First_Contact_with_MindSpore.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "xLRAOOqzdpyj"
   },
   "source": [
    "# Programming Lenet5 with MindSpore"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "PsBAWbnpdpyk"
   },
   "source": [
    "## 0. Configuring the Running Information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "QyZ3gcbKdpym"
   },
   "outputs": [],
   "source": [
    "from mindspore import context\n",
    "\n",
    "context.set_context(mode=context.GRAPH_MODE, device_target='CPU', enable_mem_reuse=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "5TeRJwsfdpyt"
   },
   "source": [
    "## 1. Downloading and preprocessing the Dataset\n",
    "### Download the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "zN8MtJ9_dpyu"
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import urllib.request\n",
    "from urllib.parse import urlparse\n",
    "import gzip\n",
    "\n",
    "def unzipfile(gzip_path):\n",
    "    open_file = open(gzip_path.replace('.gz',''), 'wb')\n",
    "    gz_file = gzip.GzipFile(gzip_path)\n",
    "    open_file.write(gz_file.read())\n",
    "    gz_file.close()\n",
    "\n",
    "def download_dataset():\n",
    "    train_path = \"./MNIST_Data/train/\"\n",
    "    test_path = \"./MNIST_Data/test/\"\n",
    "    train_path_check = os.path.exists(\"./MNIST_Data/train/\")\n",
    "    test_path_check = os.path.exists(\"./MNIST_Data/test/\")\n",
    "    if train_path_check == False and test_path_check ==False:\n",
    "        os.makedirs(train_path)\n",
    "        os.makedirs(test_path)\n",
    "    train_url = {\"http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz\", \"http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz\"}\n",
    "    test_url = {\"http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz\", \"http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz\"}\n",
    "    for url in train_url:\n",
    "        url_parse = urlparse(url)\n",
    "        # split the file name from url\n",
    "        file_name = os.path.join(train_path,url_parse.path.split('/')[-1])\n",
    "        if not os.path.exists(file_name.replace('.gz','')):\n",
    "            file = urllib.request.urlretrieve(url, file_name)\n",
    "            unzipfile(file_name)\n",
    "            os.remove(file_name)\n",
    "    for url in test_url:\n",
    "        url_parse = urlparse(url)\n",
    "        # split the file name from url\n",
    "        file_name = os.path.join(test_path,url_parse.path.split('/')[-1])\n",
    "        if not os.path.exists(file_name.replace('.gz','')):\n",
    "            file = urllib.request.urlretrieve(url, file_name)\n",
    "            unzipfile(file_name)\n",
    "            os.remove(file_name)\n",
    "\n",
    "download_dataset()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Hfemxo7Gdpyz"
   },
   "source": [
    "### Processing Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ugyHt6Okdpy0",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import mindspore.dataset.transforms.vision.c_transforms as CV\n",
    "import mindspore.dataset.transforms.c_transforms as C\n",
    "from mindspore.dataset.transforms.vision import Inter\n",
    "from mindspore.common import dtype as mstype\n",
    "\n",
    "import mindspore.dataset as ds\n",
    "\n",
    "\n",
    "def processing_data(data_path, batch_size=32):\n",
    "\n",
    "    # define dataset\n",
    "    mnist_ds = ds.MnistDataset(data_path)\n",
    "\n",
    "    # define operation parameters\n",
    "    resize_height, resize_width = 32, 32\n",
    "    rescale = 1.0 / 255.0\n",
    "    shift = 0.0\n",
    "    rescale_nml = 1 / 0.3081\n",
    "    shift_nml = -1 * 0.1307 / 0.3081\n",
    "\n",
    "   # define map operations\n",
    "    resize_op =      CV.Resize((resize_height, resize_width), interpolation=Inter.LINEAR)  # Resize images to (32, 32)\n",
    "    rescale_nml_op = CV.Rescale(rescale_nml, shift_nml) # normalize images\n",
    "    rescale_op =     CV.Rescale(rescale, shift) # rescale images\n",
    "    hwc2chw_op =     CV.HWC2CHW() # change shape from (height, width, channel) to (channel, height, width) to fit network.\n",
    "    type_cast_op =   C.TypeCast(mstype.int32) # change data type of label to int32 to fit network\n",
    "\n",
    "    # apply map operations on images\n",
    "    mnist_ds = mnist_ds.map(input_columns=\"label\", operations=type_cast_op)\n",
    "    mnist_ds = mnist_ds.map(input_columns=\"image\", operations=resize_op)\n",
    "    mnist_ds = mnist_ds.map(input_columns=\"image\", operations=rescale_op)\n",
    "    mnist_ds = mnist_ds.map(input_columns=\"image\", operations=rescale_nml_op)\n",
    "    mnist_ds = mnist_ds.map(input_columns=\"image\", operations=hwc2chw_op)\n",
    "\n",
    "    # apply DatasetOps\n",
    "    mnist_ds = mnist_ds.shuffle(buffer_size=10000)\n",
    "    mnist_ds = mnist_ds.batch(batch_size, drop_remainder=True)\n",
    "\n",
    "    return mnist_ds\n",
    "\n",
    "ds_train = processing_data(\"./MNIST_Data/train\", 32)\n",
    "ds_eval  = processing_data(\"./MNIST_Data/test\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "T-wadLHBdpy4"
   },
   "source": [
    "### Check the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "g1ncs0oqdpy4",
    "outputId": "22c11cc4-3567-4723-ab97-3fa0a08ff23d",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 20 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "iterator = ds_train.create_dict_iterator()\n",
    "item = next (iterator)\n",
    "\n",
    "image = np.squeeze(item[\"image\"],1)\n",
    "label = item[\"label\"]\n",
    "    \n",
    "fig = plt.figure(figsize=(8, 8))\n",
    "for idx in np.arange(20):\n",
    "    ax = fig.add_subplot(4, 10/2, idx+1, xticks=[], yticks=[])\n",
    "    ax.imshow(image[idx,:,:], cmap=plt.cm.binary)\n",
    "    ax.set_title(str(label[idx]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "HJKXuuKsdpy8"
   },
   "source": [
    "## 2. Defining the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "B4M77Gr2dpy9"
   },
   "outputs": [],
   "source": [
    "import mindspore.nn as nn\n",
    "import mindspore.ops.operations as P\n",
    "\n",
    "from mindspore.common.initializer import TruncatedNormal\n",
    "\n",
    "def weight_variable():\n",
    "    \"\"\"Weight initial.\"\"\"\n",
    "    return TruncatedNormal(0.02)\n",
    "\n",
    "\n",
    "class LeNet5(nn.Cell):\n",
    "    \n",
    "    def __init__(self):\n",
    "        super(LeNet5, self).__init__()\n",
    "        self.batch_size = 32\n",
    "        self.conv1 = nn.Conv2d(1, 6, kernel_size=5, \n",
    "                               weight_init=weight_variable(), has_bias=False, \n",
    "                               pad_mode=\"valid\"\n",
    "                              )\n",
    "        self.conv2 = nn.Conv2d(6, 16, kernel_size=5, \n",
    "                               weight_init=weight_variable(),has_bias=False, \n",
    "                               pad_mode=\"valid\"\n",
    "                              )\n",
    "        self.fc1 =  nn.Dense(in_channels= 16 * 5 * 5, out_channels= 120 ,weight_init= weight_variable(),bias_init=weight_variable())\n",
    "        self.fc2 =  nn.Dense(in_channels= 120,  out_channels= 84,weight_init=weight_variable(),bias_init=weight_variable())\n",
    "        self.fc3 =  nn.Dense(in_channels=  84, out_channels= 10,weight_init=weight_variable(),bias_init=weight_variable())\n",
    "        self.relu = nn.ReLU()\n",
    "        self.max_pool2d = nn.MaxPool2d(kernel_size=2, stride=2)\n",
    "        self.reshape = P.Reshape()\n",
    "   \n",
    "    def construct(self, x):\n",
    "        x = self.conv1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.max_pool2d(x)\n",
    "        x = self.conv2(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.max_pool2d(x)\n",
    "        x = self.reshape(x, (self.batch_size, -1))\n",
    "        x = self.fc1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.fc2(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.fc3(x)\n",
    "        return x\n",
    "\n",
    "    \n",
    "network = LeNet5()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "n9s2WEDHdpy_"
   },
   "source": [
    "## 3. Configuring the learning process"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MESNJEdsdpzA"
   },
   "outputs": [],
   "source": [
    "from mindspore.nn.loss import SoftmaxCrossEntropyWithLogits\n",
    "from mindspore.nn import Momentum\n",
    "\n",
    "criterion = SoftmaxCrossEntropyWithLogits(is_grad=False, sparse=True, reduction='mean')\n",
    "optimizer = Momentum(params=network.trainable_params(), learning_rate=0.01, momentum=0.9)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "C7tvN2BMdpzD"
   },
   "source": [
    "## 4. Training the Network\n",
    "### Saving the Configured Model\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "L-fzZuWpdpzD"
   },
   "outputs": [],
   "source": [
    "from mindspore.train.callback import ModelCheckpoint, CheckpointConfig, LossMonitor\n",
    "\n",
    "epochs = 1\n",
    "\n",
    "\n",
    "config_ck = CheckpointConfig(save_checkpoint_steps=1875, keep_checkpoint_max=10)\n",
    "\n",
    "ckpoint_cb = ModelCheckpoint(prefix=\"checkpoint_lenet\", config=config_ck)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "RyxR7s6WdpzG"
   },
   "source": [
    "### Calling the High-level Model API To Train and Save the Model File"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Xyfr6FAxdpzG"
   },
   "outputs": [],
   "source": [
    "from mindspore.nn.metrics import Accuracy\n",
    "\n",
    "from mindspore.train import Model\n",
    "\n",
    "model = Model(network, criterion, optimizer, metrics={\"Accuracy\": Accuracy()})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "eDFn7DdsdpzL",
    "outputId": "a2aca815-961d-4e34-a513-df42053731bc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 1 step: 1, loss is 2.3029962\n",
      "epoch: 1 step: 2, loss is 2.3023033\n",
      "epoch: 1 step: 3, loss is 2.3042462\n",
      "epoch: 1 step: 4, loss is 2.3055491\n",
      "epoch: 1 step: 5, loss is 2.300112\n",
      "epoch: 1 step: 6, loss is 2.30389\n",
      "epoch: 1 step: 7, loss is 2.3018756\n",
      "epoch: 1 step: 8, loss is 2.3052802\n",
      "epoch: 1 step: 9, loss is 2.3092847\n",
      "epoch: 1 step: 10, loss is 2.2985802\n",
      "epoch: 1 step: 11, loss is 2.3039274\n",
      "epoch: 1 step: 12, loss is 2.3022556\n",
      "epoch: 1 step: 13, loss is 2.3035452\n",
      "epoch: 1 step: 14, loss is 2.3043737\n",
      "epoch: 1 step: 15, loss is 2.3053923\n",
      "epoch: 1 step: 16, loss is 2.3018212\n",
      "epoch: 1 step: 17, loss is 2.304602\n",
      "epoch: 1 step: 18, loss is 2.3022635\n",
      "epoch: 1 step: 19, loss is 2.3065553\n",
      "epoch: 1 step: 20, loss is 2.3035917\n",
      "epoch: 1 step: 21, loss is 2.2996135\n",
      "epoch: 1 step: 22, loss is 2.3034377\n",
      "epoch: 1 step: 23, loss is 2.3002315\n",
      "epoch: 1 step: 24, loss is 2.301465\n",
      "epoch: 1 step: 25, loss is 2.2971945\n",
      "epoch: 1 step: 26, loss is 2.2951753\n",
      "epoch: 1 step: 27, loss is 2.2972674\n",
      "epoch: 1 step: 28, loss is 2.3033264\n",
      "epoch: 1 step: 29, loss is 2.2955678\n",
      "epoch: 1 step: 30, loss is 2.3087702\n",
      "epoch: 1 step: 31, loss is 2.300435\n",
      "epoch: 1 step: 32, loss is 2.2991755\n",
      "epoch: 1 step: 33, loss is 2.3100293\n",
      "epoch: 1 step: 34, loss is 2.3073454\n",
      "epoch: 1 step: 35, loss is 2.2935483\n",
      "epoch: 1 step: 36, loss is 2.3077915\n",
      "epoch: 1 step: 37, loss is 2.2999291\n",
      "epoch: 1 step: 38, loss is 2.2987452\n",
      "epoch: 1 step: 39, loss is 2.3109624\n",
      "epoch: 1 step: 40, loss is 2.3057714\n",
      "epoch: 1 step: 41, loss is 2.303825\n",
      "epoch: 1 step: 42, loss is 2.3007812\n",
      "epoch: 1 step: 43, loss is 2.301349\n",
      "epoch: 1 step: 44, loss is 2.3067784\n",
      "epoch: 1 step: 45, loss is 2.3052866\n",
      "epoch: 1 step: 46, loss is 2.2982605\n",
      "epoch: 1 step: 47, loss is 2.300379\n",
      "epoch: 1 step: 48, loss is 2.303829\n",
      "epoch: 1 step: 49, loss is 2.302877\n",
      "epoch: 1 step: 50, loss is 2.298593\n",
      "epoch: 1 step: 51, loss is 2.3019252\n",
      "epoch: 1 step: 52, loss is 2.3062139\n",
      "epoch: 1 step: 53, loss is 2.305516\n",
      "epoch: 1 step: 54, loss is 2.29586\n",
      "epoch: 1 step: 55, loss is 2.2994218\n",
      "epoch: 1 step: 56, loss is 2.3066695\n",
      "epoch: 1 step: 57, loss is 2.3060849\n",
      "epoch: 1 step: 58, loss is 2.2910738\n",
      "epoch: 1 step: 59, loss is 2.296553\n",
      "epoch: 1 step: 60, loss is 2.2962945\n",
      "epoch: 1 step: 61, loss is 2.3071089\n",
      "epoch: 1 step: 62, loss is 2.2984219\n",
      "epoch: 1 step: 63, loss is 2.3182068\n",
      "epoch: 1 step: 64, loss is 2.300341\n",
      "epoch: 1 step: 65, loss is 2.30137\n",
      "epoch: 1 step: 66, loss is 2.2996256\n",
