mnist_birk.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "sqH2SEO-X9T6"
      },
      "cell_type": "markdown",
      "source": [
        "# Neural Networks on MNIST\n",
        "\n",
        "This Jupyter notebook explains various approaches for implementing neural networks that recognize digits on [MNIST](http://yann.lecun.com/exdb/mnist/) dataset."
      ]
    },
    {
      "metadata": {
        "id": "OaLeoEcDYVO_"
      },
      "cell_type": "markdown",
      "source": [
        "## Preparing the MNIST dataset\n",
        "\n",
        "Most deep learning frameworks provide APIs for loading famous datasets like MNIST (e.g., `torchvision.datasets.MNIST` in pytorch). The APIs are handy, but hide the important step for preparing a training data for a deep learning framework; when graduating from an example dataset to the real data, we must convert a training data of our interest into the data structure that is acceptable by a deep learning framework.\n",
        "\n",
        "The cell below downloads the original distribution of the MNIST dataset on the Web, converts the dataset into `numpy` arrays, and saves the arrays as the file `mnist.npz` with keyword names."
      ]
    },
    {
      "metadata": {
        "id": "OMaOcZMBPuQY"
      },
      "cell_type": "code",
      "source": [
        "import gzip\n",
        "import os\n",
        "import sys\n",
        "import struct\n",
        "import numpy as np\n",
        "\n",
        "def read_image(fi):\n",
        "    magic, n, rows, columns = struct.unpack(\">IIII\", fi.read(16))\n",
        "    assert magic == 0x00000803\n",
        "    assert rows == 28\n",
        "    assert columns == 28\n",
        "    rawbuffer = fi.read()\n",
        "    assert len(rawbuffer) == n * rows * columns\n",
        "    rawdata = np.frombuffer(rawbuffer, dtype='>u1', count=n*rows*columns)\n",
        "    return rawdata.reshape(n, rows, columns).astype(np.float32) / 255.0\n",
        "\n",
        "def read_label(fi):\n",
        "    magic, n = struct.unpack(\">II\", fi.read(8))\n",
        "    assert magic == 0x00000801\n",
        "    rawbuffer = fi.read()\n",
        "    assert len(rawbuffer) == n\n",
        "    return np.frombuffer(rawbuffer, dtype='>u1', count=n)\n",
        "\n",
        "if __name__ == '__main__':\n",
        "    os.system('wget -N http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz')\n",
        "    os.system('wget -N http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz')\n",
        "    os.system('wget -N http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz')\n",
        "    os.system('wget -N http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz')\n",
        "\n",
        "    np.savez_compressed(\n",
        "        'mnist',\n",
        "        train_x=read_image(gzip.open('train-images-idx3-ubyte.gz', 'rb')),\n",
        "        train_y=read_label(gzip.open('train-labels-idx1-ubyte.gz', 'rb')),\n",
        "        test_x=read_image(gzip.open('t10k-images-idx3-ubyte.gz', 'rb')),\n",
        "        test_y=read_label(gzip.open('t10k-labels-idx1-ubyte.gz', 'rb'))\n",
        "    )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "TStlGwaUaZKC"
      },
      "cell_type": "markdown",
      "source": [
        "The file contains four numpy arrays (one tensor and array for each split of training and test sets) with the keywords:\n",
        "\n",
        "+ `train_x`: $60000 \\text{ (images)} \\times 28 \\text{ (y)} \\times 28 \\text{ (x)}$\n",
        "+ `train_y`: $60000 \\text{ (labels)}$\n",
        "+ `test_x`: $10000 \\text{ (images)} \\times 28 \\text{ (y)} \\times 28 \\text{ (x)}$\n",
        "+ `test_y`: $10000 \\text{ (labels)}$\n"
      ]
    },
    {
      "metadata": {
        "id": "2RbqOtJsZH35",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "51f8ade2-0c40-4cf2-8662-58f533117de8"
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "data = np.load('mnist.npz')\n",
        "\n",
        "print(data['train_x'].shape, data['train_x'].dtype)\n",
        "print(data['train_y'].shape, data['train_y'].dtype)\n",
        "print(data['test_x'].shape, data['test_x'].dtype)\n",
        "print(data['test_y'].shape, data['test_y'].dtype)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(60000, 28, 28) float32\n",
            "(60000,) uint8\n",
            "(10000, 28, 28) float32\n",
            "(10000,) uint8\n"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "_EgmYVMP_iKU"
      },
      "cell_type": "markdown",
      "source": [
        "## Visualize the dataset"
      ]
    },
    {
      "metadata": {
        "id": "Pr_UACmc3Qgc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "37118e27-a964-4d6c-cfbe-66d6d121093a"
      },
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import seaborn as sns\n",
        "\n",
        "# Index number of an instance (change this to view another instance).\n",
        "i = 0\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "image = data['train_x'][i]\n",
        "label = data['train_y'][i]\n",
        "\n",
        "print(label)\n",
        "f, ax = plt.subplots(figsize=(16, 16))\n",
        "sns.heatmap(image, annot=True, fmt='.1f', square=True, cmap=\"YlGnBu\")\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "5\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1600x1600 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "id": "NckT0WSxISfK"
      },
      "cell_type": "markdown",
      "source": [
        "## Install a library for plotting loss and accuracy values"
      ]
    },
    {
      "metadata": {
        "id": "uRvi94eAHBq8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8a810ec6-493e-4dcc-8f0a-a40fa80700b2"
      },
      "cell_type": "code",
      "source": [
        "!pip install livelossplot"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting livelossplot\n",
            "  Downloading livelossplot-0.5.5-py3-none-any.whl (22 kB)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.10/dist-packages (from livelossplot) (3.7.1)\n",
            "Requirement already satisfied: bokeh in /usr/local/lib/python3.10/dist-packages (from livelossplot) (3.2.2)\n",
            "Requirement already satisfied: Jinja2>=2.9 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (3.1.2)\n",
            "Requirement already satisfied: contourpy>=1 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (1.1.0)\n",
            "Requirement already satisfied: numpy>=1.16 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (1.23.5)\n",
            "Requirement already satisfied: packaging>=16.8 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (23.1)\n",
            "Requirement already satisfied: pandas>=1.2 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (1.5.3)\n",
            "Requirement already satisfied: pillow>=7.1.0 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (9.4.0)\n",
            "Requirement already satisfied: PyYAML>=3.10 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (6.0.1)\n",
            "Requirement already satisfied: tornado>=5.1 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (6.3.2)\n",
            "Requirement already satisfied: xyzservices>=2021.09.1 in /usr/local/lib/python3.10/dist-packages (from bokeh->livelossplot) (2023.7.0)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.10/dist-packages (from matplotlib->livelossplot) (0.11.0)\n",
            "Requirement already satisfied: fonttools>=4.22.0 in /usr/local/lib/python3.10/dist-packages (from matplotlib->livelossplot) (4.42.1)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->livelossplot) (1.4.5)\n",
            "Requirement already satisfied: pyparsing>=2.3.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->livelossplot) (3.1.1)\n",
            "Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.10/dist-packages (from matplotlib->livelossplot) (2.8.2)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from Jinja2>=2.9->bokeh->livelossplot) (2.1.3)\n",
            "Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.10/dist-packages (from pandas>=1.2->bokeh->livelossplot) (2023.3.post1)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-packages (from python-dateutil>=2.7->matplotlib->livelossplot) (1.16.0)\n",
            "Installing collected packages: livelossplot\n",
            "Successfully installed livelossplot-0.5.5\n"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "YCr4T9-8cTFH"
      },
      "cell_type": "markdown",
      "source": [
        "## Perceptron algorithm for single-layer neural network"
      ]
    },
    {
      "metadata": {
        "id": "jr3tb8zIP6H3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "d92d8e92-df78-4cd1-dbeb-e87ef1451a63"
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "from livelossplot import PlotLosses\n",
        "\n",
        "def image_to_vector(X):\n",
        "    X = np.reshape(X, (len(X), -1))     # Flatten: (N x 28 x 28) -> (N x 784)\n",
        "    return np.c_[X, np.ones(len(X))]    # Append 1: (N x 784) -> (N x 785)\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "Xtrain = image_to_vector(data['train_x'])\n",
        "Ytrain = data['train_y']\n",
        "Xtest = image_to_vector(data['test_x'])\n",
        "Ytest = data['test_y']\n",
        "\n",
        "W = np.random.randn(10, 28*28+1)\n",
        "\n",
        "eta = 0.001\n",
        "liveloss = PlotLosses()\n",
        "for t in range(100):\n",
        "    # Structured perceptron for updating weights.\n",
        "    num_correct_train = 0\n",
        "    for x, y in zip(Xtrain, Ytrain):\n",
        "        y_pred = np.argmax(np.dot(W, x))\n",
        "        if y_pred != y:\n",
        "            W[y] += x * eta\n",
        "            W[y_pred] -= x * eta\n",
        "        else:\n",
        "            num_correct_train += 1\n",
        "\n",
        "    # Evaluate and store the accuracy on the test set.\n",
        "    num_correct_test = 0\n",
        "    for x, y in zip(Xtest, Ytest):\n",
        "        y_pred = np.argmax(np.dot(W, x))\n",
        "        if y_pred == y:\n",
        "            num_correct_test += 1\n",
        "\n",
        "    # Visualize accuracy values on the training and test sets.\n",
        "    liveloss.update({\n",
        "        'accuracy': float(num_correct_train) / len(Ytrain),\n",
        "        'val_accuracy': float(num_correct_test) / len(Ytest)\n",
        "    })\n",
        "    liveloss.draw()\n",
        "\n",
        "print('Accuracy: {:.4f} (test), {:.4f} (train)'.format(\n",
        "    float(num_correct_test) / len(Ytest),\n",
        "    float(num_correct_train) / len(Ytrain)\n",
        "    ))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x800 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "accuracy\n",
            "\ttraining         \t (min:    0.582, max:    0.889, cur:    0.889)\n",
