37981.02 s\n",
      "{ 'best_config': { 'augmentation': { 'auto_augment': None,\n",
      "                                     'color_jitter': 0.4,\n",
      "                                     'cutmix': 0.0,\n",
      "                                     'cutmix_minmax': None,\n",
      "                                     'drop': 0.0,\n",
      "                                     'drop_block': None,\n",
      "                                     'drop_path': None,\n",
      "                                     'hflip': 0.5,\n",
      "                                     'mixup': 0.0,\n",
      "                                     'mixup_mode': 'batch',\n",
      "                                     'mixup_off_epoch': 0,\n",
      "                                     'mixup_prob': 1.0,\n",
      "                                     'mixup_switch_prob': 0.5,\n",
      "                                     'no_aug': False,\n",
      "                                     'ratio': (0.75, 1.3333333333333333),\n",
      "                                     'scale': (0.08, 1.0),\n",
      "                                     'smoothing': 0.1,\n",
      "                                     'train_interpolation': 'random',\n",
      "                                     'vflip': 0.0},\n",
      "                   'data': { 'crop_pct': 0.99,\n",
      "                             'img_size': None,\n",
      "                             'input_size': None,\n",
      "                             'interpolation': '',\n",
      "                             'mean': None,\n",
      "                             'std': None,\n",
      "                             'validation_batch_size_multiplier': 1},\n",
      "                   'estimator': <class 'gluoncv.auto.estimators.torch_image_classification.torch_image_classification.TorchImageClassificationEstimator'>,\n",
      "                   'gpus': [0],\n",
      "                   'img_cls': { 'global_pool_type': None,\n",
      "                                'model': 'resnet50d',\n",
      "                                'pretrained': True},\n",
      "                   'misc': { 'amp': False,\n",
      "                             'apex_amp': False,\n",
      "                             'eval_metric': 'top1',\n",
      "                             'log_interval': 50,\n",
      "                             'native_amp': False,\n",
      "                             'num_workers': 8,\n",
      "                             'pin_mem': False,\n",
      "                             'prefetcher': False,\n",
      "                             'save_images': False,\n",
      "                             'seed': 676,\n",
      "                             'torchscript': False,\n",
      "                             'tta': 0,\n",
      "                             'use_multi_epochs_loader': False},\n",
      "                   'model_ema': { 'model_ema': True,\n",
      "                                  'model_ema_decay': 0.9998,\n",
      "                                  'model_ema_force_cpu': False},\n",
      "                   'optimizer': { 'clip_grad': None,\n",
      "                                  'clip_mode': 'norm',\n",
      "                                  'momentum': 0.9,\n",
      "                                  'opt': 'sgd',\n",
      "                                  'opt_betas': None,\n",
      "                                  'opt_eps': None,\n",
      "                                  'weight_decay': 0.0001},\n",
      "                   'train': { 'batch_size': 8,\n",
      "                              'bn_eps': None,\n",
      "                              'bn_momentum': None,\n",
      "                              'cooldown_epochs': 10,\n",
      "                              'decay_epochs': 30,\n",
      "                              'decay_rate': 0.1,\n",
      "                              'early_stop_baseline': -inf,\n",
      "                              'early_stop_max_value': inf,\n",
      "                              'early_stop_min_delta': 0.001,\n",
      "                              'early_stop_patience': 50,\n",
      "                              'epochs': 50,\n",
      "                              'lr': 0.0003162278,\n",
      "                              'lr_cycle_limit': 1,\n",
      "                              'lr_cycle_mul': 1.0,\n",
      "                              'lr_noise': None,\n",
      "                              'lr_noise_pct': 0.67,\n",
      "                              'lr_noise_std': 1.0,\n",
      "                              'min_lr': 1e-05,\n",
      "                              'output_lr_mult': 0.1,\n",
      "                              'patience_epochs': 10,\n",
      "                              'sched': 'step',\n",
      "                              'start_epoch': 0,\n",
      "                              'sync_bn': False,\n",
      "                              'transfer_lr_mult': 0.01,\n",
      "                              'warmup_epochs': 3,\n",
      "                              'warmup_lr': 0.0001}},\n",
      "  'total_time': 37980.76288104057,\n",
      "  'train_acc': 0,\n",
      "  'valid_acc': 0.5441482509225651}\n"
     ]
    }
   ],
   "source": [
    "predictor = ImagePredictor()\n",
    "predictor.fit(train, \n",
    "              time_limit=10*60*60,\n",
    "              presets=['best_quality'],\n",
    "              hyperparameters={\n",