      "epoch: 1 step: 67, loss is 2.2986288\n",
      "epoch: 1 step: 68, loss is 2.296656\n",
      "epoch: 1 step: 69, loss is 2.3011372\n",
      "epoch: 1 step: 70, loss is 2.3009942\n",
      "epoch: 1 step: 71, loss is 2.3065152\n",
      "epoch: 1 step: 72, loss is 2.3069751\n",
      "epoch: 1 step: 73, loss is 2.3032205\n",
      "epoch: 1 step: 74, loss is 2.2994812\n",
      "epoch: 1 step: 75, loss is 2.3052266\n",
      "epoch: 1 step: 76, loss is 2.3122694\n",
      "epoch: 1 step: 77, loss is 2.289251\n",
      "epoch: 1 step: 78, loss is 2.3088999\n",
      "epoch: 1 step: 79, loss is 2.3023975\n",
      "epoch: 1 step: 80, loss is 2.302027\n",
      "epoch: 1 step: 81, loss is 2.3014977\n",
      "epoch: 1 step: 82, loss is 2.297378\n",
      "epoch: 1 step: 83, loss is 2.3067386\n",
      "epoch: 1 step: 84, loss is 2.296199\n",
      "epoch: 1 step: 85, loss is 2.3154993\n",
      "epoch: 1 step: 86, loss is 2.3011074\n",
      "epoch: 1 step: 87, loss is 2.301231\n",
      "epoch: 1 step: 88, loss is 2.2964973\n",
      "epoch: 1 step: 89, loss is 2.3155732\n",
      "epoch: 1 step: 90, loss is 2.29347\n",
      "epoch: 1 step: 91, loss is 2.304638\n",
      "epoch: 1 step: 92, loss is 2.303334\n",
      "epoch: 1 step: 93, loss is 2.2971401\n",
      "epoch: 1 step: 94, loss is 2.3094468\n",
      "epoch: 1 step: 95, loss is 2.2878506\n",
      "epoch: 1 step: 96, loss is 2.3105247\n",
      "epoch: 1 step: 97, loss is 2.3097484\n",
      "epoch: 1 step: 98, loss is 2.2997887\n",
      "epoch: 1 step: 99, loss is 2.30565\n",
      "epoch: 1 step: 100, loss is 2.3055649\n",
      "epoch: 1 step: 101, loss is 2.2981207\n",
      "epoch: 1 step: 102, loss is 2.294114\n",
      "epoch: 1 step: 103, loss is 2.299725\n",
      "epoch: 1 step: 104, loss is 2.294454\n",
      "epoch: 1 step: 105, loss is 2.3098547\n",
      "epoch: 1 step: 106, loss is 2.3143775\n",
      "epoch: 1 step: 107, loss is 2.304016\n",
      "epoch: 1 step: 108, loss is 2.304703\n",
      "epoch: 1 step: 109, loss is 2.2976425\n",
      "epoch: 1 step: 110, loss is 2.3034306\n",
      "epoch: 1 step: 111, loss is 2.2953718\n",
      "epoch: 1 step: 112, loss is 2.292166\n",
      "epoch: 1 step: 113, loss is 2.2926478\n",
      "epoch: 1 step: 114, loss is 2.3128269\n",
      "epoch: 1 step: 115, loss is 2.3087351\n",
      "epoch: 1 step: 116, loss is 2.3060093\n",
      "epoch: 1 step: 117, loss is 2.2990327\n",
      "epoch: 1 step: 118, loss is 2.2932293\n",
      "epoch: 1 step: 119, loss is 2.3000698\n",
      "epoch: 1 step: 120, loss is 2.3007784\n",
      "epoch: 1 step: 121, loss is 2.3074322\n",
      "epoch: 1 step: 122, loss is 2.3049119\n",
      "epoch: 1 step: 123, loss is 2.3058274\n",
      "epoch: 1 step: 124, loss is 2.295801\n",
      "epoch: 1 step: 125, loss is 2.3048244\n",
      "epoch: 1 step: 126, loss is 2.298454\n",
      "epoch: 1 step: 127, loss is 2.314505\n",
      "epoch: 1 step: 128, loss is 2.2952\n",
      "epoch: 1 step: 129, loss is 2.3137836\n",
      "epoch: 1 step: 130, loss is 2.3023436\n",
      "epoch: 1 step: 131, loss is 2.3137512\n",
      "epoch: 1 step: 132, loss is 2.312273\n",
      "epoch: 1 step: 133, loss is 2.2883968\n",
      "epoch: 1 step: 134, loss is 2.298714\n",
      "epoch: 1 step: 135, loss is 2.2922437\n",
      "epoch: 1 step: 136, loss is 2.3075507\n",
      "epoch: 1 step: 137, loss is 2.3169024\n",
      "epoch: 1 step: 138, loss is 2.3014128\n",
      "epoch: 1 step: 139, loss is 2.3025618\n",
      "epoch: 1 step: 140, loss is 2.29023\n",
      "epoch: 1 step: 141, loss is 2.2972646\n",
      "epoch: 1 step: 142, loss is 2.2932143\n",
      "epoch: 1 step: 143, loss is 2.2979784\n",
      "epoch: 1 step: 144, loss is 2.2992787\n",
      "epoch: 1 step: 145, loss is 2.3015609\n",
      "epoch: 1 step: 146, loss is 2.2804303\n",
      "epoch: 1 step: 147, loss is 2.2864897\n",
      "epoch: 1 step: 148, loss is 2.301916\n",
      "epoch: 1 step: 149, loss is 2.308669\n",
      "epoch: 1 step: 150, loss is 2.3070023\n",
      "epoch: 1 step: 151, loss is 2.2941904\n",
      "epoch: 1 step: 152, loss is 2.3121314\n",
      "epoch: 1 step: 153, loss is 2.3035994\n",
      "epoch: 1 step: 154, loss is 2.2975783\n",
      "epoch: 1 step: 155, loss is 2.2998052\n",
      "epoch: 1 step: 156, loss is 2.3027737\n",
      "epoch: 1 step: 157, loss is 2.302718\n",
      "epoch: 1 step: 158, loss is 2.299946\n",
      "epoch: 1 step: 159, loss is 2.314553\n",
      "epoch: 1 step: 160, loss is 2.2926931\n",
      "epoch: 1 step: 161, loss is 2.2983484\n",
      "epoch: 1 step: 162, loss is 2.2899234\n",
      "epoch: 1 step: 163, loss is 2.3045704\n",
      "epoch: 1 step: 164, loss is 2.2876542\n",
      "epoch: 1 step: 165, loss is 2.3032312\n",
      "epoch: 1 step: 166, loss is 2.3125315\n",
      "epoch: 1 step: 167, loss is 2.2829337\n",
      "epoch: 1 step: 168, loss is 2.299346\n",
      "epoch: 1 step: 169, loss is 2.299763\n",
      "epoch: 1 step: 170, loss is 2.3104362\n",
      "epoch: 1 step: 171, loss is 2.300923\n",
      "epoch: 1 step: 172, loss is 2.289386\n",
      "epoch: 1 step: 173, loss is 2.284899\n",
      "epoch: 1 step: 174, loss is 2.2957635\n",
      "epoch: 1 step: 175, loss is 2.2976785\n",
      "epoch: 1 step: 176, loss is 2.3004596\n",
      "epoch: 1 step: 177, loss is 2.2909946\n",
      "epoch: 1 step: 178, loss is 2.316205\n",
      "epoch: 1 step: 179, loss is 2.293432\n",
      "epoch: 1 step: 180, loss is 2.3221316\n",
      "epoch: 1 step: 181, loss is 2.298076\n",
      "epoch: 1 step: 182, loss is 2.2887475\n",
      "epoch: 1 step: 183, loss is 2.3076138\n",
      "epoch: 1 step: 184, loss is 2.3071427\n",
      "epoch: 1 step: 185, loss is 2.2988234\n",
      "epoch: 1 step: 186, loss is 2.3176947\n",
      "epoch: 1 step: 187, loss is 2.3154936\n",
      "epoch: 1 step: 188, loss is 2.2978868\n",
      "epoch: 1 step: 189, loss is 2.3188145\n",
      "epoch: 1 step: 190, loss is 2.3265867\n",
      "epoch: 1 step: 191, loss is 2.2840192\n",
      "epoch: 1 step: 192, loss is 2.2779248\n",
      "epoch: 1 step: 193, loss is 2.2942634\n",
      "epoch: 1 step: 194, loss is 2.293752\n",
      "epoch: 1 step: 195, loss is 2.3143952\n",
      "epoch: 1 step: 196, loss is 2.3167205\n",
      "epoch: 1 step: 197, loss is 2.3180604\n",
      "epoch: 1 step: 198, loss is 2.2889986\n",
      "epoch: 1 step: 199, loss is 2.2721848\n",
      "epoch: 1 step: 200, loss is 2.2668009\n",
      "epoch: 1 step: 201, loss is 2.2934585\n",
      "epoch: 1 step: 202, loss is 2.2905836\n",
      "epoch: 1 step: 203, loss is 2.316973\n",
      "epoch: 1 step: 204, loss is 2.2936542\n",
      "epoch: 1 step: 205, loss is 2.2948396\n",
      "epoch: 1 step: 206, loss is 2.311099\n",
      "epoch: 1 step: 207, loss is 2.3039887\n",
      "epoch: 1 step: 208, loss is 2.304843\n",
      "epoch: 1 step: 209, loss is 2.294315\n",
      "epoch: 1 step: 210, loss is 2.315599\n",
      "epoch: 1 step: 211, loss is 2.319413\n",
      "epoch: 1 step: 212, loss is 2.3205626\n",
      "epoch: 1 step: 213, loss is 2.2874556\n",
      "epoch: 1 step: 214, loss is 2.304689\n",
      "epoch: 1 step: 215, loss is 2.3083367\n",
      "epoch: 1 step: 216, loss is 2.3075728\n",
      "epoch: 1 step: 217, loss is 2.297926\n",
      "epoch: 1 step: 218, loss is 2.3040934\n",
      "epoch: 1 step: 219, loss is 2.2911623\n",
      "epoch: 1 step: 220, loss is 2.3222873\n",
      "epoch: 1 step: 221, loss is 2.3072026\n",
      "epoch: 1 step: 222, loss is 2.313298\n",
      "epoch: 1 step: 223, loss is 2.2956254\n",
      "epoch: 1 step: 224, loss is 2.3223882\n",
      "epoch: 1 step: 225, loss is 2.296574\n",
      "epoch: 1 step: 226, loss is 2.3079398\n",
      "epoch: 1 step: 227, loss is 2.3019006\n",
      "epoch: 1 step: 228, loss is 2.3020067\n",
      "epoch: 1 step: 229, loss is 2.3077672\n",
      "epoch: 1 step: 230, loss is 2.312134\n",
      "epoch: 1 step: 231, loss is 2.2865784\n",
      "epoch: 1 step: 232, loss is 2.3093846\n",
      "epoch: 1 step: 233, loss is 2.3218632\n",
      "epoch: 1 step: 234, loss is 2.3055515\n",
      "epoch: 1 step: 235, loss is 2.2963436\n",
      "epoch: 1 step: 236, loss is 2.2988284\n",
      "epoch: 1 step: 237, loss is 2.3028433\n",
      "epoch: 1 step: 238, loss is 2.292194\n",
      "epoch: 1 step: 239, loss is 2.3276615\n",
      "epoch: 1 step: 240, loss is 2.3103213\n",
      "epoch: 1 step: 241, loss is 2.3087182\n",
      "epoch: 1 step: 242, loss is 2.3181994\n",
      "epoch: 1 step: 243, loss is 2.3034706\n",
      "epoch: 1 step: 244, loss is 2.2965226\n",
      "epoch: 1 step: 245, loss is 2.313918\n",
      "epoch: 1 step: 246, loss is 2.3156176\n",
      "epoch: 1 step: 247, loss is 2.310821\n",
      "epoch: 1 step: 248, loss is 2.2855873\n",
      "epoch: 1 step: 249, loss is 2.29935\n",
      "epoch: 1 step: 250, loss is 2.304107\n",
      "epoch: 1 step: 251, loss is 2.3136735\n",
      "epoch: 1 step: 252, loss is 2.2967284\n",
      "epoch: 1 step: 253, loss is 2.3159103\n",
      "epoch: 1 step: 254, loss is 2.3025062\n",
      "epoch: 1 step: 255, loss is 2.3121564\n",
      "epoch: 1 step: 256, loss is 2.2974794\n",
      "epoch: 1 step: 257, loss is 2.3053102\n",
      "epoch: 1 step: 258, loss is 2.3190336\n",
      "epoch: 1 step: 259, loss is 2.3142445\n",
      "epoch: 1 step: 260, loss is 2.3037894\n",
      "epoch: 1 step: 261, loss is 2.2994525\n",
      "epoch: 1 step: 262, loss is 2.3020527\n",
      "epoch: 1 step: 263, loss is 2.3111484\n",
      "epoch: 1 step: 264, loss is 2.2998466\n",
      "epoch: 1 step: 265, loss is 2.3100696\n",
      "epoch: 1 step: 266, loss is 2.3053765\n",
      "epoch: 1 step: 267, loss is 2.3011577\n",
      "epoch: 1 step: 268, loss is 2.3156648\n",
      "epoch: 1 step: 269, loss is 2.3020537\n",
      "epoch: 1 step: 270, loss is 2.2961323\n",
      "epoch: 1 step: 271, loss is 2.2995791\n",
      "epoch: 1 step: 272, loss is 2.3054132\n",
      "epoch: 1 step: 273, loss is 2.3004441\n",
      "epoch: 1 step: 274, loss is 2.3012667\n",
      "epoch: 1 step: 275, loss is 2.295972\n",
      "epoch: 1 step: 276, loss is 2.305127\n",
      "epoch: 1 step: 277, loss is 2.3037634\n",
      "epoch: 1 step: 278, loss is 2.2985766\n",
      "epoch: 1 step: 279, loss is 2.3033671\n",
      "epoch: 1 step: 280, loss is 2.3031647\n",
      "epoch: 1 step: 281, loss is 2.3014297\n",
      "epoch: 1 step: 282, loss is 2.3030472\n",
      "epoch: 1 step: 283, loss is 2.2892826\n",
      "epoch: 1 step: 284, loss is 2.3113015\n",
      "epoch: 1 step: 285, loss is 2.3032727\n",
      "epoch: 1 step: 286, loss is 2.293394\n",
      "epoch: 1 step: 287, loss is 2.2984722\n",
      "epoch: 1 step: 288, loss is 2.295557\n",
      "epoch: 1 step: 289, loss is 2.293338\n",
      "epoch: 1 step: 290, loss is 2.3060725\n",
      "epoch: 1 step: 291, loss is 2.3040712\n",
      "epoch: 1 step: 292, loss is 2.2979252\n",
      "epoch: 1 step: 293, loss is 2.300102\n",
      "epoch: 1 step: 294, loss is 2.2985814\n",
      "epoch: 1 step: 295, loss is 2.3065863\n",
      "epoch: 1 step: 296, loss is 2.306228\n",
      "epoch: 1 step: 297, loss is 2.2980921\n",
      "epoch: 1 step: 298, loss is 2.2984505\n",
      "epoch: 1 step: 299, loss is 2.3009896\n",
      "epoch: 1 step: 300, loss is 2.2985115\n",
      "epoch: 1 step: 301, loss is 2.293402\n",
      "epoch: 1 step: 302, loss is 2.3019166\n",
      "epoch: 1 step: 303, loss is 2.3060656\n",
      "epoch: 1 step: 304, loss is 2.307879\n",
      "epoch: 1 step: 305, loss is 2.2920325\n",
      "epoch: 1 step: 306, loss is 2.3173528\n",
      "epoch: 1 step: 307, loss is 2.2851608\n",
      "epoch: 1 step: 308, loss is 2.2922473\n",
      "epoch: 1 step: 309, loss is 2.3066397\n",
      "epoch: 1 step: 310, loss is 2.3007293\n",
      "epoch: 1 step: 311, loss is 2.311826\n",
      "epoch: 1 step: 312, loss is 2.3087308\n",
      "epoch: 1 step: 313, loss is 2.2803051\n",
      "epoch: 1 step: 314, loss is 2.286456\n",
      "epoch: 1 step: 315, loss is 2.303073\n",
      "epoch: 1 step: 316, loss is 2.316142\n",
      "epoch: 1 step: 317, loss is 2.3150346\n",
      "epoch: 1 step: 318, loss is 2.3046112\n",
      "epoch: 1 step: 319, loss is 2.3055978\n",
      "epoch: 1 step: 320, loss is 2.2989411\n",
      "epoch: 1 step: 321, loss is 2.310385\n",
      "epoch: 1 step: 322, loss is 2.3257015\n",
      "epoch: 1 step: 323, loss is 2.3001907\n",
      "epoch: 1 step: 324, loss is 2.3065062\n",
      "epoch: 1 step: 325, loss is 2.3031416\n",
      "epoch: 1 step: 326, loss is 2.3080535\n",
      "epoch: 1 step: 327, loss is 2.29632\n",
      "epoch: 1 step: 328, loss is 2.2978456\n",
      "epoch: 1 step: 329, loss is 2.3114698\n",
      "epoch: 1 step: 330, loss is 2.294661\n",
      "epoch: 1 step: 331, loss is 2.3227844\n",
      "epoch: 1 step: 332, loss is 2.3011503\n",
      "epoch: 1 step: 333, loss is 2.3071456\n",
      "epoch: 1 step: 334, loss is 2.3033295\n",
      "epoch: 1 step: 335, loss is 2.3080847\n",
      "epoch: 1 step: 336, loss is 2.2955217\n",