            "\tvalidation       \t (min:    0.755, max:    0.886, cur:    0.886)\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-5-585285162a08>\u001b[0m in \u001b[0;36m<cell line: 18>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     20\u001b[0m     \u001b[0mnum_correct_train\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mXtrain\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mYtrain\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 22\u001b[0;31m         \u001b[0my_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     23\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0my_pred\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     24\u001b[0m             \u001b[0mW\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mx\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0meta\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/numpy/core/overrides.py\u001b[0m in \u001b[0;36margmax\u001b[0;34m(*args, **kwargs)\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/numpy/core/fromnumeric.py\u001b[0m in \u001b[0;36margmax\u001b[0;34m(a, axis, out, keepdims)\u001b[0m\n\u001b[1;32m   1214\u001b[0m     \"\"\"\n\u001b[1;32m   1215\u001b[0m     \u001b[0mkwds\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0;34m'keepdims'\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mkeepdims\u001b[0m\u001b[0;34m}\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mkeepdims\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_NoValue\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1216\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0m_wrapfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'argmax'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0maxis\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1217\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1218\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.10/dist-packages/numpy/core/fromnumeric.py\u001b[0m in \u001b[0;36m_wrapfunc\u001b[0;34m(obj, method, *args, **kwds)\u001b[0m\n\u001b[1;32m     49\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     50\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 51\u001b[0;31m \u001b[0;32mdef\u001b[0m \u001b[0m_wrapfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mobj\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     52\u001b[0m     \u001b[0mbound\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mobj\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     53\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mbound\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "jk8DoJxPdXax"
      },
      "cell_type": "markdown",
      "source": [
        "## Stochastic gradient descent for single-layer neural network"
      ]
    },
    {
      "metadata": {
        "id": "mZtrUqeSROVa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 341
        },
        "outputId": "248ab809-5432-450a-9e9a-e70cb56d6c96"
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "from livelossplot import PlotLosses\n",
        "\n",
        "def softmax(x):\n",
        "    # Result of softmax are invariant even if we add/subtract a constant.\n",
        "    ex = np.exp(x - np.max(x))  # Subtract such that the maximum value is one.\n",
        "    return ex / ex.sum(axis=0)\n",
        "\n",
        "def image_to_vector(X):\n",
        "    X = np.reshape(X, (len(X), -1))     # Flatten: (N x 28 x 28) -> (N x 784)\n",
        "    return np.c_[X, np.ones(len(X))]    # Append 1: (N x 784) -> (N x 785)\n",
        "\n",
        "def label_to_onehot(Y, K):\n",
        "    return np.eye(K)[Y]              # e.g., 3 -> [0, 0, 0, 1, 0, 0, 0, 0, 0]\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "Xtrain = image_to_vector(data['train_x'])\n",
        "Ytrain = label_to_onehot(data['train_y'], 10)\n",
        "Xtest = image_to_vector(data['test_x'])\n",
        "Ytest = data['test_y']\n",
        "\n",
        "W = np.random.randn(10, 28*28+1)\n",
        "\n",
        "eta = 0.001\n",
        "liveloss = PlotLosses()\n",
        "for t in range(100):\n",
        "    num_correct_train = 0\n",
        "\n",
        "    # Stochastic gradient descent.\n",
        "    for x, y in zip(Xtrain, Ytrain):\n",
        "        y_pred = softmax(np.dot(W, x))\n",
        "        W += np.outer(eta * (y - y_pred), x)\n",
        "        if  np.argmax(y_pred) == np.argmax(y):\n",
        "            num_correct_train += 1\n",
        "\n",
        "    # Evaluate and store the accuracy on the test set.\n",
        "    num_correct_test = 0\n",
        "    for x, y in zip(Xtest, Ytest):\n",
        "        y_pred = np.argmax(np.dot(W, x))\n",
        "        if y_pred == y:\n",
        "            num_correct_test += 1\n",
        "\n",
        "    # Visualize accuracy values on the training and test sets.\n",
        "    liveloss.update({\n",
        "        'accuracy': float(num_correct_train) / len(Ytrain),\n",
        "        'val_accuracy': float(num_correct_test) / len(Ytest)\n",
        "    })\n",
        "    liveloss.draw()\n",
        "\n",
        "print('Accuracy: {:.4f} (test), {:.4f} (train)'.format(\n",
        "    float(num_correct_test) / len(Ytest),\n",
        "    float(num_correct_train) / len(Ytrain)\n",
        "    ))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbgAAAExCAYAAAAOdbejAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH2Jl7PWFdmb0AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fc50b706e48>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.9182 (test), 0.9246 (train)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "twPUp-bbeww2"
      },
      "cell_type": "markdown",
      "source": [
        "## Training with pytorch"
      ]
    },
    {
      "metadata": {
        "id": "QwF495MIe0-e"
      },
      "cell_type": "markdown",
      "source": [
        "### Install pytorch"
      ]
    },
    {
      "metadata": {
        "id": "h8U839iGXCH3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "83092568-9df4-4de2-b587-f0c0982e8b64"
      },
      "cell_type": "code",
      "source": [
        "!pip install torch torchvision"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: torch in /usr/local/lib/python3.10/dist-packages (2.0.1+cu118)\n",
            "Requirement already satisfied: torchvision in /usr/local/lib/python3.10/dist-packages (0.15.2+cu118)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch) (3.12.2)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.10/dist-packages (from torch) (4.5.0)\n",
            "Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch) (1.12)\n",
            "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch) (3.1)\n",
            "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch) (3.1.2)\n",
            "Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.10/dist-packages (from torch) (2.0.0)\n",
            "Requirement already satisfied: cmake in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch) (3.27.4.1)\n",
            "Requirement already satisfied: lit in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch) (16.0.6)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from torchvision) (1.23.5)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from torchvision) (2.31.0)\n",
            "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.10/dist-packages (from torchvision) (9.4.0)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch) (2.1.3)\n",
            "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (3.2.0)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (3.4)\n",
            "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (2.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (2023.7.22)\n",
            "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch) (1.3.0)\n"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "xvQg0U8GMUYS"
      },
      "cell_type": "markdown",
      "source": [
        "### Convert the numpy arrays into pytorch tensor"
      ]
    },
    {
      "metadata": {
        "id": "MWIINUezQn6H"
      },
      "cell_type": "code",
      "source": [
        "def create_dataset(x, y, flatten=False):\n",
        "    if flatten:\n",
        "        # Convert it into a matrix (N [samples], 28*28 [dims])\n",
        "        xt = torch.from_numpy(x).view(len(x), -1)\n",
        "    else:\n",
        "        # Convert it into a 4D tensor (N [samples], 1 [ch], 28 [px], 28 [px])\n",
        "        xt = torch.from_numpy(x).unsqueeze(1)\n",
        "    yt = torch.from_numpy(y).long()\n",
        "    return TensorDataset(xt, yt)\n",
        "\n",
        "def test_model(model, loss_fn, test_loader, device):\n",
        "    model.eval()\n",
        "\n",
        "    loss = 0.\n",
        "    num_correct = 0.\n",
        "    for batch_idx, (x, y) in enumerate(test_loader):\n",
        "        x, y = x.to(device), y.to(device)\n",
        "        y_pred = model(x)\n",
        "        loss += loss_fn(y_pred, y).item()\n",
        "        _, predicted = torch.max(y_pred.data, 1)\n",
        "        num_correct += (predicted == y).sum().item()\n",
        "\n",
        "    model.train()\n",
        "    loss /= len(test_loader.dataset)\n",
        "    num_correct /= len(test_loader.dataset)\n",
        "    return loss, num_correct"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Sr0ejX6YM7ej"
      },
      "cell_type": "markdown",
      "source": [
        "### Single-layer neural network"
      ]
    },
    {
      "metadata": {
        "id": "cTkt86amTqCP"
      },
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "from torch.utils.data import TensorDataset, DataLoader\n",
        "import numpy as np\n",
        "\n",
        "#device = torch.device(\"cpu\") # Uncomment this to run on CPU\n",
        "device = torch.device(\"cuda:0\") # Uncomment this to run on GPU\n",
        "\n",
        "def create_dataset(x, y):\n",
        "    xt = torch.from_numpy(x).view(len(x), -1)\n",
        "    yt = torch.from_numpy(y).long()\n",
        "    return TensorDataset(xt, yt)\n",
        "\n",
        "model = nn.Sequential()\n",
        "model.add_module('fc1', nn.Linear(784, 10, bias=True))\n",
        "model.to(device)\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "train_set = create_dataset(data['train_x'], data['train_y'])\n",
        "train_loader = DataLoader(train_set, batch_size=256, shuffle=True)\n",
        "\n",
        "loss_fn = nn.CrossEntropyLoss(size_average=False)\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.001)\n",
        "\n",
        "for t in range(100):\n",
        "    # Training loop for mini-batches\n",
        "    for batch_idx, (x, y) in enumerate(train_loader):\n",
        "        # Make predictions with the current parameters.\n",
        "        x, y = x.to(device), y.to(device)\n",
        "        y_pred = model(x)\n",
        "\n",
        "        # Compute the loss value.\n",
        "        loss = loss_fn(y_pred, y)\n",