    "                  'epochs': 50,\n",
    "                  'model': ag.Categorical('resnet50d', 'efficientnet_b1', 'mobilenetv3_large_100'),\n",
    "              })\n",
    "predictor.save()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ed19e539",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'train_acc': 0,\n",
       " 'valid_acc': 0.5441482509225651,\n",
       " 'total_time': 37980.76288104057,\n",
       " 'best_config': {'img_cls': {'model': 'resnet50d',\n",
       "   'pretrained': True,\n",
       "   'global_pool_type': None},\n",
       "  'data': {'img_size': None,\n",
       "   'input_size': None,\n",
       "   'crop_pct': 0.99,\n",
       "   'mean': None,\n",
       "   'std': None,\n",
       "   'interpolation': '',\n",
       "   'validation_batch_size_multiplier': 1},\n",
       "  'optimizer': {'opt': 'sgd',\n",
       "   'opt_eps': None,\n",
       "   'opt_betas': None,\n",
       "   'momentum': 0.9,\n",
       "   'weight_decay': 0.0001,\n",
       "   'clip_grad': None,\n",
       "   'clip_mode': 'norm'},\n",
       "  'train': {'batch_size': 8,\n",
       "   'sched': 'step',\n",
       "   'lr': 0.0003162278,\n",
       "   'lr_noise': None,\n",
       "   'lr_noise_pct': 0.67,\n",
       "   'lr_noise_std': 1.0,\n",
       "   'lr_cycle_mul': 1.0,\n",
       "   'lr_cycle_limit': 1,\n",
       "   'transfer_lr_mult': 0.01,\n",
       "   'output_lr_mult': 0.1,\n",
       "   'warmup_lr': 0.0001,\n",
       "   'min_lr': 1e-05,\n",
       "   'epochs': 50,\n",
       "   'start_epoch': 0,\n",
       "   'decay_epochs': 30,\n",
       "   'warmup_epochs': 3,\n",
       "   'cooldown_epochs': 10,\n",
       "   'patience_epochs': 10,\n",
       "   'decay_rate': 0.1,\n",
       "   'bn_momentum': None,\n",
       "   'bn_eps': None,\n",
       "   'sync_bn': False,\n",
       "   'early_stop_patience': 50,\n",
       "   'early_stop_min_delta': 0.001,\n",
       "   'early_stop_baseline': -inf,\n",
       "   'early_stop_max_value': inf},\n",
       "  'augmentation': {'no_aug': False,\n",
       "   'scale': (0.08, 1.0),\n",
       "   'ratio': (0.75, 1.3333333333333333),\n",
       "   'hflip': 0.5,\n",
       "   'vflip': 0.0,\n",
       "   'color_jitter': 0.4,\n",
       "   'auto_augment': None,\n",
       "   'mixup': 0.0,\n",
       "   'cutmix': 0.0,\n",
       "   'cutmix_minmax': None,\n",
       "   'mixup_prob': 1.0,\n",
       "   'mixup_switch_prob': 0.5,\n",
       "   'mixup_mode': 'batch',\n",
       "   'mixup_off_epoch': 0,\n",
       "   'smoothing': 0.1,\n",
       "   'train_interpolation': 'random',\n",
       "   'drop': 0.0,\n",
       "   'drop_path': None,\n",
       "   'drop_block': None},\n",
       "  'model_ema': {'model_ema': True,\n",
       "   'model_ema_force_cpu': False,\n",
       "   'model_ema_decay': 0.9998},\n",
       "  'misc': {'seed': 676,\n",
       "   'log_interval': 50,\n",
       "   'num_workers': 8,\n",
       "   'save_images': False,\n",
       "   'amp': False,\n",
       "   'apex_amp': False,\n",
       "   'native_amp': False,\n",
       "   'pin_mem': False,\n",
       "   'prefetcher': False,\n",
       "   'eval_metric': 'top1',\n",
       "   'tta': 0,\n",
       "   'use_multi_epochs_loader': False,\n",
       "   'torchscript': False},\n",
       "  'gpus': [0],\n",
       "  'estimator': gluoncv.auto.estimators.torch_image_classification.torch_image_classification.TorchImageClassificationEstimator},\n",
       " 'fit_history': {'best_reward': 0.5435536616846304,\n",
       "  'best_config': {'batch_size▁choice': 0,\n",
       "   'lr': 0.0003162278,\n",
       "   'model▁choice': 0},\n",
       "  'total_time': 37980.76288104057,\n",
       "  'config_history': OrderedDict([(0,\n",
       "                {'batch_size▁choice': 0,\n",
       "                 'lr': 0.0003162278,\n",
       "                 'model▁choice': 0})]),\n",
       "  'reward_attr': 'acc_reward'}}"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictor.fit_summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "e06ffa1d",
   "metadata": {},
   "outputs": [],
   "source": [
    "predictor.save()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "e321ca04",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ImagePredictor sets accuracy as default eval_metric for classification problems.\n",
      "Converting raw DataFrame to ImageDataset...\n",
      "Detected 15 unique classes: ['Atelectasis', 'Cardiomegaly', 'Consolidation', 'Edema', 'Effusion', 'Emphysema', 'Fibrosis', 'Hernia', 'Infiltration', 'Mass', 'No Finding', 'Nodule', 'Pleural_Thickening', 'Pneumonia', 'Pneumothorax']\n",
      "If you feel the `classes` is inaccurate, please construct the dataset explicitly, e.g. train_data = ImageDataset(train_data, classes=[\"foo\", \"bar\"])\n",
      "Randomly split train_data into train[8112]/validation[902] splits.\n",
      "Reset labels to [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]\n",
      "The number of requested GPUs is greater than the number of available GPUs.Reduce the number to 1\n",