      "epoch: 1 step: 337, loss is 2.2973197\n",
      "epoch: 1 step: 338, loss is 2.2828119\n",
      "epoch: 1 step: 339, loss is 2.301963\n",
      "epoch: 1 step: 340, loss is 2.2933507\n",
      "epoch: 1 step: 341, loss is 2.3032506\n",
      "epoch: 1 step: 342, loss is 2.303399\n",
      "epoch: 1 step: 343, loss is 2.2998736\n",
      "epoch: 1 step: 344, loss is 2.2973485\n",
      "epoch: 1 step: 345, loss is 2.30359\n",
      "epoch: 1 step: 346, loss is 2.315813\n",
      "epoch: 1 step: 347, loss is 2.3046315\n",
      "epoch: 1 step: 348, loss is 2.302012\n",
      "epoch: 1 step: 349, loss is 2.2908013\n",
      "epoch: 1 step: 350, loss is 2.2876906\n",
      "epoch: 1 step: 351, loss is 2.2998405\n",
      "epoch: 1 step: 352, loss is 2.3007448\n",
      "epoch: 1 step: 353, loss is 2.3041644\n",
      "epoch: 1 step: 354, loss is 2.3336453\n",
      "epoch: 1 step: 355, loss is 2.2961583\n",
      "epoch: 1 step: 356, loss is 2.2848806\n",
      "epoch: 1 step: 357, loss is 2.2959967\n",
      "epoch: 1 step: 358, loss is 2.3033547\n",
      "epoch: 1 step: 359, loss is 2.3027961\n",
      "epoch: 1 step: 360, loss is 2.3013756\n",
      "epoch: 1 step: 361, loss is 2.303025\n",
      "epoch: 1 step: 362, loss is 2.3178566\n",
      "epoch: 1 step: 363, loss is 2.305174\n",
      "epoch: 1 step: 364, loss is 2.3036296\n",
      "epoch: 1 step: 365, loss is 2.3086734\n",
      "epoch: 1 step: 366, loss is 2.3011744\n",
      "epoch: 1 step: 367, loss is 2.302851\n",
      "epoch: 1 step: 368, loss is 2.2845628\n",
      "epoch: 1 step: 369, loss is 2.307473\n",
      "epoch: 1 step: 370, loss is 2.299728\n",
      "epoch: 1 step: 371, loss is 2.3042734\n",
      "epoch: 1 step: 372, loss is 2.29048\n",
      "epoch: 1 step: 373, loss is 2.284125\n",
      "epoch: 1 step: 374, loss is 2.3247106\n",
      "epoch: 1 step: 375, loss is 2.3026748\n",
      "epoch: 1 step: 376, loss is 2.3087828\n",
      "epoch: 1 step: 377, loss is 2.2892354\n",
      "epoch: 1 step: 378, loss is 2.3002157\n",
      "epoch: 1 step: 379, loss is 2.2899628\n",
      "epoch: 1 step: 380, loss is 2.308765\n",
      "epoch: 1 step: 381, loss is 2.309237\n",
      "epoch: 1 step: 382, loss is 2.3017335\n",
      "epoch: 1 step: 383, loss is 2.2984562\n",
      "epoch: 1 step: 384, loss is 2.3044064\n",
      "epoch: 1 step: 385, loss is 2.3125868\n",
      "epoch: 1 step: 386, loss is 2.300575\n",
      "epoch: 1 step: 387, loss is 2.2998352\n",
      "epoch: 1 step: 388, loss is 2.302331\n",
      "epoch: 1 step: 389, loss is 2.2935581\n",
      "epoch: 1 step: 390, loss is 2.2964294\n",
      "epoch: 1 step: 391, loss is 2.2811058\n",
      "epoch: 1 step: 392, loss is 2.3075457\n",
      "epoch: 1 step: 393, loss is 2.2941382\n",
      "epoch: 1 step: 394, loss is 2.3065095\n",
      "epoch: 1 step: 395, loss is 2.3111439\n",
      "epoch: 1 step: 396, loss is 2.3035471\n",
      "epoch: 1 step: 397, loss is 2.3107104\n",
      "epoch: 1 step: 398, loss is 2.3065703\n",
      "epoch: 1 step: 399, loss is 2.305151\n",
      "epoch: 1 step: 400, loss is 2.322019\n",
      "epoch: 1 step: 401, loss is 2.2844164\n",
      "epoch: 1 step: 402, loss is 2.3187273\n",
      "epoch: 1 step: 403, loss is 2.290029\n",
      "epoch: 1 step: 404, loss is 2.291284\n",
      "epoch: 1 step: 405, loss is 2.303893\n",
      "epoch: 1 step: 406, loss is 2.3036213\n",
      "epoch: 1 step: 407, loss is 2.288257\n",
      "epoch: 1 step: 408, loss is 2.296296\n",
      "epoch: 1 step: 409, loss is 2.2906597\n",
      "epoch: 1 step: 410, loss is 2.3063765\n",
      "epoch: 1 step: 411, loss is 2.2915761\n",
      "epoch: 1 step: 412, loss is 2.293186\n",
      "epoch: 1 step: 413, loss is 2.3023005\n",
      "epoch: 1 step: 414, loss is 2.294557\n",
      "epoch: 1 step: 415, loss is 2.2823164\n",
      "epoch: 1 step: 416, loss is 2.2923117\n",
      "epoch: 1 step: 417, loss is 2.2991107\n",
      "epoch: 1 step: 418, loss is 2.3020618\n",
      "epoch: 1 step: 419, loss is 2.3011408\n",
      "epoch: 1 step: 420, loss is 2.3108017\n",
      "epoch: 1 step: 421, loss is 2.3099983\n",
      "epoch: 1 step: 422, loss is 2.3018787\n",
      "epoch: 1 step: 423, loss is 2.293848\n",
      "epoch: 1 step: 424, loss is 2.305976\n",
      "epoch: 1 step: 425, loss is 2.306494\n",
      "epoch: 1 step: 426, loss is 2.2829568\n",
      "epoch: 1 step: 427, loss is 2.305397\n",
      "epoch: 1 step: 428, loss is 2.3174531\n",
      "epoch: 1 step: 429, loss is 2.2901013\n",
      "epoch: 1 step: 430, loss is 2.3009746\n",
      "epoch: 1 step: 431, loss is 2.2954738\n",
      "epoch: 1 step: 432, loss is 2.3100688\n",
      "epoch: 1 step: 433, loss is 2.3199031\n",
      "epoch: 1 step: 434, loss is 2.2986515\n",
      "epoch: 1 step: 435, loss is 2.2960784\n",
      "epoch: 1 step: 436, loss is 2.3039968\n",
      "epoch: 1 step: 437, loss is 2.3270285\n",
      "epoch: 1 step: 438, loss is 2.3145463\n",
      "epoch: 1 step: 439, loss is 2.296881\n",
      "epoch: 1 step: 440, loss is 2.2900448\n",
      "epoch: 1 step: 441, loss is 2.3113263\n",
      "epoch: 1 step: 442, loss is 2.3014781\n",
      "epoch: 1 step: 443, loss is 2.2870824\n",
      "epoch: 1 step: 444, loss is 2.312058\n",
      "epoch: 1 step: 445, loss is 2.31273\n",
      "epoch: 1 step: 446, loss is 2.3066604\n",
      "epoch: 1 step: 447, loss is 2.2907703\n",
      "epoch: 1 step: 448, loss is 2.3023183\n",
      "epoch: 1 step: 449, loss is 2.3100533\n",
      "epoch: 1 step: 450, loss is 2.2921178\n",
      "epoch: 1 step: 451, loss is 2.2871926\n",
      "epoch: 1 step: 452, loss is 2.2958455\n",
      "epoch: 1 step: 453, loss is 2.3064432\n",
      "epoch: 1 step: 454, loss is 2.3048096\n",
      "epoch: 1 step: 455, loss is 2.3054698\n",
      "epoch: 1 step: 456, loss is 2.3007739\n",
      "epoch: 1 step: 457, loss is 2.315443\n",
      "epoch: 1 step: 458, loss is 2.3080106\n",
      "epoch: 1 step: 459, loss is 2.3110929\n",
      "epoch: 1 step: 460, loss is 2.302502\n",
      "epoch: 1 step: 461, loss is 2.3120713\n",
      "epoch: 1 step: 462, loss is 2.3122973\n",
      "epoch: 1 step: 463, loss is 2.3017654\n",
      "epoch: 1 step: 464, loss is 2.2971268\n",
      "epoch: 1 step: 465, loss is 2.3025398\n",
      "epoch: 1 step: 466, loss is 2.3220384\n",
      "epoch: 1 step: 467, loss is 2.3108387\n",
      "epoch: 1 step: 468, loss is 2.2918506\n",
      "epoch: 1 step: 469, loss is 2.3040514\n",
      "epoch: 1 step: 470, loss is 2.3093767\n",
      "epoch: 1 step: 471, loss is 2.289769\n",
      "epoch: 1 step: 472, loss is 2.3165872\n",
      "epoch: 1 step: 473, loss is 2.3038907\n",
      "epoch: 1 step: 474, loss is 2.2924385\n",
      "epoch: 1 step: 475, loss is 2.298873\n",
      "epoch: 1 step: 476, loss is 2.2965446\n",
      "epoch: 1 step: 477, loss is 2.3139517\n",
      "epoch: 1 step: 478, loss is 2.3150735\n",
      "epoch: 1 step: 479, loss is 2.2846558\n",
      "epoch: 1 step: 480, loss is 2.2976274\n",
      "epoch: 1 step: 481, loss is 2.301548\n",
      "epoch: 1 step: 482, loss is 2.310239\n",
      "epoch: 1 step: 483, loss is 2.3120482\n",
      "epoch: 1 step: 484, loss is 2.291098\n",
      "epoch: 1 step: 485, loss is 2.304554\n",
      "epoch: 1 step: 486, loss is 2.3146229\n",
      "epoch: 1 step: 487, loss is 2.3250508\n",
      "epoch: 1 step: 488, loss is 2.2986696\n",
      "epoch: 1 step: 489, loss is 2.2971182\n",
      "epoch: 1 step: 490, loss is 2.2978551\n",
      "epoch: 1 step: 491, loss is 2.3185365\n",
      "epoch: 1 step: 492, loss is 2.303065\n",
      "epoch: 1 step: 493, loss is 2.3031669\n",
      "epoch: 1 step: 494, loss is 2.3059745\n",
      "epoch: 1 step: 495, loss is 2.2904499\n",
      "epoch: 1 step: 496, loss is 2.303825\n",
      "epoch: 1 step: 497, loss is 2.2981582\n",
      "epoch: 1 step: 498, loss is 2.296399\n",
      "epoch: 1 step: 499, loss is 2.300419\n",
      "epoch: 1 step: 500, loss is 2.291586\n",
      "epoch: 1 step: 501, loss is 2.2993937\n",
      "epoch: 1 step: 502, loss is 2.3039415\n",
      "epoch: 1 step: 503, loss is 2.2921247\n",
      "epoch: 1 step: 504, loss is 2.3131497\n",
      "epoch: 1 step: 505, loss is 2.309704\n",
      "epoch: 1 step: 506, loss is 2.2970307\n",
      "epoch: 1 step: 507, loss is 2.3101943\n",
      "epoch: 1 step: 508, loss is 2.3142653\n",
      "epoch: 1 step: 509, loss is 2.2968862\n",
      "epoch: 1 step: 510, loss is 2.2937593\n",
      "epoch: 1 step: 511, loss is 2.290421\n",
      "epoch: 1 step: 512, loss is 2.3042552\n",
      "epoch: 1 step: 513, loss is 2.3048475\n",
      "epoch: 1 step: 514, loss is 2.3251848\n",
      "epoch: 1 step: 515, loss is 2.2983809\n",
      "epoch: 1 step: 516, loss is 2.3202536\n",
      "epoch: 1 step: 517, loss is 2.2996657\n",
      "epoch: 1 step: 518, loss is 2.2972994\n",
      "epoch: 1 step: 519, loss is 2.292424\n",
      "epoch: 1 step: 520, loss is 2.300064\n",
      "epoch: 1 step: 521, loss is 2.3106108\n",
      "epoch: 1 step: 522, loss is 2.3150246\n",
      "epoch: 1 step: 523, loss is 2.2892306\n",
      "epoch: 1 step: 524, loss is 2.288677\n",
      "epoch: 1 step: 525, loss is 2.3062646\n",
      "epoch: 1 step: 526, loss is 2.3213189\n",
      "epoch: 1 step: 527, loss is 2.2941377\n",
      "epoch: 1 step: 528, loss is 2.2902625\n",
      "epoch: 1 step: 529, loss is 2.3222272\n",
      "epoch: 1 step: 530, loss is 2.3104053\n",
      "epoch: 1 step: 531, loss is 2.3103926\n",
      "epoch: 1 step: 532, loss is 2.2912743\n",
      "epoch: 1 step: 533, loss is 2.2943175\n",
      "epoch: 1 step: 534, loss is 2.3066075\n",
      "epoch: 1 step: 535, loss is 2.318593\n",
      "epoch: 1 step: 536, loss is 2.3111165\n",
      "epoch: 1 step: 537, loss is 2.2986705\n",
      "epoch: 1 step: 538, loss is 2.2988207\n",
      "epoch: 1 step: 539, loss is 2.3088887\n",
      "epoch: 1 step: 540, loss is 2.3088908\n",
      "epoch: 1 step: 541, loss is 2.3003678\n",
      "epoch: 1 step: 542, loss is 2.3038821\n",
      "epoch: 1 step: 543, loss is 2.3007236\n",
      "epoch: 1 step: 544, loss is 2.3029187\n",
      "epoch: 1 step: 545, loss is 2.2960396\n",
      "epoch: 1 step: 546, loss is 2.3017156\n",
      "epoch: 1 step: 547, loss is 2.3079653\n",
      "epoch: 1 step: 548, loss is 2.3065453\n",
      "epoch: 1 step: 549, loss is 2.3145773\n",
      "epoch: 1 step: 550, loss is 2.2970443\n",
      "epoch: 1 step: 551, loss is 2.296608\n",
      "epoch: 1 step: 552, loss is 2.3030853\n",
      "epoch: 1 step: 553, loss is 2.298063\n",
      "epoch: 1 step: 554, loss is 2.2940123\n",
      "epoch: 1 step: 555, loss is 2.2945318\n",
      "epoch: 1 step: 556, loss is 2.299582\n",
      "epoch: 1 step: 557, loss is 2.299751\n",
      "epoch: 1 step: 558, loss is 2.303287\n",
      "epoch: 1 step: 559, loss is 2.307415\n",
      "epoch: 1 step: 560, loss is 2.3070636\n",
      "epoch: 1 step: 561, loss is 2.2869992\n",
      "epoch: 1 step: 562, loss is 2.3014994\n",
      "epoch: 1 step: 563, loss is 2.3049612\n",
      "epoch: 1 step: 564, loss is 2.2908304\n",
      "epoch: 1 step: 565, loss is 2.3071134\n",
      "epoch: 1 step: 566, loss is 2.2965384\n",
      "epoch: 1 step: 567, loss is 2.3024287\n",
      "epoch: 1 step: 568, loss is 2.3046951\n",
      "epoch: 1 step: 569, loss is 2.2968037\n",
      "epoch: 1 step: 570, loss is 2.3132975\n",
      "epoch: 1 step: 571, loss is 2.311821\n",
      "epoch: 1 step: 572, loss is 2.2897503\n",
      "epoch: 1 step: 573, loss is 2.2970297\n",
      "epoch: 1 step: 574, loss is 2.306521\n",
      "epoch: 1 step: 575, loss is 2.3038888\n",
      "epoch: 1 step: 576, loss is 2.2962608\n",
      "epoch: 1 step: 577, loss is 2.3017333\n",
      "epoch: 1 step: 578, loss is 2.2940834\n",
      "epoch: 1 step: 579, loss is 2.3025856\n",
      "epoch: 1 step: 580, loss is 2.3082569\n",
      "epoch: 1 step: 581, loss is 2.2967856\n",
      "epoch: 1 step: 582, loss is 2.3011725\n",
      "epoch: 1 step: 583, loss is 2.299956\n",
      "epoch: 1 step: 584, loss is 2.2891507\n",
      "epoch: 1 step: 585, loss is 2.3144338\n",
      "epoch: 1 step: 586, loss is 2.295268\n",
      "epoch: 1 step: 587, loss is 2.2931628\n",
      "epoch: 1 step: 588, loss is 2.2796664\n",
      "epoch: 1 step: 589, loss is 2.3036134\n",
      "epoch: 1 step: 590, loss is 2.303571\n",
      "epoch: 1 step: 591, loss is 2.2991529\n",
      "epoch: 1 step: 592, loss is 2.298493\n",
      "epoch: 1 step: 593, loss is 2.3055289\n",
      "epoch: 1 step: 594, loss is 2.3093574\n",
      "epoch: 1 step: 595, loss is 2.3060079\n",
      "epoch: 1 step: 596, loss is 2.3018682\n",
      "epoch: 1 step: 597, loss is 2.27737\n",
      "epoch: 1 step: 598, loss is 2.289336\n",
      "epoch: 1 step: 599, loss is 2.3037992\n",
      "epoch: 1 step: 600, loss is 2.3057344\n",
      "epoch: 1 step: 601, loss is 2.292116\n",
      "epoch: 1 step: 602, loss is 2.2992759\n",
      "epoch: 1 step: 603, loss is 2.2959216\n",
      "epoch: 1 step: 604, loss is 2.2936053\n",
      "epoch: 1 step: 605, loss is 2.2860718\n",