        "\n",
        "        # Update the parameters.\n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "VWeH4oiZW_nf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 293
        },
        "outputId": "6fe51e3e-3f0b-400a-c859-32e5592655dd"
      },
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "from torch.utils.data import TensorDataset, DataLoader\n",
        "import numpy as np\n",
        "from livelossplot import PlotLosses\n",
        "\n",
        "#device = torch.device(\"cpu\") # Uncomment this to run on CPU\n",
        "device = torch.device(\"cuda:0\") # Uncomment this to run on GPU\n",
        "\n",
        "model = nn.Sequential()\n",
        "model.add_module('fc1', nn.Linear(784, 10, bias=True))\n",
        "print(model)\n",
        "model.to(device)\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "train_set = create_dataset(data['train_x'], data['train_y'], flatten=True)\n",
        "test_set = create_dataset(data['test_x'], data['test_y'], flatten=True)\n",
        "train_loader = DataLoader(train_set, batch_size=256, shuffle=True)\n",
        "test_loader = DataLoader(test_set, batch_size=128)\n",
        "\n",
        "loss_fn = nn.CrossEntropyLoss(size_average=False)\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.001)\n",
        "\n",
        "liveloss = PlotLosses()\n",
        "for t in range(100):\n",
        "    train_loss = 0.\n",
        "    train_correct = 0\n",
        "\n",
        "    # Training loop for mini-batches\n",
        "    for batch_idx, (x, y) in enumerate(train_loader):\n",
        "        # Make predictions with the current parameters.\n",
        "        x, y = x.to(device), y.to(device)\n",
        "        y_pred = model(x)\n",
        "        _, predicted = torch.max(y_pred.data, 1)\n",
        "        train_correct += (predicted == y).sum().item()\n",
        "\n",
        "        # Compute the loss value.\n",
        "        loss = loss_fn(y_pred, y)\n",
        "        train_loss += loss.item()\n",
        "\n",
        "        # Update the parameters.\n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "    # Compute the average loss and accuracy.\n",
        "    train_loss /= len(train_loader.dataset)\n",
        "    train_correct /= float(len(train_loader.dataset))\n",
        "\n",
        "    # Evaluate the model on the test set.\n",
        "    test_loss, test_correct = test_model(model, loss_fn, test_loader, device)\n",
        "\n",
        "    # Visualize the loss and accuracy values.\n",
        "    liveloss.update({\n",
        "        'log loss': train_loss,\n",
        "        'val_log loss': test_loss,\n",
        "        'accuracy': train_correct,\n",
        "        'val_accuracy': test_correct,\n",
        "    })\n",
        "    liveloss.draw()\n",
        "\n",
        "print('Accuracy: {:.4f} (test), {:.4f} (train)'.format(test_correct, train_correct))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Sequential(\n",
            "  (fc1): Linear(in_features=784, out_features=10, bias=True)\n",
            ")\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "TypeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-10-87cdc8205e55>\u001b[0m in \u001b[0;36m<cell line: 17>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     15\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     16\u001b[0m \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mload\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'mnist.npz'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 17\u001b[0;31m \u001b[0mtrain_set\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcreate_dataset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'train_x'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'train_y'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mflatten\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     18\u001b[0m \u001b[0mtest_set\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcreate_dataset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'test_x'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdata\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'test_y'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mflatten\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m \u001b[0mtrain_loader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataLoader\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_set\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m256\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mTypeError\u001b[0m: create_dataset() got an unexpected keyword argument 'flatten'"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "3tEreUaGItQr"
      },
      "cell_type": "markdown",
      "source": [
        "### Three-layer neural network"
      ]
    },
    {
      "metadata": {
        "id": "cIaz8ZITRcJ2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 341
        },
        "outputId": "2a506787-3e8f-4d20-8524-3e0c24d8e78c"
      },
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "from torch.utils.data import TensorDataset, DataLoader\n",
        "import numpy as np\n",
        "from livelossplot import PlotLosses\n",
        "\n",
        "#device = torch.device(\"cpu\") # Uncomment this to run on CPU\n",
        "device = torch.device(\"cuda:0\") # Uncomment this to run on GPU\n",
        "\n",
        "model = nn.Sequential()\n",
        "model.add_module('fc1',      nn.Linear(784, 256))\n",
        "model.add_module('relu1',    nn.ReLU())\n",
        "model.add_module('dropout1', nn.Dropout())\n",
        "model.add_module('fc2',      nn.Linear(256, 256))\n",
        "model.add_module('relu2',    nn.ReLU())\n",
        "model.add_module('dropout2', nn.Dropout())\n",
        "model.add_module('fc3',      nn.Linear(256, 10))\n",
        "print(model)\n",
        "model.to(device)\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "train_set = create_dataset(data['train_x'], data['train_y'], flatten=True)\n",
        "test_set = create_dataset(data['test_x'], data['test_y'], flatten=True)\n",
        "train_loader = DataLoader(train_set, batch_size=256, shuffle=True)\n",
        "test_loader = DataLoader(test_set, batch_size=128)\n",
        "\n",
        "loss_fn = nn.CrossEntropyLoss(size_average=False)\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.001)\n",
        "\n",
        "liveloss = PlotLosses()\n",
        "for t in range(100):\n",
        "    train_loss = 0.\n",
        "    train_correct = 0\n",
        "\n",
        "    # Training loop for mini-batches\n",
        "    for batch_idx, (x, y) in enumerate(train_loader):\n",
        "        # Make predictions with the current parameters.\n",
        "        x, y = x.to(device), y.to(device)\n",
        "        y_pred = model(x)\n",
        "        _, predicted = torch.max(y_pred.data, 1)\n",
        "        train_correct += (predicted == y).sum().item()\n",
        "\n",
        "        # Compute the loss value.\n",
        "        loss = loss_fn(y_pred, y)\n",
        "        train_loss += loss.item()\n",
        "\n",
        "        # Update the parameters.\n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "    # Compute the average loss and accuracy.\n",
        "    train_loss /= len(train_loader.dataset)\n",
        "    train_correct /= float(len(train_loader.dataset))\n",
        "\n",
        "    # Evaluate the model on the test set.\n",
        "    test_loss, test_correct = test_model(model, loss_fn, test_loader, device)\n",
        "\n",
        "    # Visualize the loss and accuracy values.\n",
        "    liveloss.update({\n",
        "        'log loss': train_loss,\n",
        "        'val_log loss': test_loss,\n",
        "        'accuracy': train_correct,\n",
        "        'val_accuracy': test_correct,\n",
        "    })\n",
        "    liveloss.draw()\n",
        "\n",
        "print('Accuracy: {:.4f} (test), {:.4f} (train)'.format(test_correct, train_correct))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1QAAAExCAYAAAB7x+OdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fc50b0daac8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.9832 (test), 0.9908 (train)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "kjBd7K4lI3W5"
      },
      "cell_type": "markdown",
      "source": [
        "### Convolutional Neural Network (CNN)"
      ]
    },
    {
      "metadata": {
        "id": "v8H-xg9pom28",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 341
        },
        "outputId": "a720061a-2f17-4bbf-a8d1-6c37e7d60a63"
      },
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "from torch.utils.data import TensorDataset, DataLoader\n",
        "import numpy as np\n",
        "from livelossplot import PlotLosses\n",
        "\n",
        "#device = torch.device(\"cpu\") # Uncomment this to run on CPU\n",
        "device = torch.device(\"cuda:0\") # Uncomment this to run on GPU\n",
        "\n",
        "class Flatten(torch.nn.Module):\n",
        "    def __init__(self):\n",
        "        super(Flatten, self).__init__()\n",
        "\n",
        "    def forward(self, x):\n",
        "        return x.view(-1, 512)\n",
        "\n",
        "model = torch.nn.Sequential(\n",
        "    torch.nn.Conv2d(1, 16, (5, 5)),\n",
        "    torch.nn.MaxPool2d(2),\n",
        "    torch.nn.ReLU(),\n",
        "    torch.nn.Dropout(),\n",
        "    torch.nn.Conv2d(16, 32, (5, 5)),\n",
        "    torch.nn.MaxPool2d(2),\n",
        "    torch.nn.ReLU(),\n",
        "    torch.nn.Dropout(),\n",
        "    Flatten(),\n",
        "    torch.nn.Linear(512, 256),\n",
        "    torch.nn.ReLU(),\n",
        "    torch.nn.Dropout(),\n",
        "    torch.nn.Linear(256, 10),\n",
        ")\n",
        "print(model)\n",
        "model.to(device)\n",
        "\n",
        "data = np.load('mnist.npz')\n",
        "train_set = create_dataset(data['train_x'], data['train_y'])\n",
        "test_set = create_dataset(data['test_x'], data['test_y'])\n",
        "train_loader = DataLoader(train_set, batch_size=256, shuffle=True)\n",
        "test_loader = DataLoader(test_set, batch_size=128)\n",
        "\n",
        "loss_fn = nn.CrossEntropyLoss(size_average=False)\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.001)\n",
        "\n",
        "liveloss = PlotLosses()\n",
        "for t in range(100):\n",
        "    train_loss = 0.\n",
        "    train_correct = 0\n",
        "\n",
        "    # Training loop for mini-batches\n",
        "    for batch_idx, (x, y) in enumerate(train_loader):\n",
        "        # Make predictions with the current parameters.\n",
        "        x, y = x.to(device), y.to(device)\n",
        "        y_pred = model(x)\n",
        "        _, predicted = torch.max(y_pred.data, 1)\n",
        "        train_correct += (predicted == y).sum().item()\n",
        "\n",