      "Starting fit without HPO\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "Model not found in timm model zoo. Install mxnet if it supports the model.",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-32-8b3f4efda91a>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      7\u001b[0m               hyperparameters={\n\u001b[1;32m      8\u001b[0m                   \u001b[0;34m'epochs'\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;36m50\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m                   \u001b[0;34m'model'\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mag\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mCategorical\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m               })\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/autogluon/vision/configs/presets_configs.py\u001b[0m in \u001b[0;36m_call\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     16\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0m_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     17\u001b[0m             \u001b[0mgargs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mset_presets\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpreset_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 18\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mgargs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mgkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     19\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0m_call\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     20\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0m_unpack_inner\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/autogluon/vision/predictor/predictor.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, train_data, tuning_data, time_limit, presets, hyperparameters, **kwargs)\u001b[0m\n\u001b[1;32m    419\u001b[0m             \u001b[0;31m# TODO: MXNetErrorCatcher was removed because it didn't return traceback\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    420\u001b[0m             \u001b[0;31m#  Re-add once it returns full traceback regardless of which exception was caught\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 421\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_classifier\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtask\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtuning_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mholdout_frac\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrandom_state\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    422\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_classifier\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_logger\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msetLevel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlog_level\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    423\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_classifier\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_logger\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpropagate\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mTrue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/gluoncv/auto/tasks/image_classification.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, train_data, val_data, train_size, random_state, time_limit)\u001b[0m\n\u001b[1;32m    405\u001b[0m             \u001b[0mrand_config\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mRandomSearcher\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_train_image_classification\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_config\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    406\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_logger\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minfo\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Starting fit without HPO\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 407\u001b[0;31m             \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_train_image_classification\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_train_image_classification\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrand_config\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    408\u001b[0m             \u001b[0mbest_config\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msample_config\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_train_image_classification\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrand_config\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    409\u001b[0m             \u001b[0mbest_config\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'train_data'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/autogluon/core/decorator.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, args, config, **kwargs)\u001b[0m\n\u001b[1;32m     58\u001b[0m             \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'reporter'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mFakeReporter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     59\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 60\u001b[0;31m         \u001b[0moutput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     61\u001b[0m         \u001b[0mlogger\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdebug\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Reporter Done!'