      "epoch: 1 step: 606, loss is 2.288163\n",
      "epoch: 1 step: 607, loss is 2.3116295\n",
      "epoch: 1 step: 608, loss is 2.3154473\n",
      "epoch: 1 step: 609, loss is 2.2970304\n",
      "epoch: 1 step: 610, loss is 2.319508\n",
      "epoch: 1 step: 611, loss is 2.3093991\n",
      "epoch: 1 step: 612, loss is 2.2802548\n",
      "epoch: 1 step: 613, loss is 2.2980065\n",
      "epoch: 1 step: 614, loss is 2.317182\n",
      "epoch: 1 step: 615, loss is 2.3106182\n",
      "epoch: 1 step: 616, loss is 2.2980468\n",
      "epoch: 1 step: 617, loss is 2.3078258\n",
      "epoch: 1 step: 618, loss is 2.2892785\n",
      "epoch: 1 step: 619, loss is 2.3124888\n",
      "epoch: 1 step: 620, loss is 2.307985\n",
      "epoch: 1 step: 621, loss is 2.3153574\n",
      "epoch: 1 step: 622, loss is 2.3155668\n",
      "epoch: 1 step: 623, loss is 2.316438\n",
      "epoch: 1 step: 624, loss is 2.2947693\n",
      "epoch: 1 step: 625, loss is 2.2873824\n",
      "epoch: 1 step: 626, loss is 2.2995577\n",
      "epoch: 1 step: 627, loss is 2.3054855\n",
      "epoch: 1 step: 628, loss is 2.308799\n",
      "epoch: 1 step: 629, loss is 2.3081622\n",
      "epoch: 1 step: 630, loss is 2.3038895\n",
      "epoch: 1 step: 631, loss is 2.299131\n",
      "epoch: 1 step: 632, loss is 2.302995\n",
      "epoch: 1 step: 633, loss is 2.2897153\n",
      "epoch: 1 step: 634, loss is 2.3028705\n",
      "epoch: 1 step: 635, loss is 2.3014042\n",
      "epoch: 1 step: 636, loss is 2.2881668\n",
      "epoch: 1 step: 637, loss is 2.3226752\n",
      "epoch: 1 step: 638, loss is 2.3069408\n",
      "epoch: 1 step: 639, loss is 2.2857933\n",
      "epoch: 1 step: 640, loss is 2.3093433\n",
      "epoch: 1 step: 641, loss is 2.3251307\n",
      "epoch: 1 step: 642, loss is 2.2811055\n",
      "epoch: 1 step: 643, loss is 2.291658\n",
      "epoch: 1 step: 644, loss is 2.3028111\n",
      "epoch: 1 step: 645, loss is 2.3064015\n",
      "epoch: 1 step: 646, loss is 2.3136027\n",
      "epoch: 1 step: 647, loss is 2.2963989\n",
      "epoch: 1 step: 648, loss is 2.2895792\n",
      "epoch: 1 step: 649, loss is 2.2930448\n",
      "epoch: 1 step: 650, loss is 2.2900379\n",
      "epoch: 1 step: 651, loss is 2.309043\n",
      "epoch: 1 step: 652, loss is 2.296683\n",
      "epoch: 1 step: 653, loss is 2.2916539\n",
      "epoch: 1 step: 654, loss is 2.2991543\n",
      "epoch: 1 step: 655, loss is 2.295014\n",
      "epoch: 1 step: 656, loss is 2.3055043\n",
      "epoch: 1 step: 657, loss is 2.304653\n",
      "epoch: 1 step: 658, loss is 2.2913005\n",
      "epoch: 1 step: 659, loss is 2.3050559\n",
      "epoch: 1 step: 660, loss is 2.3191848\n",
      "epoch: 1 step: 661, loss is 2.312621\n",
      "epoch: 1 step: 662, loss is 2.3211725\n",
      "epoch: 1 step: 663, loss is 2.3200266\n",
      "epoch: 1 step: 664, loss is 2.3185844\n",
      "epoch: 1 step: 665, loss is 2.3034518\n",
      "epoch: 1 step: 666, loss is 2.3009553\n",
      "epoch: 1 step: 667, loss is 2.313295\n",
      "epoch: 1 step: 668, loss is 2.293387\n",
      "epoch: 1 step: 669, loss is 2.3134294\n",
      "epoch: 1 step: 670, loss is 2.299529\n",
      "epoch: 1 step: 671, loss is 2.3073337\n",
      "epoch: 1 step: 672, loss is 2.3088694\n",
      "epoch: 1 step: 673, loss is 2.28715\n",
      "epoch: 1 step: 674, loss is 2.305023\n",
      "epoch: 1 step: 675, loss is 2.3089\n",
      "epoch: 1 step: 676, loss is 2.3045926\n",
      "epoch: 1 step: 677, loss is 2.2977874\n",
      "epoch: 1 step: 678, loss is 2.2979376\n",
      "epoch: 1 step: 679, loss is 2.314486\n",
      "epoch: 1 step: 680, loss is 2.2917325\n",
      "epoch: 1 step: 681, loss is 2.3021462\n",
      "epoch: 1 step: 682, loss is 2.2639887\n",
      "epoch: 1 step: 683, loss is 2.2909288\n",
      "epoch: 1 step: 684, loss is 2.2944221\n",
      "epoch: 1 step: 685, loss is 2.300287\n",
      "epoch: 1 step: 686, loss is 2.308497\n",
      "epoch: 1 step: 687, loss is 2.2935278\n",
      "epoch: 1 step: 688, loss is 2.2742045\n",
      "epoch: 1 step: 689, loss is 2.3093963\n",
      "epoch: 1 step: 690, loss is 2.3229187\n",
      "epoch: 1 step: 691, loss is 2.2971833\n",
      "epoch: 1 step: 692, loss is 2.2969968\n",
      "epoch: 1 step: 693, loss is 2.3050008\n",
      "epoch: 1 step: 694, loss is 2.3034408\n",
      "epoch: 1 step: 695, loss is 2.3094168\n",
      "epoch: 1 step: 696, loss is 2.3099923\n",
      "epoch: 1 step: 697, loss is 2.2916842\n",
      "epoch: 1 step: 698, loss is 2.2925377\n",
      "epoch: 1 step: 699, loss is 2.3033874\n",
      "epoch: 1 step: 700, loss is 2.2927265\n",
      "epoch: 1 step: 701, loss is 2.2975602\n",
      "epoch: 1 step: 702, loss is 2.2938962\n",
      "epoch: 1 step: 703, loss is 2.2962148\n",
      "epoch: 1 step: 704, loss is 2.3154366\n",
      "epoch: 1 step: 705, loss is 2.277434\n",
      "epoch: 1 step: 706, loss is 2.3041136\n",
      "epoch: 1 step: 707, loss is 2.3260593\n",
      "epoch: 1 step: 708, loss is 2.3031688\n",
      "epoch: 1 step: 709, loss is 2.2972908\n",
      "epoch: 1 step: 710, loss is 2.3048732\n",
      "epoch: 1 step: 711, loss is 2.3212683\n",
      "epoch: 1 step: 712, loss is 2.315587\n",
      "epoch: 1 step: 713, loss is 2.3177829\n",
      "epoch: 1 step: 714, loss is 2.3051364\n",
      "epoch: 1 step: 715, loss is 2.2999024\n",
      "epoch: 1 step: 716, loss is 2.2903762\n",
      "epoch: 1 step: 717, loss is 2.3024626\n",
      "epoch: 1 step: 718, loss is 2.3191202\n",
      "epoch: 1 step: 719, loss is 2.3058953\n",
      "epoch: 1 step: 720, loss is 2.3031185\n",
      "epoch: 1 step: 721, loss is 2.29788\n",
      "epoch: 1 step: 722, loss is 2.2941303\n",
      "epoch: 1 step: 723, loss is 2.298979\n",
      "epoch: 1 step: 724, loss is 2.2792814\n",
      "epoch: 1 step: 725, loss is 2.2959552\n",
      "epoch: 1 step: 726, loss is 2.2890294\n",
      "epoch: 1 step: 727, loss is 2.315591\n",
      "epoch: 1 step: 728, loss is 2.3063195\n",
      "epoch: 1 step: 729, loss is 2.302895\n",
      "epoch: 1 step: 730, loss is 2.3118057\n",
      "epoch: 1 step: 731, loss is 2.2895277\n",
      "epoch: 1 step: 732, loss is 2.308765\n",
      "epoch: 1 step: 733, loss is 2.300786\n",
      "epoch: 1 step: 734, loss is 2.3070493\n",
      "epoch: 1 step: 735, loss is 2.3090646\n",
      "epoch: 1 step: 736, loss is 2.2978358\n",
      "epoch: 1 step: 737, loss is 2.291006\n",
      "epoch: 1 step: 738, loss is 2.293623\n",
      "epoch: 1 step: 739, loss is 2.3090348\n",
      "epoch: 1 step: 740, loss is 2.295259\n",
      "epoch: 1 step: 741, loss is 2.2999032\n",
      "epoch: 1 step: 742, loss is 2.3122606\n",
      "epoch: 1 step: 743, loss is 2.309352\n",
      "epoch: 1 step: 744, loss is 2.3024263\n",
      "epoch: 1 step: 745, loss is 2.2824428\n",
      "epoch: 1 step: 746, loss is 2.3054252\n",
      "epoch: 1 step: 747, loss is 2.3124676\n",
      "epoch: 1 step: 748, loss is 2.3124907\n",
      "epoch: 1 step: 749, loss is 2.298224\n",
      "epoch: 1 step: 750, loss is 2.2954435\n",
      "epoch: 1 step: 751, loss is 2.293177\n",
      "epoch: 1 step: 752, loss is 2.290289\n",
      "epoch: 1 step: 753, loss is 2.3164544\n",
      "epoch: 1 step: 754, loss is 2.3050995\n",
      "epoch: 1 step: 755, loss is 2.3029983\n",
      "epoch: 1 step: 756, loss is 2.2915401\n",
      "epoch: 1 step: 757, loss is 2.3114169\n",
      "epoch: 1 step: 758, loss is 2.3024666\n",
      "epoch: 1 step: 759, loss is 2.284373\n",
      "epoch: 1 step: 760, loss is 2.2963898\n",
      "epoch: 1 step: 761, loss is 2.2949927\n",
      "epoch: 1 step: 762, loss is 2.3025217\n",
      "epoch: 1 step: 763, loss is 2.3014166\n",
      "epoch: 1 step: 764, loss is 2.2871323\n",
      "epoch: 1 step: 765, loss is 2.3130922\n",
      "epoch: 1 step: 766, loss is 2.295463\n",
      "epoch: 1 step: 767, loss is 2.307328\n",
      "epoch: 1 step: 768, loss is 2.3014517\n",
      "epoch: 1 step: 769, loss is 2.303382\n",
      "epoch: 1 step: 770, loss is 2.3083441\n",
      "epoch: 1 step: 771, loss is 2.3008223\n",
      "epoch: 1 step: 772, loss is 2.295353\n",
      "epoch: 1 step: 773, loss is 2.3029294\n",
      "epoch: 1 step: 774, loss is 2.3065205\n",
      "epoch: 1 step: 775, loss is 2.295032\n",
      "epoch: 1 step: 776, loss is 2.290621\n",
      "epoch: 1 step: 777, loss is 2.304025\n",
      "epoch: 1 step: 778, loss is 2.297865\n",
      "epoch: 1 step: 779, loss is 2.3000555\n",
      "epoch: 1 step: 780, loss is 2.2885468\n",
      "epoch: 1 step: 781, loss is 2.2942421\n",
      "epoch: 1 step: 782, loss is 2.2837758\n",
      "epoch: 1 step: 783, loss is 2.2785351\n",
      "epoch: 1 step: 784, loss is 2.3038654\n",
      "epoch: 1 step: 785, loss is 2.3177552\n",
      "epoch: 1 step: 786, loss is 2.3020754\n",
      "epoch: 1 step: 787, loss is 2.2825086\n",
      "epoch: 1 step: 788, loss is 2.294215\n",
      "epoch: 1 step: 789, loss is 2.3071294\n",
      "epoch: 1 step: 790, loss is 2.2719228\n",
      "epoch: 1 step: 791, loss is 2.2896006\n",
      "epoch: 1 step: 792, loss is 2.2917256\n",
      "epoch: 1 step: 793, loss is 2.287161\n",
      "epoch: 1 step: 794, loss is 2.2772179\n",
      "epoch: 1 step: 795, loss is 2.2895703\n",
      "epoch: 1 step: 796, loss is 2.2942226\n",
      "epoch: 1 step: 797, loss is 2.2760415\n",
      "epoch: 1 step: 798, loss is 2.3147106\n",
      "epoch: 1 step: 799, loss is 2.295553\n",
      "epoch: 1 step: 800, loss is 2.3016484\n",
      "epoch: 1 step: 801, loss is 2.3114347\n",
      "epoch: 1 step: 802, loss is 2.309609\n",
      "epoch: 1 step: 803, loss is 2.2842622\n",
      "epoch: 1 step: 804, loss is 2.2823975\n",
      "epoch: 1 step: 805, loss is 2.3002036\n",
      "epoch: 1 step: 806, loss is 2.2802875\n",
      "epoch: 1 step: 807, loss is 2.2740474\n",
      "epoch: 1 step: 808, loss is 2.2819157\n",
      "epoch: 1 step: 809, loss is 2.290746\n",
      "epoch: 1 step: 810, loss is 2.2921329\n",
      "epoch: 1 step: 811, loss is 2.2876852\n",
      "epoch: 1 step: 812, loss is 2.2973895\n",
      "epoch: 1 step: 813, loss is 2.3077996\n",
      "epoch: 1 step: 814, loss is 2.288783\n",
      "epoch: 1 step: 815, loss is 2.293751\n",
      "epoch: 1 step: 816, loss is 2.2918954\n",
      "epoch: 1 step: 817, loss is 2.2970433\n",
      "epoch: 1 step: 818, loss is 2.3016474\n",
      "epoch: 1 step: 819, loss is 2.287221\n",
      "epoch: 1 step: 820, loss is 2.2835884\n",
      "epoch: 1 step: 821, loss is 2.281673\n",
      "epoch: 1 step: 822, loss is 2.2702115\n",
      "epoch: 1 step: 823, loss is 2.2622437\n",
      "epoch: 1 step: 824, loss is 2.2914815\n",
      "epoch: 1 step: 825, loss is 2.274608\n",
      "epoch: 1 step: 826, loss is 2.3027084\n",
      "epoch: 1 step: 827, loss is 2.277711\n",
      "epoch: 1 step: 828, loss is 2.2905033\n",
      "epoch: 1 step: 829, loss is 2.2724183\n",
      "epoch: 1 step: 830, loss is 2.262413\n",
      "epoch: 1 step: 831, loss is 2.2677562\n",
      "epoch: 1 step: 832, loss is 2.2899811\n",
      "epoch: 1 step: 833, loss is 2.2761147\n",
      "epoch: 1 step: 834, loss is 2.2625613\n",
      "epoch: 1 step: 835, loss is 2.2585483\n",
      "epoch: 1 step: 836, loss is 2.2625985\n",
      "epoch: 1 step: 837, loss is 2.2462459\n",
      "epoch: 1 step: 838, loss is 2.2383316\n",
      "epoch: 1 step: 839, loss is 2.2362328\n",
      "epoch: 1 step: 840, loss is 2.2618058\n",
      "epoch: 1 step: 841, loss is 2.2541654\n",
      "epoch: 1 step: 842, loss is 2.2698278\n",
      "epoch: 1 step: 843, loss is 2.2408156\n",
      "epoch: 1 step: 844, loss is 2.187005\n",
      "epoch: 1 step: 845, loss is 2.226328\n",
      "epoch: 1 step: 846, loss is 2.2494578\n",
      "epoch: 1 step: 847, loss is 2.18062\n",
      "epoch: 1 step: 848, loss is 2.2452517\n",
      "epoch: 1 step: 849, loss is 2.3042376\n",
      "epoch: 1 step: 850, loss is 2.1845973\n",
      "epoch: 1 step: 851, loss is 2.205352\n",
      "epoch: 1 step: 852, loss is 2.186552\n",
      "epoch: 1 step: 853, loss is 2.0561802\n",
      "epoch: 1 step: 854, loss is 2.068172\n",
      "epoch: 1 step: 855, loss is 2.1811347\n",
      "epoch: 1 step: 856, loss is 2.0707438\n",
      "epoch: 1 step: 857, loss is 2.1002336\n",
      "epoch: 1 step: 858, loss is 1.8924875\n",
      "epoch: 1 step: 859, loss is 2.1339006\n",
      "epoch: 1 step: 860, loss is 1.7597522\n",
      "epoch: 1 step: 861, loss is 1.916883\n",
      "epoch: 1 step: 862, loss is 1.9835427\n",
      "epoch: 1 step: 863, loss is 1.8475674\n",
      "epoch: 1 step: 864, loss is 1.6478062\n",
      "epoch: 1 step: 865, loss is 2.2501626\n",
      "epoch: 1 step: 866, loss is 1.5490353\n",
      "epoch: 1 step: 867, loss is 2.0394492\n",
      "epoch: 1 step: 868, loss is 2.0791183\n",
      "epoch: 1 step: 869, loss is 1.6989685\n",
      "epoch: 1 step: 870, loss is 2.0145597\n",
      "epoch: 1 step: 871, loss is 1.8304303\n",
      "epoch: 1 step: 872, loss is 2.312253\n",
      "epoch: 1 step: 873, loss is 1.8823107\n",
      "epoch: 1 step: 874, loss is 1.7674185\n",