        "        # Compute the loss value.\n",
        "        loss = loss_fn(y_pred, y)\n",
        "        train_loss += loss.item()\n",
        "\n",
        "        # Update the parameters.\n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "    # Compute the average loss and accuracy.\n",
        "    train_loss /= len(train_loader.dataset)\n",
        "    train_correct /= float(len(train_loader.dataset))\n",
        "\n",
        "    # Evaluate the model on the test set.\n",
        "    test_loss, test_correct = test_model(model, loss_fn, test_loader, device)\n",
        "\n",
        "    # Visualize the loss and accuracy values.\n",
        "    liveloss.update({\n",
        "        'log loss': train_loss,\n",
        "        'val_log loss': test_loss,\n",
        "        'accuracy': train_correct,\n",
        "        'val_accuracy': test_correct,\n",
        "    })\n",
        "    liveloss.draw()\n",
        "\n",
        "print('Accuracy: {:.4f} (test), {:.4f} (train)'.format(test_correct, train_correct))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1QAAAExCAYAAAB7x+OdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nqHLkbuysvAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fc50b690f28>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.9949 (test), 0.9856 (train)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "data = \"Transformers is a media franchise produced by American toy company Hasbro and Japanese toy company Takara Tomy   It primarily follows the heroic Autobots and the villainous Decepticons   two alien robot factions at war that can transform into other forms   such as vehicles and animals   The franchise encompasses toys   animation   comic books   video games and films   As of            it generated more than      trillion          billion   in revenue         making it one of the highest  grossing media franchises of all time  The franchise began in          with the Transformers toy line   comprising transforming mecha toys from Takara  s Diaclone and Micro Change toylines rebranded for Western markets         The term   Generation      covers both the animated television series The Transformers and the comic book series of the same name   which are further divided into Japanese   British and Canadian spin  offs   respectively   Sequels followed   such as the Generation    comic book and Beast Wars TV series   which became its own mini  universe   Generation    characters underwent two reboots with Dreamwave Productions in          and IDW Publishing in            with a third starting in            There have been other incarnations of the story based on different toy lines during and after the     th century   The first was the Robots in Disguise series   followed by three shows   Armada   Energon   and Cybertron   that constitute a single universe called the   Unicron Trilogy    A live  action film series started in            again distinct from previous incarnations   while the Transformers   Animated series merged concepts from the G   continuity   the          live  action film and the   Unicron Trilogy     For most of the         s   in an attempt to mitigate the wave of reboots   the   Aligned Continuity   was established   In            Transformers   Cyberverse debuted   once again   distinct from the previous incarnations  Although initially a separate and competing franchise started in            Tonka  s GoBots became the intellectual property of Hasbro after their buyout of Tonka in            Subsequently   the universe depicted in the animated series Challenge of the GoBots and follow  up film GoBots   Battle of the Rock Lords was retroactively established as an alternate universe within the Transformers multiverse        FictionTransformers   Generation                          Main articles   Transformers   Generation    and The Transformers   TV series  Classic Transformers franchise logo used until         Spider  Man battles Megatron on the cover of The Transformers       Generation One is a retroactive term for the Transformers characters that appeared between          and            The Transformers began with the         s Japanese toy lines Micro Change and Diaclone   They presented robots able to transform into everyday vehicles   electronic items or weapons   Hasbro bought the Micro Change and Diaclone toys   and partnered with Takara         Marvel Comics was hired by Hasbro to create the backstory   editor  in  chief Jim Shooter wrote an overall story   and gave the task of creating the characters to writer Dennis O  Neil         Unhappy with O  Neil  s work   although O  Neil created the name   Optimus Prime       Shooter chose Bob Budiansky to create the characters        The Transformers mecha were largely designed by Shōji Kawamori   the creator of the Japanese mecha anime franchise Macross   which was adapted into the Robotech franchise in North America           Kawamori came up with the idea of transforming mechs while working on the Diaclone and Macross franchises in the early         s   such as the VF     Valkyrie in Macross and Robotech     with his Diaclone mechs later providing the basis for Transformers        The primary concept of Generation One is that the heroic Optimus Prime   the villainous Megatron   and their finest soldiers crash land on pre  historic Earth in the Ark and the Nemesis before awakening in            Cybertron hurtling through the Neutral zone as an effect of the war   The Marvel comic was originally part of the main Marvel Universe   with appearances from Spider  Man and Nick Fury   plus some cameos         as well as a visit to the Savage Land          The Transformers TV series began around the same time   Produced by Sunbow Productions and Marvel Productions   later Hasbro Productions   from the start it contradicted Budiansky  s backstories   The TV series shows the Autobots looking for new energy sources   and crash landing as the Decepticons attack           Marvel interpreted the Autobots as destroying a rogue asteroid approaching Cybertron           Shockwave is loyal to Megatron in the TV series   keeping Cybertron in a stalemate during his absence           but in the comic book he attempts to take command of the Decepticons           The TV series would also differ wildly from the origins Budiansky had created for the Dinobots                   the Decepticon turned Autobot Jetfire           known as Skyfire on TV             the Constructicons   who combine to form Devastator                     and Omega Supreme                   The Marvel comic establishes early on that Prime wields the Creation Matrix   which gives life to machines   In the second season   the two  part episode The Key to Vector Sigma introduced the ancient Vector Sigma computer   which served the same original purpose as the Creation Matrix   giving life to Transformers     and its guardian Alpha Trion  In            the cartoon became the film The Transformers   The Movie   which is set in the year            It introduced the Matrix as the   Autobot Matrix of Leadership     as a fatally wounded Prime gives it to Ultra Magnus   however   as Prime dies he drops the matrix   which is then caught by Hot Rod who subsequently becomes Rodimus Prime later on in the film   Unicron   a transformer who devours planets   fears its power and recreates a heavily damaged Megatron as Galvatron   as well as Bombshell or Skywarp becoming Cyclonus   Thundercracker becoming Scourge and two other Insecticons becoming Scourge  s huntsmen   the Sweeps   Eventually   Rodimus Prime takes out the Matrix and destroys Unicron           In the United Kingdom   the weekly comic book interspliced original material to keep up with U  S   reprints           and The Movie provided much new material   Writer Simon Furman proceeded to expand the continuity with movie spin  offs involving the time travelling Galvatron                   The Movie also featured guest voices from Leonard Nimoy as Galvatron   Scatman Crothers as Jazz   Casey Kasem as Cliffjumper   Orson Welles as Unicron and Eric Idle as the leader of the Junkions   Wreck  Gar   though unnamed in the movie     The Transformers theme tune for the film was performed by Lion with   Weird Al   Yankovic adding a song to the soundtrack  The third season followed up The Movie   with the revelation of the Quintessons having used Cybertron as a factory   Their robots rebel   and in time the workers become the Autobots and the soldiers become the Decepticons     Note   This appears to contradict background presented in the first two seasons of the series     It is the Autobots who develop transformation           Due to popular demand           Optimus Prime is resurrected at the conclusion of the third season           and the series ended with a three  episode story arc   However   the Japanese broadcast of the series was supplemented with a newly produced OVA   Scramble City   before creating entirely new series to continue the storyline   ignoring the          end of the American series   The extended Japanese run consisted of The Headmasters   Super  God Masterforce   Victory and Zone   then in illustrated magazine form as Battlestars   Return of Convoy and Operation   Combination   Just as the TV series was wrapping up   Marvel continued to expand its continuity   It followed The Movie  s example by killing Prime         and Megatron           albeit in the present day   Dinobot leader Grimlock takes over as Autobot leader           There was a G  I   Joe crossover         and the limited series The Transformers   Headmasters   which further expanded the scope to the planet Nebulon           It led on to the main title resurrecting Prime as a Powermaster          \"\n",
        "\n",
        "\n",
        "#\"The franchise began in 1984 with the Transformers toy line, comprising transforming mecha toys from Takara's Diaclone and Micro Change toylines rebranded for Western markets.[2] The term Generation 1 covers both the animated television series The Transformers and the comic book series of the same name, which are further divided into Japanese, British and Canadian spin-offs, respectively. Sequels followed, such as the Generation 2 comic book and Beast Wars TV series, which became its own mini-universe. Generation 1 characters underwent two reboots with Dreamwave Productions in 2001 and IDW Publishing in 2005, with a third starting in 2019. There have been other incarnations of the story based on different toy lines during and after the 20th century. The first was the Robots in Disguise series, followed by three shows (Armada, Energon, and Cybertron) that constitute a single universe called the Unicron Trilogy\"\n",
        "\n",
        "data = data.replace(\".\", \"\")\n",
        "data.replace(\",\",\"\")\n",
        "\n",
        "data = data.lower()\n",
        "data = data.split(\" \")\n",
        "\n",
        "\n",
        "uniques = len(set(data))\n",