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     62\u001b[0m         \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'reporter'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdone\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/autogluon/core/decorator.py\u001b[0m in \u001b[0;36mwrapper_call\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    141\u001b[0m         \u001b[0;34m@\u001b[0m\u001b[0mfunctools\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwraps\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfunc\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    142\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0mwrapper_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 143\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    144\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    145\u001b[0m         \u001b[0mdefault\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mkwvars\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'_default_config'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/gluoncv/auto/tasks/image_classification.py\u001b[0m in \u001b[0;36m_train_image_classification\u001b[0;34m(args, reporter)\u001b[0m\n\u001b[1;32m    142\u001b[0m                     \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Model not found in gluoncv model zoo. Install torch and timm if it supports the model.'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    143\u001b[0m                 \u001b[0;32melif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mmxnet_model_list\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 144\u001b[0;31m                     \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Model not found in timm model zoo. Install mxnet if it supports the model.'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    145\u001b[0m                 \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    146\u001b[0m                     \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Model not supported because it does not exist in both timm and gluoncv model zoo.'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mValueError\u001b[0m: Model not found in timm model zoo. Install mxnet if it supports the model."
     ]
    }
   ],
   "source": [
    "model = xrv.models.DenseNet(weights=\"densenet121-res224-nih\")\n",
    "\n",
    "predictor = ImagePredictor()\n",
    "predictor.fit(train, \n",
    "              time_limit=2*60*60,\n",
    "#               presets=['best_quality'],\n",
    "              hyperparameters={\n",
    "                  'epochs': 50,\n",
    "                  'model': ag.Categorical(model),\n",
    "              })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "fdabb9bf",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('adv_inception_v3',\n",
       " 'bat_resnext26ts',\n",
       " 'botnet26t_256',\n",
       " 'botnet50ts_256',\n",
       " 'cait_m36_384',\n",
       " 'cait_m48_448',\n",
       " 'cait_s24_224',\n",
       " 'cait_s24_384',\n",
       " 'cait_s36_384',\n",
       " 'cait_xs24_384',\n",
       " 'cait_xxs24_224',\n",
       " 'cait_xxs24_384',\n",
       " 'cait_xxs36_224',\n",
       " 'cait_xxs36_384',\n",
       " 'coat_lite_mini',\n",
       " 'coat_lite_small',\n",
       " 'coat_lite_tiny',\n",
       " 'coat_mini',\n",
       " 'coat_tiny',\n",
       " 'convit_base',\n",
       " 'convit_small',\n",
       " 'convit_tiny',\n",
       " 'cspdarknet53',\n",
       " 'cspdarknet53_iabn',\n",
       " 'cspresnet50',\n",
       " 'cspresnet50d',\n",
       " 'cspresnet50w',\n",
       " 'cspresnext50',\n",
       " 'cspresnext50_iabn',\n",
       " 'darknet53',\n",
       " 'deit_base_distilled_patch16_224',\n",
       " 'deit_base_distilled_patch16_384',\n",
       " 'deit_base_patch16_224',\n",
       " 'deit_base_patch16_384',\n",
       " 'deit_small_distilled_patch16_224',\n",
       " 'deit_small_patch16_224',\n",
       " 'deit_tiny_distilled_patch16_224',\n",
       " 'deit_tiny_patch16_224',\n",
       " 'densenet121',\n",
       " 'densenet121d',\n",
       " 'densenet161',\n",
       " 'densenet169',\n",
       " 'densenet201',\n",
       " 'densenet264',\n",
       " 'densenet264d_iabn',\n",
       " 'densenetblur121d',\n",
       " 'dla34',\n",
       " 'dla46_c',\n",
       " 'dla46x_c',\n",
       " 'dla60',\n",
       " 'dla60_res2net',\n",
       " 'dla60_res2next',\n",
       " 'dla60x',\n",
       " 'dla60x_c',\n",
       " 'dla102',\n",
       " 'dla102x',\n",
       " 'dla102x2',\n",
       " 'dla169',\n",
       " 'dm_nfnet_f0',\n",
       " 'dm_nfnet_f1',\n",
       " 'dm_nfnet_f2',\n",
       " 'dm_nfnet_f3',\n",
       " 'dm_nfnet_f4',\n",
       " 'dm_nfnet_f5',\n",
       " 'dm_nfnet_f6',\n",
       " 'dpn68',\n",
       " 'dpn68b',\n",
       " 'dpn92',\n",
       " 'dpn98',\n",
       " 'dpn107',\n",
       " 'dpn131',\n",
       " 'eca_botnext26ts_256',\n",
       " 'eca_efficientnet_b0',\n",
       " 'eca_halonext26ts',\n",
       " 'eca_lambda_resnext26ts',\n",
       " 'eca_nfnet_l0',\n",
       " 'eca_nfnet_l1',\n",
       " 'eca_nfnet_l2',\n",
       " 'eca_nfnet_l3',\n",
       " 'eca_swinnext26ts_256',\n",
       " 'eca_vovnet39b',\n",
       " 'ecaresnet26t',\n",
       " 'ecaresnet50d',\n",
       " 'ecaresnet50d_pruned',\n",
       " 'ecaresnet50t',\n",
       " 'ecaresnet101d',\n",
       " 'ecaresnet101d_pruned',\n",
       " 'ecaresnet200d',\n",
       " 'ecaresnet269d',\n",
       " 'ecaresnetlight',\n",
       " 'ecaresnext26t_32x4d',\n",
       " 'ecaresnext50t_32x4d',\n",
       " 'efficientnet_b0',\n",
       " 'efficientnet_b1',\n",
       " 'efficientnet_b1_pruned',\n",
       " 'efficientnet_b2',\n",
       " 'efficientnet_b2_pruned',\n",
       " 'efficientnet_b2a',\n",
       " 'efficientnet_b3',\n",
       " 'efficientnet_b3_pruned',\n",
       " 'efficientnet_b3a',\n",
       " 'efficientnet_b4',\n",
       " 'efficientnet_b5',\n",
       " 'efficientnet_b6',\n",
       " 'efficientnet_b7',\n",
       " 'efficientnet_b8',\n",
       " 'efficientnet_cc_b0_4e',\n",
       " 'efficientnet_cc_b0_8e',\n",
       " 'efficientnet_cc_b1_8e',\n",
       " 'efficientnet_el',\n",
       " 'efficientnet_el_pruned',\n",