      "epoch: 1 step: 875, loss is 1.8107212\n",
      "epoch: 1 step: 876, loss is 1.8160436\n",
      "epoch: 1 step: 877, loss is 1.8629727\n",
      "epoch: 1 step: 878, loss is 1.8265799\n",
      "epoch: 1 step: 879, loss is 1.9635834\n",
      "epoch: 1 step: 880, loss is 1.6989686\n",
      "epoch: 1 step: 881, loss is 1.6135037\n",
      "epoch: 1 step: 882, loss is 1.9003522\n",
      "epoch: 1 step: 883, loss is 1.751629\n",
      "epoch: 1 step: 884, loss is 1.7907355\n",
      "epoch: 1 step: 885, loss is 1.9996349\n",
      "epoch: 1 step: 886, loss is 1.7002867\n",
      "epoch: 1 step: 887, loss is 1.5135735\n",
      "epoch: 1 step: 888, loss is 1.747382\n",
      "epoch: 1 step: 889, loss is 1.5088216\n",
      "epoch: 1 step: 890, loss is 1.5707498\n",
      "epoch: 1 step: 891, loss is 1.1628383\n",
      "epoch: 1 step: 892, loss is 1.3351604\n",
      "epoch: 1 step: 893, loss is 1.2702289\n",
      "epoch: 1 step: 894, loss is 1.2823908\n",
      "epoch: 1 step: 895, loss is 1.219234\n",
      "epoch: 1 step: 896, loss is 1.125713\n",
      "epoch: 1 step: 897, loss is 1.0116317\n",
      "epoch: 1 step: 898, loss is 1.4350045\n",
      "epoch: 1 step: 899, loss is 1.523111\n",
      "epoch: 1 step: 900, loss is 1.5645814\n",
      "epoch: 1 step: 901, loss is 1.96007\n",
      "epoch: 1 step: 902, loss is 1.4109422\n",
      "epoch: 1 step: 903, loss is 1.5345273\n",
      "epoch: 1 step: 904, loss is 1.1789153\n",
      "epoch: 1 step: 905, loss is 1.4632277\n",
      "epoch: 1 step: 906, loss is 1.4485998\n",
      "epoch: 1 step: 907, loss is 1.1467905\n",
      "epoch: 1 step: 908, loss is 1.3450749\n",
      "epoch: 1 step: 909, loss is 1.3621356\n",
      "epoch: 1 step: 910, loss is 1.2990779\n",
      "epoch: 1 step: 911, loss is 1.4800344\n",
      "epoch: 1 step: 912, loss is 1.6163831\n",
      "epoch: 1 step: 913, loss is 1.2741752\n",
      "epoch: 1 step: 914, loss is 1.0923687\n",
      "epoch: 1 step: 915, loss is 1.2469255\n",
      "epoch: 1 step: 916, loss is 1.0773689\n",
      "epoch: 1 step: 917, loss is 0.87907875\n",
      "epoch: 1 step: 918, loss is 0.88364774\n",
      "epoch: 1 step: 919, loss is 1.1096395\n",
      "epoch: 1 step: 920, loss is 1.5142299\n",
      "epoch: 1 step: 921, loss is 1.5571738\n",
      "epoch: 1 step: 922, loss is 1.0765679\n",
      "epoch: 1 step: 923, loss is 0.856564\n",
      "epoch: 1 step: 924, loss is 1.0117152\n",
      "epoch: 1 step: 925, loss is 1.5034511\n",
      "epoch: 1 step: 926, loss is 0.9170122\n",
      "epoch: 1 step: 927, loss is 1.1024946\n",
      "epoch: 1 step: 928, loss is 1.1239359\n",
      "epoch: 1 step: 929, loss is 1.103117\n",
      "epoch: 1 step: 930, loss is 0.9770429\n",
      "epoch: 1 step: 931, loss is 1.262664\n",
      "epoch: 1 step: 932, loss is 0.8041433\n",
      "epoch: 1 step: 933, loss is 0.935048\n",
      "epoch: 1 step: 934, loss is 0.75911117\n",
      "epoch: 1 step: 935, loss is 0.7567863\n",
      "epoch: 1 step: 936, loss is 0.93873054\n",
      "epoch: 1 step: 937, loss is 0.8306789\n",
      "epoch: 1 step: 938, loss is 1.0800498\n",
      "epoch: 1 step: 939, loss is 1.107067\n",
      "epoch: 1 step: 940, loss is 0.6703848\n",
      "epoch: 1 step: 941, loss is 1.3019388\n",
      "epoch: 1 step: 942, loss is 0.755839\n",
      "epoch: 1 step: 943, loss is 0.8514472\n",
      "epoch: 1 step: 944, loss is 0.758751\n",
      "epoch: 1 step: 945, loss is 0.5247954\n",
      "epoch: 1 step: 946, loss is 0.8065117\n",
      "epoch: 1 step: 947, loss is 0.6568605\n",
      "epoch: 1 step: 948, loss is 0.6803141\n",
      "epoch: 1 step: 949, loss is 0.84117275\n",
      "epoch: 1 step: 950, loss is 0.6742877\n",
      "epoch: 1 step: 951, loss is 0.5383589\n",
      "epoch: 1 step: 952, loss is 0.9961929\n",
      "epoch: 1 step: 953, loss is 0.82625717\n",
      "epoch: 1 step: 954, loss is 0.7956822\n",
      "epoch: 1 step: 955, loss is 0.60098994\n",
      "epoch: 1 step: 956, loss is 0.41206518\n",
      "epoch: 1 step: 957, loss is 0.891399\n",
      "epoch: 1 step: 958, loss is 1.0385866\n",
      "epoch: 1 step: 959, loss is 0.5596\n",
      "epoch: 1 step: 960, loss is 0.8061647\n",
      "epoch: 1 step: 961, loss is 0.66609156\n",
      "epoch: 1 step: 962, loss is 0.48925188\n",
      "epoch: 1 step: 963, loss is 0.5132302\n",
      "epoch: 1 step: 964, loss is 0.398133\n",
      "epoch: 1 step: 965, loss is 0.546315\n",
      "epoch: 1 step: 966, loss is 0.8728975\n",
      "epoch: 1 step: 967, loss is 0.40564066\n",
      "epoch: 1 step: 968, loss is 0.61154956\n",
      "epoch: 1 step: 969, loss is 0.47586235\n",
      "epoch: 1 step: 970, loss is 0.6082181\n",
      "epoch: 1 step: 971, loss is 0.6652099\n",
      "epoch: 1 step: 972, loss is 0.6129927\n",
      "epoch: 1 step: 973, loss is 0.54125834\n",
      "epoch: 1 step: 974, loss is 0.7356366\n",
      "epoch: 1 step: 975, loss is 0.42982894\n",
      "epoch: 1 step: 976, loss is 0.77939165\n",
      "epoch: 1 step: 977, loss is 0.66934925\n",
      "epoch: 1 step: 978, loss is 0.68424666\n",
      "epoch: 1 step: 979, loss is 0.4005886\n",
      "epoch: 1 step: 980, loss is 0.6534044\n",
      "epoch: 1 step: 981, loss is 0.8669177\n",
      "epoch: 1 step: 982, loss is 0.34319177\n",
      "epoch: 1 step: 983, loss is 0.2632935\n",
      "epoch: 1 step: 984, loss is 0.5050466\n",
      "epoch: 1 step: 985, loss is 0.483961\n",
      "epoch: 1 step: 986, loss is 0.6676218\n",
      "epoch: 1 step: 987, loss is 0.51638585\n",
      "epoch: 1 step: 988, loss is 0.5111797\n",
      "epoch: 1 step: 989, loss is 0.6092554\n",
      "epoch: 1 step: 990, loss is 0.52976257\n",
      "epoch: 1 step: 991, loss is 0.48427296\n",
      "epoch: 1 step: 992, loss is 0.5987566\n",
      "epoch: 1 step: 993, loss is 0.46781301\n",
      "epoch: 1 step: 994, loss is 0.793617\n",
      "epoch: 1 step: 995, loss is 0.40782598\n",
      "epoch: 1 step: 996, loss is 0.8390678\n",
      "epoch: 1 step: 997, loss is 0.4645052\n",
      "epoch: 1 step: 998, loss is 0.28586888\n",
      "epoch: 1 step: 999, loss is 0.24029538\n",
      "epoch: 1 step: 1000, loss is 0.24771957\n",
      "epoch: 1 step: 1001, loss is 0.2775087\n",
      "epoch: 1 step: 1002, loss is 0.43482822\n",
      "epoch: 1 step: 1003, loss is 0.7205806\n",
      "epoch: 1 step: 1004, loss is 0.53696877\n",
      "epoch: 1 step: 1005, loss is 0.48667476\n",
      "epoch: 1 step: 1006, loss is 0.7680383\n",
      "epoch: 1 step: 1007, loss is 0.43496725\n",
      "epoch: 1 step: 1008, loss is 0.40782598\n",
      "epoch: 1 step: 1009, loss is 0.42481676\n",
      "epoch: 1 step: 1010, loss is 0.42777532\n",
      "epoch: 1 step: 1011, loss is 0.2651268\n",
      "epoch: 1 step: 1012, loss is 0.7326452\n",
      "epoch: 1 step: 1013, loss is 0.4556732\n",
      "epoch: 1 step: 1014, loss is 0.44973922\n",
      "epoch: 1 step: 1015, loss is 0.58492124\n",
      "epoch: 1 step: 1016, loss is 0.65810984\n",
      "epoch: 1 step: 1017, loss is 0.39078867\n",
      "epoch: 1 step: 1018, loss is 0.36803812\n",
      "epoch: 1 step: 1019, loss is 0.45131096\n",
      "epoch: 1 step: 1020, loss is 0.43583098\n",
      "epoch: 1 step: 1021, loss is 0.3816088\n",
      "epoch: 1 step: 1022, loss is 0.29335684\n",
      "epoch: 1 step: 1023, loss is 0.577488\n",
      "epoch: 1 step: 1024, loss is 0.35376105\n",
      "epoch: 1 step: 1025, loss is 0.4463961\n",
      "epoch: 1 step: 1026, loss is 0.19638999\n",
      "epoch: 1 step: 1027, loss is 0.40797007\n",
      "epoch: 1 step: 1028, loss is 0.29675192\n",
      "epoch: 1 step: 1029, loss is 0.32451487\n",
      "epoch: 1 step: 1030, loss is 0.38516042\n",
      "epoch: 1 step: 1031, loss is 0.1727915\n",
      "epoch: 1 step: 1032, loss is 0.16765311\n",
      "epoch: 1 step: 1033, loss is 0.28395718\n",
      "epoch: 1 step: 1034, loss is 0.40852985\n",
      "epoch: 1 step: 1035, loss is 0.58658063\n",
      "epoch: 1 step: 1036, loss is 0.36349064\n",
      "epoch: 1 step: 1037, loss is 0.42598987\n",
      "epoch: 1 step: 1038, loss is 0.84372765\n",
      "epoch: 1 step: 1039, loss is 0.6160477\n",
      "epoch: 1 step: 1040, loss is 0.64532274\n",
      "epoch: 1 step: 1041, loss is 0.3730353\n",
      "epoch: 1 step: 1042, loss is 0.24602337\n",
      "epoch: 1 step: 1043, loss is 0.6055812\n",
      "epoch: 1 step: 1044, loss is 0.27826613\n",
      "epoch: 1 step: 1045, loss is 0.4080508\n",
      "epoch: 1 step: 1046, loss is 0.67776775\n",
      "epoch: 1 step: 1047, loss is 0.66588885\n",
      "epoch: 1 step: 1048, loss is 0.6095781\n",
      "epoch: 1 step: 1049, loss is 0.35304475\n",
      "epoch: 1 step: 1050, loss is 0.15197946\n",
      "epoch: 1 step: 1051, loss is 0.36475903\n",
      "epoch: 1 step: 1052, loss is 0.5347757\n",
      "epoch: 1 step: 1053, loss is 0.33682033\n",
      "epoch: 1 step: 1054, loss is 0.50436884\n",
      "epoch: 1 step: 1055, loss is 0.265473\n",
      "epoch: 1 step: 1056, loss is 0.5491859\n",
      "epoch: 1 step: 1057, loss is 0.4096243\n",
      "epoch: 1 step: 1058, loss is 0.3073904\n",
      "epoch: 1 step: 1059, loss is 0.4753292\n",
      "epoch: 1 step: 1060, loss is 0.42157272\n",
      "epoch: 1 step: 1061, loss is 0.24225923\n",
      "epoch: 1 step: 1062, loss is 0.42127454\n",
      "epoch: 1 step: 1063, loss is 0.26800916\n",
      "epoch: 1 step: 1064, loss is 0.3277799\n",
      "epoch: 1 step: 1065, loss is 0.5203651\n",
      "epoch: 1 step: 1066, loss is 0.85034716\n",
      "epoch: 1 step: 1067, loss is 0.32029378\n",
      "epoch: 1 step: 1068, loss is 0.3525036\n",
      "epoch: 1 step: 1069, loss is 0.3827795\n",
      "epoch: 1 step: 1070, loss is 0.16786695\n",
      "epoch: 1 step: 1071, loss is 0.37292385\n",
      "epoch: 1 step: 1072, loss is 0.17177047\n",
      "epoch: 1 step: 1073, loss is 0.47043383\n",
      "epoch: 1 step: 1074, loss is 0.45905688\n",
      "epoch: 1 step: 1075, loss is 0.36163568\n",
      "epoch: 1 step: 1076, loss is 0.21719839\n",
      "epoch: 1 step: 1077, loss is 0.53339493\n",
      "epoch: 1 step: 1078, loss is 0.4105552\n",
      "epoch: 1 step: 1079, loss is 0.568854\n",
      "epoch: 1 step: 1080, loss is 0.6397742\n",
      "epoch: 1 step: 1081, loss is 0.31131473\n",
      "epoch: 1 step: 1082, loss is 0.37624416\n",
      "epoch: 1 step: 1083, loss is 0.60565007\n",
      "epoch: 1 step: 1084, loss is 0.18607262\n",
      "epoch: 1 step: 1085, loss is 0.39079294\n",
      "epoch: 1 step: 1086, loss is 0.32334352\n",
      "epoch: 1 step: 1087, loss is 0.33626655\n",
      "epoch: 1 step: 1088, loss is 0.17269741\n",
      "epoch: 1 step: 1089, loss is 0.3980743\n",
      "epoch: 1 step: 1090, loss is 0.22321436\n",
      "epoch: 1 step: 1091, loss is 0.34364727\n",
      "epoch: 1 step: 1092, loss is 0.3758012\n",
      "epoch: 1 step: 1093, loss is 0.21214257\n",
      "epoch: 1 step: 1094, loss is 0.267317\n",
      "epoch: 1 step: 1095, loss is 0.5219899\n",
      "epoch: 1 step: 1096, loss is 0.3477505\n",
      "epoch: 1 step: 1097, loss is 0.09403778\n",
      "epoch: 1 step: 1098, loss is 0.6218136\n",
      "epoch: 1 step: 1099, loss is 0.14303952\n",
      "epoch: 1 step: 1100, loss is 0.26137254\n",
      "epoch: 1 step: 1101, loss is 0.13829537\n",
      "epoch: 1 step: 1102, loss is 0.66898227\n",
      "epoch: 1 step: 1103, loss is 0.31300277\n",
      "epoch: 1 step: 1104, loss is 0.18368171\n",
      "epoch: 1 step: 1105, loss is 0.681355\n",
      "epoch: 1 step: 1106, loss is 0.1972735\n",
      "epoch: 1 step: 1107, loss is 0.3852156\n",
      "epoch: 1 step: 1108, loss is 0.1932149\n",
      "epoch: 1 step: 1109, loss is 0.18293574\n",
      "epoch: 1 step: 1110, loss is 0.26467636\n",
      "epoch: 1 step: 1111, loss is 0.23035428\n",
      "epoch: 1 step: 1112, loss is 0.10199675\n",
      "epoch: 1 step: 1113, loss is 0.21895239\n",
      "epoch: 1 step: 1114, loss is 0.28780058\n",
      "epoch: 1 step: 1115, loss is 0.3115482\n",
      "epoch: 1 step: 1116, loss is 0.21632425\n",
      "epoch: 1 step: 1117, loss is 0.13003153\n",
      "epoch: 1 step: 1118, loss is 0.18697397\n",
      "epoch: 1 step: 1119, loss is 0.36623016\n",
      "epoch: 1 step: 1120, loss is 0.34516373\n",
      "epoch: 1 step: 1121, loss is 0.36823648\n",
      "epoch: 1 step: 1122, loss is 0.5090488\n",
      "epoch: 1 step: 1123, loss is 0.16439858\n",
      "epoch: 1 step: 1124, loss is 0.5019853\n",
      "epoch: 1 step: 1125, loss is 0.31341758\n",
      "epoch: 1 step: 1126, loss is 0.46539545\n",
      "epoch: 1 step: 1127, loss is 0.2050698\n",
      "epoch: 1 step: 1128, loss is 0.24474895\n",
      "epoch: 1 step: 1129, loss is 0.19920367\n",
      "epoch: 1 step: 1130, loss is 0.3255741\n",
      "epoch: 1 step: 1131, loss is 0.17932846\n",
      "epoch: 1 step: 1132, loss is 0.28100199\n",
      "epoch: 1 step: 1133, loss is 0.91400087\n",
      "epoch: 1 step: 1134, loss is 0.21071365\n",
      "epoch: 1 step: 1135, loss is 0.3766691\n",
      "epoch: 1 step: 1136, loss is 0.46541223\n",
      "epoch: 1 step: 1137, loss is 0.2259051\n",