        "print(uniques)\n",
        "words = list(set(data))\n",
        "\n",
        "numdata = []\n",
        "for word in data:\n",
        "  numdata.append(words.index(word))\n",
        "\n",
        "numdata = torch.IntTensor(numdata)\n",
        "print(numdata)"
      ],
      "metadata": {
        "id": "dSW1Tas9HKyi",
        "outputId": "3c0b9b0e-da22-4e3a-86c6-27ec938ce2d9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "519\n",
            "tensor([381,  41, 322,  ...,   0,   0,   0], dtype=torch.int32)\n"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "DdSGWuLffzEg",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3fcd9a19-4377-4b7b-cd5c-a48a3bc5b8fe"
      },
      "cell_type": "code",
      "source": [
        "\n",
        "import math\n",
        "\n",
        "\n",
        "\n",
        "class self_attention(nn.Module):\n",
        "  def __init__(self):\n",
        "    super().__init__()\n",
        "\n",
        "    self.Q = nn.Linear(12, 12)\n",
        "    self.K = nn.Linear(12, 12)\n",
        "    self.V = nn.Linear(12, 12)\n",
        "    self.softmax = nn.Softmax(dim=0)\n",
        "    self.embedding = nn.Embedding(519,12)\n",
        "    self.flatten = nn.Flatten(start_dim=0)\n",
        "    self.preoutput = nn.Linear(144,512)\n",
        "    self.gelu = nn.GELU()\n",
        "    self.output = nn.Linear(512, 519)\n",
        "\n",
        "  def forward(self, x):\n",
        "    x = self.embedding(x)\n",
        "    x_q = self.Q(x)\n",
        "    x_k = self.K(x)\n",
        "    x_v = self.V(x)\n",
        "    x_qk =  x_q.matmul(x_k.transpose(0,1))/math.sqrt(12)\n",
        "    x = self.softmax(x_qk).matmul(x_v)\n",
        "    x = self.flatten(x)\n",
        "    x=self.preoutput(x)\n",
        "    x = self.gelu(x)\n",
        "    x = self.output(x)\n",
        "\n",
        "    return x\n",
        "\n",
        "\n",
        "self_atten = self_attention()\n",
        "\n",
        "p = self_atten(numdata[:12])\n",
        "\n",
        "optimizer = torch.optim.Adam(self_atten.parameters(), lr = 0.0001)\n",
        "loss = nn.CrossEntropyLoss()\n",
        "\n",
        "for i in range(100):\n",
        "  for i in range(len(numdata)-13):\n",
        "    optimizer.zero_grad()\n",
        "    x = numdata[i:i+12]\n",
        "    y = torch.zeros(519)\n",
        "    y[numdata[i+13]] = 1\n",
        "    y_p = self_atten(x)\n",
        "    l = loss(y_p, y)\n",
        "    print(l)\n",
        "    l.backward()\n",
        "    optimizer.step()\n",
        "\n",
        "\n",
        "p = self_atten(numdata[:12])\n",
        "\n",
        "print(p, words[p.argmax()], data[13])\n",
        "\n",
        "for i in range(len(numdata)-13):\n",
        "  x = numdata[i:i+12]\n",
        "  y = data[i+13]\n",
        "  y_p = self_atten(x)\n",
        "  #print(y, words[y_p.argmax()])\n",
        "  if y != words[y_p.argmax()]:\n",
        "    print(\"Wrong!\")\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1;30;43mStreaming output truncated to the last 5000 lines.\u001b[0m\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.7285e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.4212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.9407e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.2769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(8.3681e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.8517e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.5020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0077, grad_fn=<DivBackward1>)\n",
            "tensor(0.0763, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7921e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.8477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.9020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.0953e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0116, grad_fn=<DivBackward1>)\n",
            "tensor(0.0126, grad_fn=<DivBackward1>)\n",
            "tensor(0.0192, grad_fn=<DivBackward1>)\n",
            "tensor(0.0144, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.9709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.3669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.3801e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.2067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(6.0914e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.0650e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.2702e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.8861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.4186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0050, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0519, grad_fn=<DivBackward1>)\n",
            "tensor(0.0083, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0049, grad_fn=<DivBackward1>)\n",
            "tensor(7.1761e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(6.5325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.0371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6106e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.4848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.3974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.0504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.7219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1260e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.6755e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.5669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1563e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.3285e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.7219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.8913e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.2966e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.1059e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.6570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.4265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.4318e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0188, grad_fn=<DivBackward1>)\n",
            "tensor(2.9900, grad_fn=<DivBackward1>)\n",
            "tensor(0.0076, grad_fn=<DivBackward1>)\n",
            "tensor(0.0105, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(8.8569e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.2636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(2.0861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0045, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0138, grad_fn=<DivBackward1>)\n",
            "tensor(0.0127, grad_fn=<DivBackward1>)\n",
            "tensor(0.0274, grad_fn=<DivBackward1>)\n",
            "tensor(0.0075, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(2.0027e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0065, grad_fn=<DivBackward1>)\n",
            "tensor(0.3023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.0333e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.9431e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.1934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4901e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9789e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.2464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.6477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.3098e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.5550e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.6332e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.9696e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(6.4371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.0967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0091, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0049, grad_fn=<DivBackward1>)\n",
            "tensor(0.0135, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.7022e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.5391e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.2875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9550e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0610e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.0173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7643e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.1484e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.0663e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.7220e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(6.8781e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(1.6928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.2888e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.2888e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.4716e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6822e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.1815e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.4120e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.8943e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.2254e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.9407e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.5656e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.9815e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0070, grad_fn=<DivBackward1>)\n",
            "tensor(0.0057, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0045, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.9339e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(6.8781e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.5153e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.0636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.1655e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0071, grad_fn=<DivBackward1>)\n",
            "tensor(2.9423, grad_fn=<DivBackward1>)\n",
            "tensor(0.0705, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.0045, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(9.1791e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.8014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.6597e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(4.3153e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0051, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(0.0097, grad_fn=<DivBackward1>)\n",
            "tensor(2.4309, grad_fn=<DivBackward1>)\n",
            "tensor(0.1476, grad_fn=<DivBackward1>)\n",
            "tensor(0.0199, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0047, grad_fn=<DivBackward1>)\n",
            "tensor(0.0108, grad_fn=<DivBackward1>)\n",