       " 'efficientnet_em',\n",
       " 'efficientnet_es',\n",
       " 'efficientnet_es_pruned',\n",
       " 'efficientnet_l2',\n",
       " 'efficientnet_lite0',\n",
       " 'efficientnet_lite1',\n",
       " 'efficientnet_lite2',\n",
       " 'efficientnet_lite3',\n",
       " 'efficientnet_lite4',\n",
       " 'efficientnetv2_l',\n",
       " 'efficientnetv2_m',\n",
       " 'efficientnetv2_rw_m',\n",
       " 'efficientnetv2_rw_s',\n",
       " 'efficientnetv2_s',\n",
       " 'ens_adv_inception_resnet_v2',\n",
       " 'ese_vovnet19b_dw',\n",
       " 'ese_vovnet19b_slim',\n",
       " 'ese_vovnet19b_slim_dw',\n",
       " 'ese_vovnet39b',\n",
       " 'ese_vovnet39b_evos',\n",
       " 'ese_vovnet57b',\n",
       " 'ese_vovnet99b',\n",
       " 'ese_vovnet99b_iabn',\n",
       " 'fbnetc_100',\n",
       " 'fbnetv3_b',\n",
       " 'fbnetv3_d',\n",
       " 'fbnetv3_g',\n",
       " 'gc_efficientnet_b0',\n",
       " 'gcresnet50t',\n",
       " 'gcresnext26ts',\n",
       " 'geresnet50t',\n",
       " 'gernet_l',\n",
       " 'gernet_m',\n",
       " 'gernet_s',\n",
       " 'ghostnet_050',\n",
       " 'ghostnet_100',\n",
       " 'ghostnet_130',\n",
       " 'gluon_inception_v3',\n",
       " 'gluon_resnet18_v1b',\n",
       " 'gluon_resnet34_v1b',\n",
       " 'gluon_resnet50_v1b',\n",
       " 'gluon_resnet50_v1c',\n",
       " 'gluon_resnet50_v1d',\n",
       " 'gluon_resnet50_v1s',\n",
       " 'gluon_resnet101_v1b',\n",
       " 'gluon_resnet101_v1c',\n",
       " 'gluon_resnet101_v1d',\n",
       " 'gluon_resnet101_v1s',\n",
       " 'gluon_resnet152_v1b',\n",
       " 'gluon_resnet152_v1c',\n",
       " 'gluon_resnet152_v1d',\n",
       " 'gluon_resnet152_v1s',\n",
       " 'gluon_resnext50_32x4d',\n",
       " 'gluon_resnext101_32x4d',\n",
       " 'gluon_resnext101_64x4d',\n",
       " 'gluon_senet154',\n",
       " 'gluon_seresnext50_32x4d',\n",
       " 'gluon_seresnext101_32x4d',\n",
       " 'gluon_seresnext101_64x4d',\n",
       " 'gluon_xception65',\n",
       " 'gmixer_12_224',\n",
       " 'gmixer_24_224',\n",
       " 'gmlp_b16_224',\n",
       " 'gmlp_s16_224',\n",
       " 'gmlp_ti16_224',\n",
       " 'halonet26t',\n",
       " 'halonet50ts',\n",
       " 'halonet_h1',\n",
       " 'halonet_h1_c4c5',\n",
       " 'hardcorenas_a',\n",
       " 'hardcorenas_b',\n",
       " 'hardcorenas_c',\n",
       " 'hardcorenas_d',\n",
       " 'hardcorenas_e',\n",
       " 'hardcorenas_f',\n",
       " 'hrnet_w18',\n",
       " 'hrnet_w18_small',\n",
       " 'hrnet_w18_small_v2',\n",
       " 'hrnet_w30',\n",
       " 'hrnet_w32',\n",
       " 'hrnet_w40',\n",
       " 'hrnet_w44',\n",
       " 'hrnet_w48',\n",
       " 'hrnet_w64',\n",
       " 'ig_resnext101_32x8d',\n",
       " 'ig_resnext101_32x16d',\n",
       " 'ig_resnext101_32x32d',\n",
       " 'ig_resnext101_32x48d',\n",
       " 'inception_resnet_v2',\n",
       " 'inception_v3',\n",
       " 'inception_v4',\n",
       " 'lambda_resnet26t',\n",
       " 'lambda_resnet50t',\n",
       " 'legacy_senet154',\n",
       " 'legacy_seresnet18',\n",
       " 'legacy_seresnet34',\n",
       " 'legacy_seresnet50',\n",
       " 'legacy_seresnet101',\n",
       " 'legacy_seresnet152',\n",
       " 'legacy_seresnext26_32x4d',\n",
       " 'legacy_seresnext50_32x4d',\n",
       " 'legacy_seresnext101_32x4d',\n",
       " 'levit_128',\n",
       " 'levit_128s',\n",
       " 'levit_192',\n",
       " 'levit_256',\n",
       " 'levit_384',\n",
       " 'mixer_b16_224',\n",
       " 'mixer_b16_224_in21k',\n",
       " 'mixer_b16_224_miil',\n",
       " 'mixer_b16_224_miil_in21k',\n",
       " 'mixer_b32_224',\n",
       " 'mixer_l16_224',\n",
       " 'mixer_l16_224_in21k',\n",
       " 'mixer_l32_224',\n",
       " 'mixer_s16_224',\n",
       " 'mixer_s32_224',\n",
       " 'mixnet_l',\n",
       " 'mixnet_m',\n",
       " 'mixnet_s',\n",
       " 'mixnet_xl',\n",
       " 'mixnet_xxl',\n",
       " 'mnasnet_050',\n",
       " 'mnasnet_075',\n",
       " 'mnasnet_100',\n",
       " 'mnasnet_140',\n",
       " 'mnasnet_a1',\n",
       " 'mnasnet_b1',\n",
       " 'mnasnet_small',\n",
       " 'mobilenetv2_100',\n",
       " 'mobilenetv2_110d',\n",
       " 'mobilenetv2_120d',\n",
       " 'mobilenetv2_140',\n",
       " 'mobilenetv3_large_075',\n",
       " 'mobilenetv3_large_100',\n",
       " 'mobilenetv3_large_100_miil',\n",
       " 'mobilenetv3_large_100_miil_in21k',\n",
       " 'mobilenetv3_rw',\n",
       " 'mobilenetv3_small_075',\n",
       " 'mobilenetv3_small_100',\n",
       " 'nasnetalarge',\n",
       " 'nf_ecaresnet26',\n",
       " 'nf_ecaresnet50',\n",
       " 'nf_ecaresnet101',\n",
       " 'nf_regnet_b0',\n",
       " 'nf_regnet_b1',\n",
       " 'nf_regnet_b2',\n",
       " 'nf_regnet_b3',\n",
       " 'nf_regnet_b4',\n",
       " 'nf_regnet_b5',\n",
       " 'nf_resnet26',\n",
       " 'nf_resnet50',\n",
       " 'nf_resnet101',\n",
       " 'nf_seresnet26',\n",
       " 'nf_seresnet50',\n",
       " 'nf_seresnet101',\n",
       " 'nfnet_f0',\n",
       " 'nfnet_f0s',\n",
       " 'nfnet_f1',\n",
       " 'nfnet_f1s',\n",
       " 'nfnet_f2',\n",
       " 'nfnet_f2s',\n",
       " 'nfnet_f3',\n",
       " 'nfnet_f3s',\n",
       " 'nfnet_f4',\n",
       " 'nfnet_f4s',\n",
       " 'nfnet_f5',\n",
       " 'nfnet_f5s',\n",
       " 'nfnet_f6',\n",
       " 'nfnet_f6s',\n",
       " 'nfnet_f7',\n",
       " 'nfnet_f7s',\n",
       " 'nfnet_l0',\n",
       " 'pit_b_224',\n",
       " 'pit_b_distilled_224',\n",
       " 'pit_s_224',\n",
       " 'pit_s_distilled_224',\n",