      "epoch: 1 step: 1138, loss is 0.27766672\n",
      "epoch: 1 step: 1139, loss is 0.4859815\n",
      "epoch: 1 step: 1140, loss is 0.40704483\n",
      "epoch: 1 step: 1141, loss is 0.49902132\n",
      "epoch: 1 step: 1142, loss is 0.3555463\n",
      "epoch: 1 step: 1143, loss is 0.20939822\n",
      "epoch: 1 step: 1144, loss is 0.18129364\n",
      "epoch: 1 step: 1145, loss is 0.39768332\n",
      "epoch: 1 step: 1146, loss is 0.14185889\n",
      "epoch: 1 step: 1147, loss is 0.61285406\n",
      "epoch: 1 step: 1148, loss is 0.35987684\n",
      "epoch: 1 step: 1149, loss is 0.3019188\n",
      "epoch: 1 step: 1150, loss is 0.32990256\n",
      "epoch: 1 step: 1151, loss is 0.4055625\n",
      "epoch: 1 step: 1152, loss is 0.19448988\n",
      "epoch: 1 step: 1153, loss is 0.35750157\n",
      "epoch: 1 step: 1154, loss is 0.325013\n",
      "epoch: 1 step: 1155, loss is 0.19630969\n",
      "epoch: 1 step: 1156, loss is 0.31592605\n",
      "epoch: 1 step: 1157, loss is 0.12473279\n",
      "epoch: 1 step: 1158, loss is 0.48753187\n",
      "epoch: 1 step: 1159, loss is 0.24019119\n",
      "epoch: 1 step: 1160, loss is 0.19086829\n",
      "epoch: 1 step: 1161, loss is 0.26464757\n",
      "epoch: 1 step: 1162, loss is 0.28130355\n",
      "epoch: 1 step: 1163, loss is 0.3402713\n",
      "epoch: 1 step: 1164, loss is 0.18293516\n",
      "epoch: 1 step: 1165, loss is 0.20245409\n",
      "epoch: 1 step: 1166, loss is 0.24487501\n",
      "epoch: 1 step: 1167, loss is 0.195078\n",
      "epoch: 1 step: 1168, loss is 0.29777342\n",
      "epoch: 1 step: 1169, loss is 0.24487859\n",
      "epoch: 1 step: 1170, loss is 0.16964816\n",
      "epoch: 1 step: 1171, loss is 0.12010643\n",
      "epoch: 1 step: 1172, loss is 0.35885292\n",
      "epoch: 1 step: 1173, loss is 0.16041157\n",
      "epoch: 1 step: 1174, loss is 0.2024799\n",
      "epoch: 1 step: 1175, loss is 0.11880303\n",
      "epoch: 1 step: 1176, loss is 0.2867076\n",
      "epoch: 1 step: 1177, loss is 0.053357534\n",
      "epoch: 1 step: 1178, loss is 0.1813792\n",
      "epoch: 1 step: 1179, loss is 0.2630782\n",
      "epoch: 1 step: 1180, loss is 0.20825446\n",
      "epoch: 1 step: 1181, loss is 0.23406103\n",
      "epoch: 1 step: 1182, loss is 0.103949316\n",
      "epoch: 1 step: 1183, loss is 0.25518113\n",
      "epoch: 1 step: 1184, loss is 0.20298633\n",
      "epoch: 1 step: 1185, loss is 0.12963001\n",
      "epoch: 1 step: 1186, loss is 0.5107653\n",
      "epoch: 1 step: 1187, loss is 0.15512712\n",
      "epoch: 1 step: 1188, loss is 0.16172488\n",
      "epoch: 1 step: 1189, loss is 0.23760203\n",
      "epoch: 1 step: 1190, loss is 0.12507851\n",
      "epoch: 1 step: 1191, loss is 0.17830302\n",
      "epoch: 1 step: 1192, loss is 0.19341001\n",
      "epoch: 1 step: 1193, loss is 0.51620233\n",
      "epoch: 1 step: 1194, loss is 0.14384474\n",
      "epoch: 1 step: 1195, loss is 0.06996977\n",
      "epoch: 1 step: 1196, loss is 0.17036422\n",
      "epoch: 1 step: 1197, loss is 0.167764\n",
      "epoch: 1 step: 1198, loss is 0.18177052\n",
      "epoch: 1 step: 1199, loss is 0.34575483\n",
      "epoch: 1 step: 1200, loss is 0.098052025\n",
      "epoch: 1 step: 1201, loss is 0.25842857\n",
      "epoch: 1 step: 1202, loss is 0.25050786\n",
      "epoch: 1 step: 1203, loss is 0.44772896\n",
      "epoch: 1 step: 1204, loss is 0.23032753\n",
      "epoch: 1 step: 1205, loss is 0.19988774\n",
      "epoch: 1 step: 1206, loss is 0.4368616\n",
      "epoch: 1 step: 1207, loss is 0.23655969\n",
      "epoch: 1 step: 1208, loss is 0.027799428\n",
      "epoch: 1 step: 1209, loss is 0.23572099\n",
      "epoch: 1 step: 1210, loss is 0.14992234\n",
      "epoch: 1 step: 1211, loss is 0.34409094\n",
      "epoch: 1 step: 1212, loss is 0.2719796\n",
      "epoch: 1 step: 1213, loss is 0.740222\n",
      "epoch: 1 step: 1214, loss is 0.2733745\n",
      "epoch: 1 step: 1215, loss is 0.4687469\n",
      "epoch: 1 step: 1216, loss is 0.19457835\n",
      "epoch: 1 step: 1217, loss is 0.22090325\n",
      "epoch: 1 step: 1218, loss is 0.02607303\n",
      "epoch: 1 step: 1219, loss is 0.37755856\n",
      "epoch: 1 step: 1220, loss is 0.43837005\n",
      "epoch: 1 step: 1221, loss is 0.1404981\n",
      "epoch: 1 step: 1222, loss is 0.29063606\n",
      "epoch: 1 step: 1223, loss is 0.11416643\n",
      "epoch: 1 step: 1224, loss is 0.10771347\n",
      "epoch: 1 step: 1225, loss is 0.17314753\n",
      "epoch: 1 step: 1226, loss is 0.15072194\n",
      "epoch: 1 step: 1227, loss is 0.5529535\n",
      "epoch: 1 step: 1228, loss is 0.2871613\n",
      "epoch: 1 step: 1229, loss is 0.25731158\n",
      "epoch: 1 step: 1230, loss is 0.24665855\n",
      "epoch: 1 step: 1231, loss is 0.19748491\n",
      "epoch: 1 step: 1232, loss is 0.20941827\n",
      "epoch: 1 step: 1233, loss is 0.44942963\n",
      "epoch: 1 step: 1234, loss is 0.20718898\n",
      "epoch: 1 step: 1235, loss is 0.29581308\n",
      "epoch: 1 step: 1236, loss is 0.28710204\n",
      "epoch: 1 step: 1237, loss is 0.49254087\n",
      "epoch: 1 step: 1238, loss is 0.1271814\n",
      "epoch: 1 step: 1239, loss is 0.086086296\n",
      "epoch: 1 step: 1240, loss is 0.22534324\n",
      "epoch: 1 step: 1241, loss is 0.1217125\n",
      "epoch: 1 step: 1242, loss is 0.123531975\n",
      "epoch: 1 step: 1243, loss is 0.1873155\n",
      "epoch: 1 step: 1244, loss is 0.19624671\n",
      "epoch: 1 step: 1245, loss is 0.18156241\n",
      "epoch: 1 step: 1246, loss is 0.27989408\n",
      "epoch: 1 step: 1247, loss is 0.08275278\n",
      "epoch: 1 step: 1248, loss is 0.14774144\n",
      "epoch: 1 step: 1249, loss is 0.4058478\n",
      "epoch: 1 step: 1250, loss is 0.40073192\n",
      "epoch: 1 step: 1251, loss is 0.28495473\n",
      "epoch: 1 step: 1252, loss is 0.1237053\n",
      "epoch: 1 step: 1253, loss is 0.17548409\n",
      "epoch: 1 step: 1254, loss is 0.05437863\n",
      "epoch: 1 step: 1255, loss is 0.13469903\n",
      "epoch: 1 step: 1256, loss is 0.37818092\n",
      "epoch: 1 step: 1257, loss is 0.113863\n",
      "epoch: 1 step: 1258, loss is 0.122255296\n",
      "epoch: 1 step: 1259, loss is 0.3768829\n",
      "epoch: 1 step: 1260, loss is 0.4544468\n",
      "epoch: 1 step: 1261, loss is 0.43782914\n",
      "epoch: 1 step: 1262, loss is 0.20563641\n",
      "epoch: 1 step: 1263, loss is 0.16872366\n",
      "epoch: 1 step: 1264, loss is 0.16520698\n",
      "epoch: 1 step: 1265, loss is 0.23534334\n",
      "epoch: 1 step: 1266, loss is 0.3269068\n",
      "epoch: 1 step: 1267, loss is 0.2172891\n",
      "epoch: 1 step: 1268, loss is 0.12134034\n",
      "epoch: 1 step: 1269, loss is 0.15150978\n",
      "epoch: 1 step: 1270, loss is 0.058139347\n",
      "epoch: 1 step: 1271, loss is 0.49493116\n",
      "epoch: 1 step: 1272, loss is 0.17863637\n",
      "epoch: 1 step: 1273, loss is 0.15265317\n",
      "epoch: 1 step: 1274, loss is 0.12883079\n",
      "epoch: 1 step: 1275, loss is 0.3263552\n",
      "epoch: 1 step: 1276, loss is 0.17939138\n",
      "epoch: 1 step: 1277, loss is 0.3570179\n",
      "epoch: 1 step: 1278, loss is 0.1442532\n",
      "epoch: 1 step: 1279, loss is 0.067968786\n",
      "epoch: 1 step: 1280, loss is 0.16431057\n",
      "epoch: 1 step: 1281, loss is 0.23393059\n",
      "epoch: 1 step: 1282, loss is 0.13455187\n",
      "epoch: 1 step: 1283, loss is 0.30417395\n",
      "epoch: 1 step: 1284, loss is 0.14489028\n",
      "epoch: 1 step: 1285, loss is 0.069990665\n",
      "epoch: 1 step: 1286, loss is 0.25869936\n",
      "epoch: 1 step: 1287, loss is 0.06395979\n",
      "epoch: 1 step: 1288, loss is 0.28279752\n",
      "epoch: 1 step: 1289, loss is 0.23797067\n",
      "epoch: 1 step: 1290, loss is 0.15145802\n",
      "epoch: 1 step: 1291, loss is 0.27138054\n",
      "epoch: 1 step: 1292, loss is 0.3386876\n",
      "epoch: 1 step: 1293, loss is 0.3196348\n",
      "epoch: 1 step: 1294, loss is 0.23677443\n",
      "epoch: 1 step: 1295, loss is 0.044388697\n",
      "epoch: 1 step: 1296, loss is 0.10516579\n",
      "epoch: 1 step: 1297, loss is 0.27632958\n",
      "epoch: 1 step: 1298, loss is 0.2651959\n",
      "epoch: 1 step: 1299, loss is 0.31447738\n",
      "epoch: 1 step: 1300, loss is 0.20059241\n",
      "epoch: 1 step: 1301, loss is 0.23801805\n",
      "epoch: 1 step: 1302, loss is 0.08839221\n",
      "epoch: 1 step: 1303, loss is 0.5580834\n",
      "epoch: 1 step: 1304, loss is 0.059053287\n",
      "epoch: 1 step: 1305, loss is 0.22305644\n",
      "epoch: 1 step: 1306, loss is 0.2728455\n",
      "epoch: 1 step: 1307, loss is 0.15261532\n",
      "epoch: 1 step: 1308, loss is 0.4921447\n",
      "epoch: 1 step: 1309, loss is 0.8566924\n",
      "epoch: 1 step: 1310, loss is 0.13176116\n",
      "epoch: 1 step: 1311, loss is 0.59724116\n",
      "epoch: 1 step: 1312, loss is 0.23499973\n",
      "epoch: 1 step: 1313, loss is 0.38317046\n",
      "epoch: 1 step: 1314, loss is 0.15177236\n",
      "epoch: 1 step: 1315, loss is 0.1314344\n",
      "epoch: 1 step: 1316, loss is 0.16009866\n",
      "epoch: 1 step: 1317, loss is 0.21556717\n",
      "epoch: 1 step: 1318, loss is 0.3883618\n",
      "epoch: 1 step: 1319, loss is 0.34944972\n",
      "epoch: 1 step: 1320, loss is 0.36038053\n",
      "epoch: 1 step: 1321, loss is 0.23635973\n",
      "epoch: 1 step: 1322, loss is 0.1740986\n",
      "epoch: 1 step: 1323, loss is 0.05727216\n",
      "epoch: 1 step: 1324, loss is 0.18018255\n",
      "epoch: 1 step: 1325, loss is 0.02096558\n",
      "epoch: 1 step: 1326, loss is 0.35229054\n",
      "epoch: 1 step: 1327, loss is 0.26898992\n",
      "epoch: 1 step: 1328, loss is 0.22324985\n",
      "epoch: 1 step: 1329, loss is 0.25657013\n",
      "epoch: 1 step: 1330, loss is 0.1910859\n",
      "epoch: 1 step: 1331, loss is 0.15136643\n",
      "epoch: 1 step: 1332, loss is 0.2028911\n",
      "epoch: 1 step: 1333, loss is 0.032072935\n",
      "epoch: 1 step: 1334, loss is 0.07248228\n",
      "epoch: 1 step: 1335, loss is 0.05776738\n",
      "epoch: 1 step: 1336, loss is 0.12570138\n",
      "epoch: 1 step: 1337, loss is 0.2147425\n",
      "epoch: 1 step: 1338, loss is 0.016995123\n",
      "epoch: 1 step: 1339, loss is 0.33247855\n",
      "epoch: 1 step: 1340, loss is 0.084627\n",
      "epoch: 1 step: 1341, loss is 0.23015916\n",
      "epoch: 1 step: 1342, loss is 0.22639859\n",
      "epoch: 1 step: 1343, loss is 0.12837972\n",
      "epoch: 1 step: 1344, loss is 0.18780792\n",
      "epoch: 1 step: 1345, loss is 0.6471439\n",
      "epoch: 1 step: 1346, loss is 0.12162343\n",
      "epoch: 1 step: 1347, loss is 0.11259567\n",
      "epoch: 1 step: 1348, loss is 0.08486323\n",
      "epoch: 1 step: 1349, loss is 0.019832794\n",
      "epoch: 1 step: 1350, loss is 0.109113015\n",
      "epoch: 1 step: 1351, loss is 0.404473\n",
      "epoch: 1 step: 1352, loss is 0.054653756\n",
      "epoch: 1 step: 1353, loss is 0.5583621\n",
      "epoch: 1 step: 1354, loss is 0.21348472\n",
      "epoch: 1 step: 1355, loss is 0.13636597\n",
      "epoch: 1 step: 1356, loss is 0.18039092\n",
      "epoch: 1 step: 1357, loss is 0.10349581\n",
      "epoch: 1 step: 1358, loss is 0.15027235\n",
      "epoch: 1 step: 1359, loss is 0.017351378\n",
      "epoch: 1 step: 1360, loss is 0.19025806\n",
      "epoch: 1 step: 1361, loss is 0.23969889\n",
      "epoch: 1 step: 1362, loss is 0.3858544\n",
      "epoch: 1 step: 1363, loss is 0.03171436\n",
      "epoch: 1 step: 1364, loss is 0.38455018\n",
      "epoch: 1 step: 1365, loss is 0.13236824\n",
      "epoch: 1 step: 1366, loss is 0.4041299\n",
      "epoch: 1 step: 1367, loss is 0.3610497\n",
      "epoch: 1 step: 1368, loss is 0.31535262\n",
      "epoch: 1 step: 1369, loss is 0.2013173\n",
      "epoch: 1 step: 1370, loss is 0.16973658\n",
      "epoch: 1 step: 1371, loss is 0.18847944\n",
      "epoch: 1 step: 1372, loss is 0.19490762\n",
      "epoch: 1 step: 1373, loss is 0.23822248\n",
      "epoch: 1 step: 1374, loss is 0.16093034\n",
      "epoch: 1 step: 1375, loss is 0.10756436\n",
      "epoch: 1 step: 1376, loss is 0.1143123\n",
      "epoch: 1 step: 1377, loss is 0.29405272\n",
      "epoch: 1 step: 1378, loss is 0.124302916\n",
      "epoch: 1 step: 1379, loss is 0.19737484\n",
      "epoch: 1 step: 1380, loss is 0.13423567\n",
      "epoch: 1 step: 1381, loss is 0.49843505\n",
      "epoch: 1 step: 1382, loss is 0.1767243\n",
      "epoch: 1 step: 1383, loss is 0.23917948\n",
      "epoch: 1 step: 1384, loss is 0.14053988\n",
      "epoch: 1 step: 1385, loss is 0.1844199\n",
      "epoch: 1 step: 1386, loss is 0.36599365\n",
      "epoch: 1 step: 1387, loss is 0.3656305\n",
      "epoch: 1 step: 1388, loss is 0.02392726\n",
      "epoch: 1 step: 1389, loss is 0.11736807\n",
      "epoch: 1 step: 1390, loss is 0.24774295\n",
      "epoch: 1 step: 1391, loss is 0.4284506\n",
      "epoch: 1 step: 1392, loss is 0.15076947\n",
      "epoch: 1 step: 1393, loss is 0.15562923\n",
      "epoch: 1 step: 1394, loss is 0.06988819\n",
      "epoch: 1 step: 1395, loss is 0.118494585\n",