            "tensor(6.8662e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.0265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.0769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.9377e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.8305e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.3987e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.2860e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.6397e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(3.0994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.9815e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(7.6053e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.1100e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.8943e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.2530e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.0636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3324e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(8.2254e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.9735e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3365e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3007e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9908e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.8517e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.0903, grad_fn=<DivBackward1>)\n",
            "tensor(2.5361, grad_fn=<DivBackward1>)\n",
            "tensor(0.0922, grad_fn=<DivBackward1>)\n",
            "tensor(0.0108, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(6.9735e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.1989e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(6.4133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(3.9457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.2983e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.8981e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(2.3365e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.7734e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(1.5139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0036, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0105, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(4.4941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.9444e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.9219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.5895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.4861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(4.5060e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.1510e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0069, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0101, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.2769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.6464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.5868e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7285e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0047, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.7047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(3.9934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.4845, grad_fn=<DivBackward1>)\n",
            "tensor(1.2196, grad_fn=<DivBackward1>)\n",
            "tensor(2.4182, grad_fn=<DivBackward1>)\n",
            "tensor(0.1007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0112, grad_fn=<DivBackward1>)\n",
            "tensor(0.0254, grad_fn=<DivBackward1>)\n",
            "tensor(0.0179, grad_fn=<DivBackward1>)\n",
            "tensor(0.0118, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0095, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0041, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.3749e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.2370e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.8398e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.6014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5259e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.9481, grad_fn=<DivBackward1>)\n",
            "tensor(0.7840, grad_fn=<DivBackward1>)\n",
            "tensor(0.1127, grad_fn=<DivBackward1>)\n",
            "tensor(0.0251, grad_fn=<DivBackward1>)\n",
            "tensor(0.0594, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(6.2941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.9193e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.1525e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0053, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.3164, grad_fn=<DivBackward1>)\n",
            "tensor(0.1841, grad_fn=<DivBackward1>)\n",
            "tensor(0.3168, grad_fn=<DivBackward1>)\n",
            "tensor(0.0426, grad_fn=<DivBackward1>)\n",
            "tensor(0.0166, grad_fn=<DivBackward1>)\n",
            "tensor(0.0062, grad_fn=<DivBackward1>)\n",
            "tensor(0.0041, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.1961e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.1590e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.4175e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.8848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.3643e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.9761e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9126e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.1948e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.9339e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.7060e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(9.7751e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.4107e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0072, grad_fn=<DivBackward1>)\n",
            "tensor(0.0081, grad_fn=<DivBackward1>)\n",
            "tensor(0.0042, grad_fn=<DivBackward1>)\n",
            "tensor(0.0113, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0052, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0068, grad_fn=<DivBackward1>)\n",
            "tensor(0.0494, grad_fn=<DivBackward1>)\n",
            "tensor(0.1055, grad_fn=<DivBackward1>)\n",
            "tensor(0.1035, grad_fn=<DivBackward1>)\n",
            "tensor(0.1003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0961, grad_fn=<DivBackward1>)\n",
            "tensor(0.0913, grad_fn=<DivBackward1>)\n",
            "tensor(0.0860, grad_fn=<DivBackward1>)\n",
            "tensor(0.0806, grad_fn=<DivBackward1>)\n",
            "tensor(0.0750, grad_fn=<DivBackward1>)\n",
            "tensor(0.0697, grad_fn=<DivBackward1>)\n",
            "tensor(0.0645, grad_fn=<DivBackward1>)\n",
            "tensor(0.0596, grad_fn=<DivBackward1>)\n",
            "tensor(0.0550, grad_fn=<DivBackward1>)\n",
            "tensor(5.7804, grad_fn=<DivBackward1>)\n",
            "tensor(5.5037, grad_fn=<DivBackward1>)\n",
            "tensor(0.2239, grad_fn=<DivBackward1>)\n",
            "tensor(0.2124, grad_fn=<DivBackward1>)\n",
            "tensor(0.0439, grad_fn=<DivBackward1>)\n",
            "tensor(0.0231, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0163, grad_fn=<DivBackward1>)\n",
            "tensor(9.5482e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0069, grad_fn=<DivBackward1>)\n",
            "tensor(0.0103, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.6093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.0623e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.7751e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.1072e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(0.0098, grad_fn=<DivBackward1>)\n",
            "tensor(2.5868e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(8.1178e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.2928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.1391e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.4080e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.2821e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.7656e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.5830e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.8252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(2.1934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0214, grad_fn=<DivBackward1>)\n",
            "tensor(0.0056, grad_fn=<DivBackward1>)\n",
            "tensor(0.0035, grad_fn=<DivBackward1>)\n",
            "tensor(1.7047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0051, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.0128, grad_fn=<DivBackward1>)\n",
            "tensor(0.0151, grad_fn=<DivBackward1>)\n",
            "tensor(0.0493, grad_fn=<DivBackward1>)\n",
            "tensor(2.5973, grad_fn=<DivBackward1>)\n",
            "tensor(0.0253, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(0.0091, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.4873e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.5708e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.7934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(7.2119e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.5086e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1525e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.6716e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.9407e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.2452e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(6.5682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(6.1989e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.2159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.9258e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(4.4226e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.9312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.8729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(7.8678e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(9.4175e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.3378e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.1974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2755e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0090, grad_fn=<DivBackward1>)\n",
            "tensor(6.5325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.9219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(3.3378e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.9219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(2.3246e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.5830e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.5218e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.8000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(1.6451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.7100e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(5.5669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3471e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.2689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0981e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8027e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.6861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9683e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0490e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5378e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.1525e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.9206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0064, grad_fn=<DivBackward1>)\n",
            "tensor(0.0146, grad_fn=<DivBackward1>)\n",
            "tensor(0.0293, grad_fn=<DivBackward1>)\n",