       " 'pit_ti_224',\n",
       " 'pit_ti_distilled_224',\n",
       " 'pit_xs_224',\n",
       " 'pit_xs_distilled_224',\n",
       " 'pnasnet5large',\n",
       " 'rednet26t',\n",
       " 'rednet50ts',\n",
       " 'regnetx_002',\n",
       " 'regnetx_004',\n",
       " 'regnetx_006',\n",
       " 'regnetx_008',\n",
       " 'regnetx_016',\n",
       " 'regnetx_032',\n",
       " 'regnetx_040',\n",
       " 'regnetx_064',\n",
       " 'regnetx_080',\n",
       " 'regnetx_120',\n",
       " 'regnetx_160',\n",
       " 'regnetx_320',\n",
       " 'regnety_002',\n",
       " 'regnety_004',\n",
       " 'regnety_006',\n",
       " 'regnety_008',\n",
       " 'regnety_016',\n",
       " 'regnety_032',\n",
       " 'regnety_040',\n",
       " 'regnety_064',\n",
       " 'regnety_080',\n",
       " 'regnety_120',\n",
       " 'regnety_160',\n",
       " 'regnety_320',\n",
       " 'repvgg_a2',\n",
       " 'repvgg_b0',\n",
       " 'repvgg_b1',\n",
       " 'repvgg_b1g4',\n",
       " 'repvgg_b2',\n",
       " 'repvgg_b2g4',\n",
       " 'repvgg_b3',\n",
       " 'repvgg_b3g4',\n",
       " 'res2net50_14w_8s',\n",
       " 'res2net50_26w_4s',\n",
       " 'res2net50_26w_6s',\n",
       " 'res2net50_26w_8s',\n",
       " 'res2net50_48w_2s',\n",
       " 'res2net101_26w_4s',\n",
       " 'res2next50',\n",
       " 'resmlp_12_224',\n",
       " 'resmlp_12_distilled_224',\n",
       " 'resmlp_24_224',\n",
       " 'resmlp_24_distilled_224',\n",
       " 'resmlp_36_224',\n",
       " 'resmlp_36_distilled_224',\n",
       " 'resmlp_big_24_224',\n",
       " 'resmlp_big_24_224_in22ft1k',\n",
       " 'resmlp_big_24_distilled_224',\n",
       " 'resnest14d',\n",
       " 'resnest26d',\n",
       " 'resnest50d',\n",
       " 'resnest50d_1s4x24d',\n",
       " 'resnest50d_4s2x40d',\n",
       " 'resnest101e',\n",
       " 'resnest200e',\n",
       " 'resnest269e',\n",
       " 'resnet18',\n",
       " 'resnet18d',\n",
       " 'resnet26',\n",
       " 'resnet26d',\n",
       " 'resnet26t',\n",
       " 'resnet34',\n",
       " 'resnet34d',\n",
       " 'resnet50',\n",
       " 'resnet50d',\n",
       " 'resnet50t',\n",
       " 'resnet51q',\n",
       " 'resnet61q',\n",
       " 'resnet101',\n",
       " 'resnet101d',\n",
       " 'resnet152',\n",
       " 'resnet152d',\n",
       " 'resnet200',\n",
       " 'resnet200d',\n",
       " 'resnetblur18',\n",
       " 'resnetblur50',\n",
       " 'resnetrs50',\n",
       " 'resnetrs101',\n",
       " 'resnetrs152',\n",
       " 'resnetrs200',\n",
       " 'resnetrs270',\n",
       " 'resnetrs350',\n",
       " 'resnetrs420',\n",
       " 'resnetv2_50',\n",
       " 'resnetv2_50d',\n",
       " 'resnetv2_50t',\n",
       " 'resnetv2_50x1_bit_distilled',\n",
       " 'resnetv2_50x1_bitm',\n",
       " 'resnetv2_50x1_bitm_in21k',\n",
       " 'resnetv2_50x3_bitm',\n",
       " 'resnetv2_50x3_bitm_in21k',\n",
       " 'resnetv2_101',\n",
       " 'resnetv2_101d',\n",
       " 'resnetv2_101x1_bitm',\n",
       " 'resnetv2_101x1_bitm_in21k',\n",
       " 'resnetv2_101x3_bitm',\n",
       " 'resnetv2_101x3_bitm_in21k',\n",
       " 'resnetv2_152',\n",
       " 'resnetv2_152d',\n",
       " 'resnetv2_152x2_bit_teacher',\n",
       " 'resnetv2_152x2_bit_teacher_384',\n",
       " 'resnetv2_152x2_bitm',\n",
       " 'resnetv2_152x2_bitm_in21k',\n",
       " 'resnetv2_152x4_bitm',\n",
       " 'resnetv2_152x4_bitm_in21k',\n",
       " 'resnext50_32x4d',\n",
       " 'resnext50d_32x4d',\n",
       " 'resnext101_32x4d',\n",
       " 'resnext101_32x8d',\n",
       " 'resnext101_64x4d',\n",
       " 'rexnet_100',\n",
       " 'rexnet_130',\n",
       " 'rexnet_150',\n",
       " 'rexnet_200',\n",
       " 'rexnetr_100',\n",
       " 'rexnetr_130',\n",
       " 'rexnetr_150',\n",
       " 'rexnetr_200',\n",
       " 'selecsls42',\n",
       " 'selecsls42b',\n",
       " 'selecsls60',\n",
       " 'selecsls60b',\n",
       " 'selecsls84',\n",
       " 'semnasnet_050',\n",
       " 'semnasnet_075',\n",
       " 'semnasnet_100',\n",
       " 'semnasnet_140',\n",
       " 'senet154',\n",
       " 'seresnet18',\n",
       " 'seresnet34',\n",
       " 'seresnet50',\n",
       " 'seresnet50t',\n",
       " 'seresnet101',\n",
       " 'seresnet152',\n",
       " 'seresnet152d',\n",
       " 'seresnet200d',\n",
       " 'seresnet269d',\n",
       " 'seresnext26d_32x4d',\n",
       " 'seresnext26t_32x4d',\n",
       " 'seresnext26tn_32x4d',\n",
       " 'seresnext50_32x4d',\n",
       " 'seresnext101_32x4d',\n",
       " 'seresnext101_32x8d',\n",
       " 'skresnet18',\n",
       " 'skresnet34',\n",
       " 'skresnet50',\n",
       " 'skresnet50d',\n",
       " 'skresnext50_32x4d',\n",
       " 'spnasnet_100',\n",
       " 'ssl_resnet18',\n",
       " 'ssl_resnet50',\n",
       " 'ssl_resnext50_32x4d',\n",
       " 'ssl_resnext101_32x4d',\n",
       " 'ssl_resnext101_32x8d',\n",
       " 'ssl_resnext101_32x16d',\n",
       " 'swin_base_patch4_window7_224',\n",
       " 'swin_base_patch4_window7_224_in22k',\n",
       " 'swin_base_patch4_window12_384',\n",
       " 'swin_base_patch4_window12_384_in22k',\n",
       " 'swin_large_patch4_window7_224',\n",
       " 'swin_large_patch4_window7_224_in22k',\n",
       " 'swin_large_patch4_window12_384',\n",
       " 'swin_large_patch4_window12_384_in22k',\n",
       " 'swin_small_patch4_window7_224',\n",
       " 'swin_tiny_patch4_window7_224',\n",
       " 'swinnet26t_256',\n",
       " 'swinnet50ts_256',\n",
       " 'swsl_resnet18',\n",
       " 'swsl_resnet50',\n",
       " 'swsl_resnext50_32x4d',\n",
       " 'swsl_resnext101_32x4d',\n",