      "epoch: 1 step: 1396, loss is 0.06103547\n",
      "epoch: 1 step: 1397, loss is 0.025950726\n",
      "epoch: 1 step: 1398, loss is 0.03272787\n",
      "epoch: 1 step: 1399, loss is 0.2627942\n",
      "epoch: 1 step: 1400, loss is 0.38103995\n",
      "epoch: 1 step: 1401, loss is 0.22816753\n",
      "epoch: 1 step: 1402, loss is 0.30665272\n",
      "epoch: 1 step: 1403, loss is 0.06672175\n",
      "epoch: 1 step: 1404, loss is 0.23718242\n",
      "epoch: 1 step: 1405, loss is 0.09051269\n",
      "epoch: 1 step: 1406, loss is 0.13866678\n",
      "epoch: 1 step: 1407, loss is 0.063452065\n",
      "epoch: 1 step: 1408, loss is 0.043927975\n",
      "epoch: 1 step: 1409, loss is 0.18343712\n",
      "epoch: 1 step: 1410, loss is 0.18493071\n",
      "epoch: 1 step: 1411, loss is 0.09312214\n",
      "epoch: 1 step: 1412, loss is 0.024791336\n",
      "epoch: 1 step: 1413, loss is 0.13190745\n",
      "epoch: 1 step: 1414, loss is 0.09261114\n",
      "epoch: 1 step: 1415, loss is 0.17346089\n",
      "epoch: 1 step: 1416, loss is 0.2319999\n",
      "epoch: 1 step: 1417, loss is 0.18715186\n",
      "epoch: 1 step: 1418, loss is 0.07728614\n",
      "epoch: 1 step: 1419, loss is 0.059753306\n",
      "epoch: 1 step: 1420, loss is 0.15847766\n",
      "epoch: 1 step: 1421, loss is 0.23588195\n",
      "epoch: 1 step: 1422, loss is 0.16232987\n",
      "epoch: 1 step: 1423, loss is 0.09556182\n",
      "epoch: 1 step: 1424, loss is 0.1647461\n",
      "epoch: 1 step: 1425, loss is 0.078246616\n",
      "epoch: 1 step: 1426, loss is 0.3195501\n",
      "epoch: 1 step: 1427, loss is 0.5479045\n",
      "epoch: 1 step: 1428, loss is 0.12741184\n",
      "epoch: 1 step: 1429, loss is 0.05938091\n",
      "epoch: 1 step: 1430, loss is 0.04010809\n",
      "epoch: 1 step: 1431, loss is 0.09019649\n",
      "epoch: 1 step: 1432, loss is 0.13935211\n",
      "epoch: 1 step: 1433, loss is 0.12239115\n",
      "epoch: 1 step: 1434, loss is 0.23763472\n",
      "epoch: 1 step: 1435, loss is 0.13594289\n",
      "epoch: 1 step: 1436, loss is 0.16978218\n",
      "epoch: 1 step: 1437, loss is 0.25450918\n",
      "epoch: 1 step: 1438, loss is 0.07837727\n",
      "epoch: 1 step: 1439, loss is 0.1796098\n",
      "epoch: 1 step: 1440, loss is 0.04301632\n",
      "epoch: 1 step: 1441, loss is 0.23331013\n",
      "epoch: 1 step: 1442, loss is 0.0408407\n",
      "epoch: 1 step: 1443, loss is 0.02763132\n",
      "epoch: 1 step: 1444, loss is 0.06837091\n",
      "epoch: 1 step: 1445, loss is 0.036505185\n",
      "epoch: 1 step: 1446, loss is 0.13437402\n",
      "epoch: 1 step: 1447, loss is 0.061323658\n",
      "epoch: 1 step: 1448, loss is 0.10246266\n",
      "epoch: 1 step: 1449, loss is 0.511012\n",
      "epoch: 1 step: 1450, loss is 0.25850484\n",
      "epoch: 1 step: 1451, loss is 0.2685744\n",
      "epoch: 1 step: 1452, loss is 0.1916143\n",
      "epoch: 1 step: 1453, loss is 0.11388509\n",
      "epoch: 1 step: 1454, loss is 0.11170808\n",
      "epoch: 1 step: 1455, loss is 0.07661552\n",
      "epoch: 1 step: 1456, loss is 0.090765215\n",
      "epoch: 1 step: 1457, loss is 0.27614677\n",
      "epoch: 1 step: 1458, loss is 0.53361624\n",
      "epoch: 1 step: 1459, loss is 0.12638046\n",
      "epoch: 1 step: 1460, loss is 0.15924461\n",
      "epoch: 1 step: 1461, loss is 0.06981458\n",
      "epoch: 1 step: 1462, loss is 0.13064973\n",
      "epoch: 1 step: 1463, loss is 0.030223887\n",
      "epoch: 1 step: 1464, loss is 0.23442186\n",
      "epoch: 1 step: 1465, loss is 0.3482758\n",
      "epoch: 1 step: 1466, loss is 0.04632106\n",
      "epoch: 1 step: 1467, loss is 0.1378041\n",
      "epoch: 1 step: 1468, loss is 0.17806955\n",
      "epoch: 1 step: 1469, loss is 0.08009932\n",
      "epoch: 1 step: 1470, loss is 0.05575101\n",
      "epoch: 1 step: 1471, loss is 0.28459722\n",
      "epoch: 1 step: 1472, loss is 0.12985027\n",
      "epoch: 1 step: 1473, loss is 0.15847993\n",
      "epoch: 1 step: 1474, loss is 0.07929554\n",
      "epoch: 1 step: 1475, loss is 0.27334186\n",
      "epoch: 1 step: 1476, loss is 0.22089928\n",
      "epoch: 1 step: 1477, loss is 0.14402185\n",
      "epoch: 1 step: 1478, loss is 0.07168905\n",
      "epoch: 1 step: 1479, loss is 0.09636022\n",
      "epoch: 1 step: 1480, loss is 0.13952945\n",
      "epoch: 1 step: 1481, loss is 0.36246562\n",
      "epoch: 1 step: 1482, loss is 0.21308535\n",
      "epoch: 1 step: 1483, loss is 0.2193711\n",
      "epoch: 1 step: 1484, loss is 0.31480852\n",
      "epoch: 1 step: 1485, loss is 0.28956798\n",
      "epoch: 1 step: 1486, loss is 0.07119675\n",
      "epoch: 1 step: 1487, loss is 0.38585618\n",
      "epoch: 1 step: 1488, loss is 0.14348675\n",
      "epoch: 1 step: 1489, loss is 0.16339234\n",
      "epoch: 1 step: 1490, loss is 0.4627546\n",
      "epoch: 1 step: 1491, loss is 0.15343949\n",
      "epoch: 1 step: 1492, loss is 0.31238317\n",
      "epoch: 1 step: 1493, loss is 0.27939704\n",
      "epoch: 1 step: 1494, loss is 0.048022185\n",
      "epoch: 1 step: 1495, loss is 0.19718769\n",
      "epoch: 1 step: 1496, loss is 0.17564905\n",
      "epoch: 1 step: 1497, loss is 0.16773748\n",
      "epoch: 1 step: 1498, loss is 0.13770558\n",
      "epoch: 1 step: 1499, loss is 0.15822032\n",
      "epoch: 1 step: 1500, loss is 0.30684966\n",
      "epoch: 1 step: 1501, loss is 0.14316566\n",
      "epoch: 1 step: 1502, loss is 0.29331002\n",
      "epoch: 1 step: 1503, loss is 0.17284389\n",
      "epoch: 1 step: 1504, loss is 0.08673446\n",
      "epoch: 1 step: 1505, loss is 0.14916626\n",
      "epoch: 1 step: 1506, loss is 0.09635977\n",
      "epoch: 1 step: 1507, loss is 0.06379104\n",
      "epoch: 1 step: 1508, loss is 0.2914315\n",
      "epoch: 1 step: 1509, loss is 0.083853364\n",
      "epoch: 1 step: 1510, loss is 0.12057252\n",
      "epoch: 1 step: 1511, loss is 0.17638841\n",
      "epoch: 1 step: 1512, loss is 0.14076307\n",
      "epoch: 1 step: 1513, loss is 0.03221818\n",
      "epoch: 1 step: 1514, loss is 0.5020863\n",
      "epoch: 1 step: 1515, loss is 0.06942843\n",
      "epoch: 1 step: 1516, loss is 0.3392204\n",
      "epoch: 1 step: 1517, loss is 0.14629944\n",
      "epoch: 1 step: 1518, loss is 0.091503546\n",
      "epoch: 1 step: 1519, loss is 0.17582779\n",
      "epoch: 1 step: 1520, loss is 0.041902367\n",
      "epoch: 1 step: 1521, loss is 0.22925241\n",
      "epoch: 1 step: 1522, loss is 0.081242725\n",
      "epoch: 1 step: 1523, loss is 0.14803237\n",
      "epoch: 1 step: 1524, loss is 0.032160327\n",
      "epoch: 1 step: 1525, loss is 0.05216776\n",
      "epoch: 1 step: 1526, loss is 0.09676644\n",
      "epoch: 1 step: 1527, loss is 0.10596394\n",
      "epoch: 1 step: 1528, loss is 0.12214963\n",
      "epoch: 1 step: 1529, loss is 0.094952025\n",
      "epoch: 1 step: 1530, loss is 0.09447362\n",
      "epoch: 1 step: 1531, loss is 0.13118969\n",
      "epoch: 1 step: 1532, loss is 0.1521298\n",
      "epoch: 1 step: 1533, loss is 0.033969376\n",
      "epoch: 1 step: 1534, loss is 0.013055153\n",
      "epoch: 1 step: 1535, loss is 0.09880686\n",
      "epoch: 1 step: 1536, loss is 0.34078103\n",
      "epoch: 1 step: 1537, loss is 0.1522503\n",
      "epoch: 1 step: 1538, loss is 0.30593607\n",
      "epoch: 1 step: 1539, loss is 0.24309729\n",
      "epoch: 1 step: 1540, loss is 0.1201354\n",
      "epoch: 1 step: 1541, loss is 0.3955048\n",
      "epoch: 1 step: 1542, loss is 0.041609854\n",
      "epoch: 1 step: 1543, loss is 0.14879486\n",
      "epoch: 1 step: 1544, loss is 0.08917011\n",
      "epoch: 1 step: 1545, loss is 0.09700656\n",
      "epoch: 1 step: 1546, loss is 0.23718902\n",
      "epoch: 1 step: 1547, loss is 0.089114375\n",
      "epoch: 1 step: 1548, loss is 0.0632399\n",
      "epoch: 1 step: 1549, loss is 0.3535574\n",
      "epoch: 1 step: 1550, loss is 0.4500244\n",
      "epoch: 1 step: 1551, loss is 0.06916758\n",
      "epoch: 1 step: 1552, loss is 0.18991096\n",
      "epoch: 1 step: 1553, loss is 0.020577528\n",
      "epoch: 1 step: 1554, loss is 0.12788294\n",
      "epoch: 1 step: 1555, loss is 0.14144918\n",
      "epoch: 1 step: 1556, loss is 0.04829053\n",
      "epoch: 1 step: 1557, loss is 0.13407132\n",
      "epoch: 1 step: 1558, loss is 0.2989988\n",
      "epoch: 1 step: 1559, loss is 0.09033528\n",
      "epoch: 1 step: 1560, loss is 0.26432642\n",
      "epoch: 1 step: 1561, loss is 0.055321023\n",
      "epoch: 1 step: 1562, loss is 0.50705963\n",
      "epoch: 1 step: 1563, loss is 0.25080243\n",
      "epoch: 1 step: 1564, loss is 0.2460534\n",
      "epoch: 1 step: 1565, loss is 0.1323284\n",
      "epoch: 1 step: 1566, loss is 0.44129437\n",
      "epoch: 1 step: 1567, loss is 0.31745183\n",
      "epoch: 1 step: 1568, loss is 0.3721412\n",
      "epoch: 1 step: 1569, loss is 0.18746369\n",
      "epoch: 1 step: 1570, loss is 0.16023256\n",
      "epoch: 1 step: 1571, loss is 0.5111789\n",
      "epoch: 1 step: 1572, loss is 0.21591721\n",
      "epoch: 1 step: 1573, loss is 0.079193324\n",
      "epoch: 1 step: 1574, loss is 0.04002896\n",
      "epoch: 1 step: 1575, loss is 0.17946535\n",
      "epoch: 1 step: 1576, loss is 0.13267623\n",
      "epoch: 1 step: 1577, loss is 0.11970565\n",
      "epoch: 1 step: 1578, loss is 0.15569805\n",
      "epoch: 1 step: 1579, loss is 0.1726857\n",
      "epoch: 1 step: 1580, loss is 0.087112114\n",
      "epoch: 1 step: 1581, loss is 0.06994753\n",
      "epoch: 1 step: 1582, loss is 0.17055519\n",
      "epoch: 1 step: 1583, loss is 0.15214217\n",
      "epoch: 1 step: 1584, loss is 0.13856311\n",
      "epoch: 1 step: 1585, loss is 0.11439528\n",
      "epoch: 1 step: 1586, loss is 0.04471768\n",
      "epoch: 1 step: 1587, loss is 0.19697186\n",
      "epoch: 1 step: 1588, loss is 0.1233261\n",
      "epoch: 1 step: 1589, loss is 0.024087928\n",
      "epoch: 1 step: 1590, loss is 0.115088284\n",
      "epoch: 1 step: 1591, loss is 0.018000001\n",
      "epoch: 1 step: 1592, loss is 0.034862876\n",
      "epoch: 1 step: 1593, loss is 0.03065581\n",
      "epoch: 1 step: 1594, loss is 0.20695415\n",
      "epoch: 1 step: 1595, loss is 0.095846534\n",
      "epoch: 1 step: 1596, loss is 0.32968402\n",
      "epoch: 1 step: 1597, loss is 0.22417696\n",
      "epoch: 1 step: 1598, loss is 0.29696387\n",
      "epoch: 1 step: 1599, loss is 0.11173569\n",
      "epoch: 1 step: 1600, loss is 0.408804\n",
      "epoch: 1 step: 1601, loss is 0.22681524\n",
      "epoch: 1 step: 1602, loss is 0.15978694\n",
      "epoch: 1 step: 1603, loss is 0.05013982\n",
      "epoch: 1 step: 1604, loss is 0.045672502\n",
      "epoch: 1 step: 1605, loss is 0.059203874\n",
      "epoch: 1 step: 1606, loss is 0.31651098\n",
      "epoch: 1 step: 1607, loss is 0.029421031\n",
      "epoch: 1 step: 1608, loss is 0.17880872\n",
      "epoch: 1 step: 1609, loss is 0.34115475\n",
      "epoch: 1 step: 1610, loss is 0.41764042\n",
      "epoch: 1 step: 1611, loss is 0.11923853\n",
      "epoch: 1 step: 1612, loss is 0.12065805\n",
      "epoch: 1 step: 1613, loss is 0.15094125\n",
      "epoch: 1 step: 1614, loss is 0.27978835\n",
      "epoch: 1 step: 1615, loss is 0.16987658\n",
      "epoch: 1 step: 1616, loss is 0.17665413\n",
      "epoch: 1 step: 1617, loss is 0.016721187\n",
      "epoch: 1 step: 1618, loss is 0.20450616\n",
      "epoch: 1 step: 1619, loss is 0.12978691\n",
      "epoch: 1 step: 1620, loss is 0.40205517\n",
      "epoch: 1 step: 1621, loss is 0.13747557\n",
      "epoch: 1 step: 1622, loss is 0.26491216\n",
      "epoch: 1 step: 1623, loss is 0.15537567\n",
      "epoch: 1 step: 1624, loss is 0.30461013\n",
      "epoch: 1 step: 1625, loss is 0.06949287\n",
      "epoch: 1 step: 1626, loss is 0.02077031\n",
      "epoch: 1 step: 1627, loss is 0.14378665\n",
      "epoch: 1 step: 1628, loss is 0.34510154\n",
      "epoch: 1 step: 1629, loss is 0.07473027\n",
      "epoch: 1 step: 1630, loss is 0.04347448\n",
      "epoch: 1 step: 1631, loss is 0.20814641\n",
      "epoch: 1 step: 1632, loss is 0.23609953\n",
      "epoch: 1 step: 1633, loss is 0.13128509\n",
      "epoch: 1 step: 1634, loss is 0.36504424\n",
      "epoch: 1 step: 1635, loss is 0.25342664\n",
      "epoch: 1 step: 1636, loss is 0.38141638\n",
      "epoch: 1 step: 1637, loss is 0.2272902\n",
      "epoch: 1 step: 1638, loss is 0.11629833\n",
      "epoch: 1 step: 1639, loss is 0.036421392\n",
      "epoch: 1 step: 1640, loss is 0.056861628\n",
      "epoch: 1 step: 1641, loss is 0.36046502\n",
      "epoch: 1 step: 1642, loss is 0.2342944\n",
      "epoch: 1 step: 1643, loss is 0.15145741\n",
      "epoch: 1 step: 1644, loss is 0.11556755\n",
      "epoch: 1 step: 1645, loss is 0.09407815\n",
      "epoch: 1 step: 1646, loss is 0.24618141\n",
      "epoch: 1 step: 1647, loss is 0.024620527\n",
      "epoch: 1 step: 1648, loss is 0.03348761\n",
      "epoch: 1 step: 1649, loss is 0.34264907\n",
      "epoch: 1 step: 1650, loss is 0.23027119\n",
      "epoch: 1 step: 1651, loss is 0.2060169\n",
      "epoch: 1 step: 1652, loss is 0.13095208\n",