            "tensor(0.0079, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.7949e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.8477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.5616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.1696e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.4702e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0149, grad_fn=<DivBackward1>)\n",
            "tensor(0.0191, grad_fn=<DivBackward1>)\n",
            "tensor(7.8678e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.6597e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(2.9683e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7166e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.9271e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.8742e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.7815e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.4186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.0411e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3603e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.9139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(6.7589e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.5469e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.5099e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.5895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.9312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.8239e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.7299e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.7299e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.6636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.0437e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.3974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.3285e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0099, grad_fn=<DivBackward1>)\n",
            "tensor(4.0941, grad_fn=<DivBackward1>)\n",
            "tensor(0.0113, grad_fn=<DivBackward1>)\n",
            "tensor(0.0660, grad_fn=<DivBackward1>)\n",
            "tensor(8.4397e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1259e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.0371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.6912e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3351e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.6941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.5565e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(4.5895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.7895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.8159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.5060e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0088, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.8678e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.9973e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.0994e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.0041, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0171, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7563e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.4067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.8556e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.6781e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.1033e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.4345e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.4451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3947e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.6291e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3961e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.4318e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3828e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1404e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8675e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1525e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0264, grad_fn=<DivBackward1>)\n",
            "tensor(0.0093, grad_fn=<DivBackward1>)\n",
            "tensor(0.0283, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0100, grad_fn=<DivBackward1>)\n",
            "tensor(3.9339e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.2173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.2875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.8014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.2254e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.6492e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3471e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.0411e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0083, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(8.7973e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.2145e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.0411e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0490e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.0027e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.2557e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.0333e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.7299e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(5.2093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.1239, grad_fn=<DivBackward1>)\n",
            "tensor(0.0326, grad_fn=<DivBackward1>)\n",
            "tensor(0.0140, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(4.7206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.5391e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.6294e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0101, grad_fn=<DivBackward1>)\n",
            "tensor(0.0284, grad_fn=<DivBackward1>)\n",
            "tensor(0.0157, grad_fn=<DivBackward1>)\n",
            "tensor(0.0451, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(5.0067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.8412e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3484e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.8926e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.2159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5259e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.4638e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0056, grad_fn=<DivBackward1>)\n",
            "tensor(0.0310, grad_fn=<DivBackward1>)\n",
            "tensor(0.1787, grad_fn=<DivBackward1>)\n",
            "tensor(0.5761, grad_fn=<DivBackward1>)\n",
            "tensor(0.5339, grad_fn=<DivBackward1>)\n",
            "tensor(0.4712, grad_fn=<DivBackward1>)\n",
            "tensor(0.3989, grad_fn=<DivBackward1>)\n",
            "tensor(0.3265, grad_fn=<DivBackward1>)\n",
            "tensor(3.8396, grad_fn=<DivBackward1>)\n",
            "tensor(4.3074, grad_fn=<DivBackward1>)\n",
            "tensor(2.2883, grad_fn=<DivBackward1>)\n",
            "tensor(0.1131, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.6835e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3722e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.9934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(3.2782e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0384, grad_fn=<DivBackward1>)\n",
            "tensor(0.0075, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.1246e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8198e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.8876e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(0.0900, grad_fn=<DivBackward1>)\n",
            "tensor(4.0758, grad_fn=<DivBackward1>)\n",
            "tensor(4.7751, grad_fn=<DivBackward1>)\n",
            "tensor(0.0506, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0076, grad_fn=<DivBackward1>)\n",
            "tensor(0.0162, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(6.0797e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0239, grad_fn=<DivBackward1>)\n",
            "tensor(0.0437, grad_fn=<DivBackward1>)\n",
            "tensor(0.4578, grad_fn=<DivBackward1>)\n",
            "tensor(0.4406, grad_fn=<DivBackward1>)\n",
            "tensor(0.4095, grad_fn=<DivBackward1>)\n",
            "tensor(0.3693, grad_fn=<DivBackward1>)\n",
            "tensor(0.3247, grad_fn=<DivBackward1>)\n",
            "tensor(0.2799, grad_fn=<DivBackward1>)\n",
            "tensor(0.2375, grad_fn=<DivBackward1>)\n",
            "tensor(5.1054, grad_fn=<DivBackward1>)\n",
            "tensor(4.4679, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0123, grad_fn=<DivBackward1>)\n",
            "tensor(0.0072, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0106, grad_fn=<DivBackward1>)\n",
            "tensor(0.0045, grad_fn=<DivBackward1>)\n",
            "tensor(0.2559, grad_fn=<DivBackward1>)\n",
            "tensor(0.2526, grad_fn=<DivBackward1>)\n",
            "tensor(0.2437, grad_fn=<DivBackward1>)\n",
            "tensor(0.2306, grad_fn=<DivBackward1>)\n",
            "tensor(0.2145, grad_fn=<DivBackward1>)\n",
            "tensor(3.4799, grad_fn=<DivBackward1>)\n",
            "tensor(4.9643, grad_fn=<DivBackward1>)\n",
            "tensor(2.8774, grad_fn=<DivBackward1>)\n",
            "tensor(0.0147, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(7.1642e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.0252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.2464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.9339e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0043, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.9509e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.4451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.4980e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.2040e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.2383e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.3020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.3259e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7762e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.4373e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.8146e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.6542e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0153, grad_fn=<DivBackward1>)\n",
            "tensor(2.4212, grad_fn=<DivBackward1>)\n",
            "tensor(2.6704, grad_fn=<DivBackward1>)\n",
            "tensor(2.8969, grad_fn=<DivBackward1>)\n",
            "tensor(0.0178, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0131, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.8265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.6674e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.4782e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.0290, grad_fn=<DivBackward1>)\n",
            "tensor(0.0848, grad_fn=<DivBackward1>)\n",
            "tensor(3.8785, grad_fn=<DivBackward1>)\n",
            "tensor(0.0098, grad_fn=<DivBackward1>)\n",
            "tensor(0.0150, grad_fn=<DivBackward1>)\n",
            "tensor(0.0074, grad_fn=<DivBackward1>)\n",