       " 'swsl_resnext101_32x8d',\n",
       " 'swsl_resnext101_32x16d',\n",
       " 'tf_efficientnet_b0',\n",
       " 'tf_efficientnet_b0_ap',\n",
       " 'tf_efficientnet_b0_ns',\n",
       " 'tf_efficientnet_b1',\n",
       " 'tf_efficientnet_b1_ap',\n",
       " 'tf_efficientnet_b1_ns',\n",
       " 'tf_efficientnet_b2',\n",
       " 'tf_efficientnet_b2_ap',\n",
       " 'tf_efficientnet_b2_ns',\n",
       " 'tf_efficientnet_b3',\n",
       " 'tf_efficientnet_b3_ap',\n",
       " 'tf_efficientnet_b3_ns',\n",
       " 'tf_efficientnet_b4',\n",
       " 'tf_efficientnet_b4_ap',\n",
       " 'tf_efficientnet_b4_ns',\n",
       " 'tf_efficientnet_b5',\n",
       " 'tf_efficientnet_b5_ap',\n",
       " 'tf_efficientnet_b5_ns',\n",
       " 'tf_efficientnet_b6',\n",
       " 'tf_efficientnet_b6_ap',\n",
       " 'tf_efficientnet_b6_ns',\n",
       " 'tf_efficientnet_b7',\n",
       " 'tf_efficientnet_b7_ap',\n",
       " 'tf_efficientnet_b7_ns',\n",
       " 'tf_efficientnet_b8',\n",
       " 'tf_efficientnet_b8_ap',\n",
       " 'tf_efficientnet_cc_b0_4e',\n",
       " 'tf_efficientnet_cc_b0_8e',\n",
       " 'tf_efficientnet_cc_b1_8e',\n",
       " 'tf_efficientnet_el',\n",
       " 'tf_efficientnet_em',\n",
       " 'tf_efficientnet_es',\n",
       " 'tf_efficientnet_l2_ns',\n",
       " 'tf_efficientnet_l2_ns_475',\n",
       " 'tf_efficientnet_lite0',\n",
       " 'tf_efficientnet_lite1',\n",
       " 'tf_efficientnet_lite2',\n",
       " 'tf_efficientnet_lite3',\n",
       " 'tf_efficientnet_lite4',\n",
       " 'tf_efficientnetv2_b0',\n",
       " 'tf_efficientnetv2_b1',\n",
       " 'tf_efficientnetv2_b2',\n",
       " 'tf_efficientnetv2_b3',\n",
       " 'tf_efficientnetv2_l',\n",
       " 'tf_efficientnetv2_l_in21ft1k',\n",
       " 'tf_efficientnetv2_l_in21k',\n",
       " 'tf_efficientnetv2_m',\n",
       " 'tf_efficientnetv2_m_in21ft1k',\n",
       " 'tf_efficientnetv2_m_in21k',\n",
       " 'tf_efficientnetv2_s',\n",
       " 'tf_efficientnetv2_s_in21ft1k',\n",
       " 'tf_efficientnetv2_s_in21k',\n",
       " 'tf_inception_v3',\n",
       " 'tf_mixnet_l',\n",
       " 'tf_mixnet_m',\n",
       " 'tf_mixnet_s',\n",
       " 'tf_mobilenetv3_large_075',\n",
       " 'tf_mobilenetv3_large_100',\n",
       " 'tf_mobilenetv3_large_minimal_100',\n",
       " 'tf_mobilenetv3_small_075',\n",
       " 'tf_mobilenetv3_small_100',\n",
       " 'tf_mobilenetv3_small_minimal_100',\n",
       " 'tnt_b_patch16_224',\n",
       " 'tnt_s_patch16_224',\n",
       " 'tresnet_l',\n",
       " 'tresnet_l_448',\n",
       " 'tresnet_m',\n",
       " 'tresnet_m_448',\n",
       " 'tresnet_m_miil_in21k',\n",
       " 'tresnet_xl',\n",
       " 'tresnet_xl_448',\n",
       " 'tv_densenet121',\n",
       " 'tv_resnet34',\n",
       " 'tv_resnet50',\n",
       " 'tv_resnet101',\n",
       " 'tv_resnet152',\n",
       " 'tv_resnext50_32x4d',\n",
       " 'twins_pcpvt_base',\n",
       " 'twins_pcpvt_large',\n",
       " 'twins_pcpvt_small',\n",
       " 'twins_svt_base',\n",
       " 'twins_svt_large',\n",
       " 'twins_svt_small',\n",
       " 'vgg11',\n",
       " 'vgg11_bn',\n",
       " 'vgg13',\n",
       " 'vgg13_bn',\n",
       " 'vgg16',\n",
       " 'vgg16_bn',\n",
       " 'vgg19',\n",
       " 'vgg19_bn',\n",
       " 'visformer_small',\n",
       " 'visformer_tiny',\n",
       " 'vit_base_patch16_224',\n",
       " 'vit_base_patch16_224_in21k',\n",
       " 'vit_base_patch16_224_miil',\n",
       " 'vit_base_patch16_224_miil_in21k',\n",
       " 'vit_base_patch16_384',\n",
       " 'vit_base_patch32_224',\n",
       " 'vit_base_patch32_224_in21k',\n",
       " 'vit_base_patch32_384',\n",
       " 'vit_base_r26_s32_224',\n",
       " 'vit_base_r50_s16_224',\n",
       " 'vit_base_r50_s16_224_in21k',\n",
       " 'vit_base_r50_s16_384',\n",
       " 'vit_base_resnet26d_224',\n",
       " 'vit_base_resnet50_224_in21k',\n",
       " 'vit_base_resnet50_384',\n",
       " 'vit_base_resnet50d_224',\n",
       " 'vit_huge_patch14_224_in21k',\n",
       " 'vit_large_patch16_224',\n",
       " 'vit_large_patch16_224_in21k',\n",
       " 'vit_large_patch16_384',\n",
       " 'vit_large_patch32_224',\n",
       " 'vit_large_patch32_224_in21k',\n",
       " 'vit_large_patch32_384',\n",
       " 'vit_large_r50_s32_224',\n",
       " 'vit_large_r50_s32_224_in21k',\n",
       " 'vit_large_r50_s32_384',\n",
       " 'vit_small_patch16_224',\n",
       " 'vit_small_patch16_224_in21k',\n",
       " 'vit_small_patch16_384',\n",
       " 'vit_small_patch32_224',\n",
       " 'vit_small_patch32_224_in21k',\n",
       " 'vit_small_patch32_384',\n",
       " 'vit_small_r26_s32_224',\n",
       " 'vit_small_r26_s32_224_in21k',\n",
       " 'vit_small_r26_s32_384',\n",
       " 'vit_small_resnet26d_224',\n",
       " 'vit_small_resnet50d_s16_224',\n",
       " 'vit_tiny_patch16_224',\n",
       " 'vit_tiny_patch16_224_in21k',\n",
       " 'vit_tiny_patch16_384',\n",
       " 'vit_tiny_r_s16_p8_224',\n",
       " 'vit_tiny_r_s16_p8_224_in21k',\n",
       " 'vit_tiny_r_s16_p8_384',\n",
       " 'vovnet39a',\n",
       " 'vovnet57a',\n",
       " 'wide_resnet50_2',\n",
       " 'wide_resnet101_2',\n",
       " 'xception',\n",
       " 'xception41',\n",
       " 'xception65',\n",
       " 'xception71')"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictor.list_models()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "abe86165",
   "metadata": {},
   "source": [
    "## Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "f57a50e5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   Atelectasis  Cardiomegaly  Consolidation     Edema  Effusion  Emphysema  \\\n",