      "epoch: 1 step: 1653, loss is 0.24109623\n",
      "epoch: 1 step: 1654, loss is 0.06883921\n",
      "epoch: 1 step: 1655, loss is 0.43111843\n",
      "epoch: 1 step: 1656, loss is 0.3535302\n",
      "epoch: 1 step: 1657, loss is 0.028467214\n",
      "epoch: 1 step: 1658, loss is 0.46712106\n",
      "epoch: 1 step: 1659, loss is 0.24111724\n",
      "epoch: 1 step: 1660, loss is 0.3493296\n",
      "epoch: 1 step: 1661, loss is 0.15126297\n",
      "epoch: 1 step: 1662, loss is 0.24973804\n",
      "epoch: 1 step: 1663, loss is 0.26643768\n",
      "epoch: 1 step: 1664, loss is 0.08662087\n",
      "epoch: 1 step: 1665, loss is 0.08193225\n",
      "epoch: 1 step: 1666, loss is 0.036313243\n",
      "epoch: 1 step: 1667, loss is 0.10465815\n",
      "epoch: 1 step: 1668, loss is 0.20943648\n",
      "epoch: 1 step: 1669, loss is 0.3126205\n",
      "epoch: 1 step: 1670, loss is 0.05857732\n",
      "epoch: 1 step: 1671, loss is 0.08354486\n",
      "epoch: 1 step: 1672, loss is 0.06806934\n",
      "epoch: 1 step: 1673, loss is 0.23448035\n",
      "epoch: 1 step: 1674, loss is 0.2247771\n",
      "epoch: 1 step: 1675, loss is 0.17726602\n",
      "epoch: 1 step: 1676, loss is 0.097608335\n",
      "epoch: 1 step: 1677, loss is 0.17564885\n",
      "epoch: 1 step: 1678, loss is 0.017570186\n",
      "epoch: 1 step: 1679, loss is 0.39220893\n",
      "epoch: 1 step: 1680, loss is 0.31990772\n",
      "epoch: 1 step: 1681, loss is 0.3191737\n",
      "epoch: 1 step: 1682, loss is 0.41520286\n",
      "epoch: 1 step: 1683, loss is 0.2248671\n",
      "epoch: 1 step: 1684, loss is 0.051752027\n",
      "epoch: 1 step: 1685, loss is 0.04471356\n",
      "epoch: 1 step: 1686, loss is 0.27796736\n",
      "epoch: 1 step: 1687, loss is 0.12455835\n",
      "epoch: 1 step: 1688, loss is 0.34683877\n",
      "epoch: 1 step: 1689, loss is 0.4937389\n",
      "epoch: 1 step: 1690, loss is 0.23877074\n",
      "epoch: 1 step: 1691, loss is 0.06186238\n",
      "epoch: 1 step: 1692, loss is 0.1484989\n",
      "epoch: 1 step: 1693, loss is 0.11527736\n",
      "epoch: 1 step: 1694, loss is 0.12796469\n",
      "epoch: 1 step: 1695, loss is 0.36207634\n",
      "epoch: 1 step: 1696, loss is 0.12245353\n",
      "epoch: 1 step: 1697, loss is 0.033674013\n",
      "epoch: 1 step: 1698, loss is 0.21124087\n",
      "epoch: 1 step: 1699, loss is 0.05177667\n",
      "epoch: 1 step: 1700, loss is 0.3483248\n",
      "epoch: 1 step: 1701, loss is 0.20217904\n",
      "epoch: 1 step: 1702, loss is 0.24331608\n",
      "epoch: 1 step: 1703, loss is 0.21419573\n",
      "epoch: 1 step: 1704, loss is 0.35407358\n",
      "epoch: 1 step: 1705, loss is 0.082868546\n",
      "epoch: 1 step: 1706, loss is 0.07906558\n",
      "epoch: 1 step: 1707, loss is 0.10647953\n",
      "epoch: 1 step: 1708, loss is 0.1007212\n",
      "epoch: 1 step: 1709, loss is 0.14943728\n",
      "epoch: 1 step: 1710, loss is 0.033027243\n",
      "epoch: 1 step: 1711, loss is 0.26331747\n",
      "epoch: 1 step: 1712, loss is 0.26108995\n",
      "epoch: 1 step: 1713, loss is 0.15750758\n",
      "epoch: 1 step: 1714, loss is 0.19396582\n",
      "epoch: 1 step: 1715, loss is 0.32395357\n",
      "epoch: 1 step: 1716, loss is 0.25291452\n",
      "epoch: 1 step: 1717, loss is 0.18621632\n",
      "epoch: 1 step: 1718, loss is 0.42969483\n",
      "epoch: 1 step: 1719, loss is 0.1787869\n",
      "epoch: 1 step: 1720, loss is 0.14787264\n",
      "epoch: 1 step: 1721, loss is 0.09639491\n",
      "epoch: 1 step: 1722, loss is 0.1614046\n",
      "epoch: 1 step: 1723, loss is 0.2581092\n",
      "epoch: 1 step: 1724, loss is 0.32767943\n",
      "epoch: 1 step: 1725, loss is 0.07481319\n",
      "epoch: 1 step: 1726, loss is 0.0842764\n",
      "epoch: 1 step: 1727, loss is 0.17477815\n",
      "epoch: 1 step: 1728, loss is 0.35208675\n",
      "epoch: 1 step: 1729, loss is 0.1192701\n",
      "epoch: 1 step: 1730, loss is 0.06613776\n",
      "epoch: 1 step: 1731, loss is 0.077270955\n",
      "epoch: 1 step: 1732, loss is 0.20168725\n",
      "epoch: 1 step: 1733, loss is 0.18767567\n",
      "epoch: 1 step: 1734, loss is 0.22555543\n",
      "epoch: 1 step: 1735, loss is 0.29379013\n",
      "epoch: 1 step: 1736, loss is 0.10256298\n",
      "epoch: 1 step: 1737, loss is 0.07038352\n",
      "epoch: 1 step: 1738, loss is 0.07165847\n",
      "epoch: 1 step: 1739, loss is 0.13625205\n",
      "epoch: 1 step: 1740, loss is 0.03283938\n",
      "epoch: 1 step: 1741, loss is 0.06385146\n",
      "epoch: 1 step: 1742, loss is 0.085911795\n",
      "epoch: 1 step: 1743, loss is 0.02286525\n",
      "epoch: 1 step: 1744, loss is 0.044718273\n",
      "epoch: 1 step: 1745, loss is 0.18159471\n",
      "epoch: 1 step: 1746, loss is 0.10331383\n",
      "epoch: 1 step: 1747, loss is 0.30849668\n",
      "epoch: 1 step: 1748, loss is 0.5066165\n",
      "epoch: 1 step: 1749, loss is 0.45901626\n",
      "epoch: 1 step: 1750, loss is 0.23225741\n",
      "epoch: 1 step: 1751, loss is 0.30529985\n",
      "epoch: 1 step: 1752, loss is 0.09082366\n",
      "epoch: 1 step: 1753, loss is 0.26820824\n",
      "epoch: 1 step: 1754, loss is 0.423691\n",
      "epoch: 1 step: 1755, loss is 0.06331016\n",
      "epoch: 1 step: 1756, loss is 0.10750544\n",
      "epoch: 1 step: 1757, loss is 0.020671733\n",
      "epoch: 1 step: 1758, loss is 0.108630165\n",
      "epoch: 1 step: 1759, loss is 0.4268688\n",
      "epoch: 1 step: 1760, loss is 0.24350269\n",
      "epoch: 1 step: 1761, loss is 0.27116027\n",
      "epoch: 1 step: 1762, loss is 0.27681768\n",
      "epoch: 1 step: 1763, loss is 0.2965446\n",
      "epoch: 1 step: 1764, loss is 0.1002032\n",
      "epoch: 1 step: 1765, loss is 0.30833852\n",
      "epoch: 1 step: 1766, loss is 0.10936732\n",
      "epoch: 1 step: 1767, loss is 0.07357388\n",
      "epoch: 1 step: 1768, loss is 0.16046534\n",
      "epoch: 1 step: 1769, loss is 0.15835305\n",
      "epoch: 1 step: 1770, loss is 0.19565645\n",
      "epoch: 1 step: 1771, loss is 0.17671187\n",
      "epoch: 1 step: 1772, loss is 0.38011336\n",
      "epoch: 1 step: 1773, loss is 0.1349776\n",
      "epoch: 1 step: 1774, loss is 0.39392158\n",
      "epoch: 1 step: 1775, loss is 0.21422423\n",
      "epoch: 1 step: 1776, loss is 0.19199337\n",
      "epoch: 1 step: 1777, loss is 0.24782489\n",
      "epoch: 1 step: 1778, loss is 0.09482623\n",
      "epoch: 1 step: 1779, loss is 0.13095535\n",
      "epoch: 1 step: 1780, loss is 0.08106783\n",
      "epoch: 1 step: 1781, loss is 0.049946986\n",
      "epoch: 1 step: 1782, loss is 0.06131782\n",
      "epoch: 1 step: 1783, loss is 0.20389895\n",
      "epoch: 1 step: 1784, loss is 0.024215585\n",
      "epoch: 1 step: 1785, loss is 0.16006042\n",
      "epoch: 1 step: 1786, loss is 0.2054097\n",
      "epoch: 1 step: 1787, loss is 0.06469754\n",
      "epoch: 1 step: 1788, loss is 0.13392538\n",
      "epoch: 1 step: 1789, loss is 0.22424562\n",
      "epoch: 1 step: 1790, loss is 0.17115052\n",
      "epoch: 1 step: 1791, loss is 0.1533849\n",
      "epoch: 1 step: 1792, loss is 0.017133618\n",
      "epoch: 1 step: 1793, loss is 0.03388121\n",
      "epoch: 1 step: 1794, loss is 0.22840668\n",
      "epoch: 1 step: 1795, loss is 0.025773516\n",
      "epoch: 1 step: 1796, loss is 0.0606889\n",
      "epoch: 1 step: 1797, loss is 0.13679142\n",
      "epoch: 1 step: 1798, loss is 0.10139223\n",
      "epoch: 1 step: 1799, loss is 0.4241843\n",
      "epoch: 1 step: 1800, loss is 0.107900724\n",
      "epoch: 1 step: 1801, loss is 0.06875972\n",
      "epoch: 1 step: 1802, loss is 0.20851834\n",
      "epoch: 1 step: 1803, loss is 0.052410915\n",
      "epoch: 1 step: 1804, loss is 0.31407413\n",
      "epoch: 1 step: 1805, loss is 0.20003366\n",
      "epoch: 1 step: 1806, loss is 0.09882748\n",
      "epoch: 1 step: 1807, loss is 0.05212685\n",
      "epoch: 1 step: 1808, loss is 0.23128\n",
      "epoch: 1 step: 1809, loss is 0.07993271\n",
      "epoch: 1 step: 1810, loss is 0.03746863\n",
      "epoch: 1 step: 1811, loss is 0.10327284\n",
      "epoch: 1 step: 1812, loss is 0.07729714\n",
      "epoch: 1 step: 1813, loss is 0.025662892\n",
      "epoch: 1 step: 1814, loss is 0.11721676\n",
      "epoch: 1 step: 1815, loss is 0.056005437\n",
      "epoch: 1 step: 1816, loss is 0.11993335\n",
      "epoch: 1 step: 1817, loss is 0.16883074\n",
      "epoch: 1 step: 1818, loss is 0.15227547\n",
      "epoch: 1 step: 1819, loss is 0.13461994\n",
      "epoch: 1 step: 1820, loss is 0.14262095\n",
      "epoch: 1 step: 1821, loss is 0.04300563\n",
      "epoch: 1 step: 1822, loss is 0.15871187\n",
      "epoch: 1 step: 1823, loss is 0.23733716\n",
      "epoch: 1 step: 1824, loss is 0.123742476\n",
      "epoch: 1 step: 1825, loss is 0.2503525\n",
      "epoch: 1 step: 1826, loss is 0.25831383\n",
      "epoch: 1 step: 1827, loss is 0.09324491\n",
      "epoch: 1 step: 1828, loss is 0.062027548\n",
      "epoch: 1 step: 1829, loss is 0.1479994\n",
      "epoch: 1 step: 1830, loss is 0.13003495\n",
      "epoch: 1 step: 1831, loss is 0.12424392\n",
      "epoch: 1 step: 1832, loss is 0.08605524\n",
      "epoch: 1 step: 1833, loss is 0.07561226\n",
      "epoch: 1 step: 1834, loss is 0.099502094\n",
      "epoch: 1 step: 1835, loss is 0.3049611\n",
      "epoch: 1 step: 1836, loss is 0.15874103\n",
      "epoch: 1 step: 1837, loss is 0.08632015\n",
      "epoch: 1 step: 1838, loss is 0.13027969\n",
      "epoch: 1 step: 1839, loss is 0.06964685\n",
      "epoch: 1 step: 1840, loss is 0.04334371\n",
      "epoch: 1 step: 1841, loss is 0.096723594\n",
      "epoch: 1 step: 1842, loss is 0.115592726\n",
      "epoch: 1 step: 1843, loss is 0.01450201\n",
      "epoch: 1 step: 1844, loss is 0.2287153\n",
      "epoch: 1 step: 1845, loss is 0.09150521\n",
      "epoch: 1 step: 1846, loss is 0.124094166\n",
      "epoch: 1 step: 1847, loss is 0.2783916\n",
      "epoch: 1 step: 1848, loss is 0.01774316\n",
      "epoch: 1 step: 1849, loss is 0.033816107\n",
      "epoch: 1 step: 1850, loss is 0.39238694\n",
      "epoch: 1 step: 1851, loss is 0.21530148\n",
      "epoch: 1 step: 1852, loss is 0.11267878\n",
      "epoch: 1 step: 1853, loss is 0.37873915\n",
      "epoch: 1 step: 1854, loss is 0.044494756\n",
      "epoch: 1 step: 1855, loss is 0.115080066\n",
      "epoch: 1 step: 1856, loss is 0.22031042\n",
      "epoch: 1 step: 1857, loss is 0.11368644\n",
      "epoch: 1 step: 1858, loss is 0.21604615\n",
      "epoch: 1 step: 1859, loss is 0.036978744\n",
      "epoch: 1 step: 1860, loss is 0.14253253\n",
      "epoch: 1 step: 1861, loss is 0.14583558\n",
      "epoch: 1 step: 1862, loss is 0.2481359\n",
      "epoch: 1 step: 1863, loss is 0.22360936\n",
      "epoch: 1 step: 1864, loss is 0.027104443\n",
      "epoch: 1 step: 1865, loss is 0.2002303\n",
      "epoch: 1 step: 1866, loss is 0.055771165\n",
      "epoch: 1 step: 1867, loss is 0.051496465\n",
      "epoch: 1 step: 1868, loss is 0.10295154\n",
      "epoch: 1 step: 1869, loss is 0.03374002\n",
      "epoch: 1 step: 1870, loss is 0.14970702\n",
      "epoch: 1 step: 1871, loss is 0.30618373\n",
      "epoch: 1 step: 1872, loss is 0.11556213\n",
      "epoch: 1 step: 1873, loss is 0.013095432\n",
      "epoch: 1 step: 1874, loss is 0.36053973\n",
      "epoch: 1 step: 1875, loss is 0.056543835\n"
     ]
    }
   ],
   "source": [
    "\n",
    "model.train(epochs, ds_train, callbacks=[ckpoint_cb, LossMonitor()], dataset_sink_mode=False)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "fWC-oi-DdpzN"
   },
   "source": [
    "## 5. Model Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Anaz27cjdpzO",
    "outputId": "5a0d7a8c-59b5-4a52-c4b8-f94dac3ae539",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "First_Contact_with_MindSpore_12Apr2020_23_30.ipynb  etc    root\r\n",
      "MNIST_Data\t\t\t\t\t    home   run\r\n",
      "analyze_fail.dat\t\t\t\t    lib    sbin\r\n",
      "bin\t\t\t\t\t\t    lib64  srv\r\n",
      "boot\t\t\t\t\t\t    media  sys\r\n",
      "checkpoint_lenet-1_1875.ckpt\t\t\t    mnt    tmp\r\n",
      "checkpoint_lenet-graph.meta\t\t\t    opt    usr\r\n",
      "dev\t\t\t\t\t\t    proc   var\r\n"
     ]
    }
   ],
   "source": [
    "!ls"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9r51xVzddpzP",
    "outputId": "f5c55608-aab6-4363-cf8f-a5b1e6a8d6ae"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'Accuracy': 0.961738782051282}\n"
     ]
    }
   ],
   "source": [
    "from mindspore.train.serialization import load_checkpoint, load_param_into_net\n",
    "\n",
    "# load the saved model for evaluation\n",
    "param_dict = load_checkpoint(\"checkpoint_lenet-1_1875.ckpt\")\n",
    "# load parameter to the network\n",
    "load_param_into_net(network, param_dict)\n",
    "# load testing dataset\n",
    "acc = model.eval(ds_eval, dataset_sink_mode=False)\n",
    "print(\"{}\".format(acc))"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "First-Contact-with-MindSpore-10Apr2020-darrer.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
} 