      "0      0.03743      0.020337       0.027051  0.014201   0.05837   0.014831   \n",
      "\n",
      "   Fibrosis    Hernia  Infiltration      Mass  No Finding    Nodule  \\\n",
      "0  0.029753  0.009042      0.188618  0.038629    0.451636  0.041629   \n",
      "\n",
      "   Pleural_Thickening  Pneumonia  Pneumothorax  \n",
      "0            0.019874   0.012368      0.036231  \n"
     ]
    }
   ],
   "source": [
    "image_path = test.iloc[0]['image']\n",
    "result = predictor.predict_proba(image_path)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "b8defe2a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "AttributeError",
     "evalue": "'tuple' object has no attribute 'to'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-55-22f6af3b0dce>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mpredictor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtest\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/autogluon/vision/predictor/predictor.py\u001b[0m in \u001b[0;36mevaluate\u001b[0;34m(self, data)\u001b[0m\n\u001b[1;32m    646\u001b[0m             \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    647\u001b[0m                 \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_ImageClassification\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDataset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mclasses\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 648\u001b[0;31m         \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_classifier\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmetric_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_eval_metric\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    649\u001b[0m         \u001b[0;31m# TODO: remove the switch if mxnet is deprecated\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    650\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mret\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdict\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/gluoncv/auto/estimators/torch_image_classification/torch_image_classification.py\u001b[0m in \u001b[0;36mevaluate\u001b[0;34m(self, val_data, metric_name)\u001b[0m\n\u001b[1;32m    604\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    605\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmetric_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 606\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_evaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mval_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmetric_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    607\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    608\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_evaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmetric_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/gluoncv/auto/estimators/torch_image_classification/torch_image_classification.py\u001b[0m in \u001b[0;36m_evaluate\u001b[0;34m(self, val_data, metric_name)\u001b[0m\n\u001b[1;32m    627\u001b[0m             \u001b[0mpin_memory\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_misc_cfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpin_mem\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    628\u001b[0m         )\n\u001b[0;32m--> 629\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalidate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnet\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_loader\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalidate_loss_fn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mamp_autocast\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_amp_autocast\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmetric_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmetric_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    630\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    631\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_predict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/gluoncv/auto/estimators/torch_image_classification/torch_image_classification.py\u001b[0m in \u001b[0;36mvalidate\u001b[0;34m(self, net, loader, loss_fn, amp_autocast, metric_name)\u001b[0m\n\u001b[1;32m    424\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_misc_cfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprefetcher\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    425\u001b[0m                     \u001b[0minput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mctx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 426\u001b[0;31m                     \u001b[0mtarget\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtarget\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mctx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    427\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    428\u001b[0m                 \u001b[0;32mwith\u001b[0m \u001b[0mamp_autocast\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mAttributeError\u001b[0m: 'tuple' object has no attribute 'to'"
     ]
    }
   ],
   "source": [
    "predictor.evaluate(test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5cd2713a",
   "metadata": {},
   "source": [
    "## Precision, Recall, F1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "13d3176a",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_pytorch_p36",
   "language": "python",
   "name": "conda_pytorch_p36"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}