            "tensor(2.3961e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.1536e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.6793e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.1046e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.9522e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(6.7589e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.3153e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.1059e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.4822e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(9.7751e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.0333e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.4729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.4107e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.2264e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.3907e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1444e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3484e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.9139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2278e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.4557e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.4941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(5.1020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.9403e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.1989e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.1297e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.5378e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.4106e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.2318e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3681e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(0.1053, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.1973, grad_fn=<DivBackward1>)\n",
            "tensor(0.0063, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0133, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0035, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.7032e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.8700e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.5218e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2888e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(0.7616, grad_fn=<DivBackward1>)\n",
            "tensor(0.7516, grad_fn=<DivBackward1>)\n",
            "tensor(0.0146, grad_fn=<DivBackward1>)\n",
            "tensor(1.1362, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.5830e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7563e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3947e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.5565e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.7550e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.7949e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0086, grad_fn=<DivBackward1>)\n",
            "tensor(0.0047, grad_fn=<DivBackward1>)\n",
            "tensor(0.3240, grad_fn=<DivBackward1>)\n",
            "tensor(1.0735, grad_fn=<DivBackward1>)\n",
            "tensor(1.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.8950, grad_fn=<DivBackward1>)\n",
            "tensor(0.7690, grad_fn=<DivBackward1>)\n",
            "tensor(0.6394, grad_fn=<DivBackward1>)\n",
            "tensor(1.4361, grad_fn=<DivBackward1>)\n",
            "tensor(4.6989, grad_fn=<DivBackward1>)\n",
            "tensor(3.0335, grad_fn=<DivBackward1>)\n",
            "tensor(0.0306, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(9.0357e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(3.2663e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0623e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(4.4464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.2093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.3047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(8.3085e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.0173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.3497e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.2557e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.1893e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.9364e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.0756e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.0623e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.2411e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.0173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(7.6529e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(1.0848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6106e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.1722e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.8000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.5431e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.0173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.6451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.8876e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.6093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.1471e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.8943e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.1842e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.1378e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.2000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.6757e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.2941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.0914e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.9416e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.0237e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(9.0953e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0765, grad_fn=<DivBackward1>)\n",
            "tensor(0.0043, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0052, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.0252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0053, grad_fn=<DivBackward1>)\n",
            "tensor(0.0249, grad_fn=<DivBackward1>)\n",
            "tensor(0.0359, grad_fn=<DivBackward1>)\n",
            "tensor(0.0216, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.3246e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.4199e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.5682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.9577e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.2132e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0060, grad_fn=<DivBackward1>)\n",
            "tensor(0.0128, grad_fn=<DivBackward1>)\n",
            "tensor(0.2614, grad_fn=<DivBackward1>)\n",
            "tensor(0.0104, grad_fn=<DivBackward1>)\n",
            "tensor(2.1537, grad_fn=<DivBackward1>)\n",
            "tensor(0.0164, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.7113e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.0279e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.8939e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(9.8224e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.9100e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0044, grad_fn=<DivBackward1>)\n",
            "tensor(7.1525e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.1260e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(3.3974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0055, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0139, grad_fn=<DivBackward1>)\n",
            "tensor(0.0041, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.4000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.6316e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.2715e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.7019e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(7.7486e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.0237e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.4067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.6464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.7022e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.9669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(7.8913e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.8861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(8.7496e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.3085e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2398e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0049, grad_fn=<DivBackward1>)\n",
            "tensor(0.0701, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1563e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.2570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3484e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.7947e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0199, grad_fn=<DivBackward1>)\n",
            "tensor(0.0130, grad_fn=<DivBackward1>)\n",
            "tensor(0.0205, grad_fn=<DivBackward1>)\n",
            "tensor(0.0120, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0039, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.7431e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.6768e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.4928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(6.0797e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.8079e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.8133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(5.0663e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.5047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.3828e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0037, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0527, grad_fn=<DivBackward1>)\n",
            "tensor(0.0080, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(7.2000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.6610e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.7486e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.6967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.8371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.8211e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.8172e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.5830e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.8462e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1086e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.5537e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.4175e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7683e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.8781e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1444e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.3417e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.6808e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.7245e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.8477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.4107e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0150, grad_fn=<DivBackward1>)\n",
            "tensor(2.9617, grad_fn=<DivBackward1>)\n",
            "tensor(0.0077, grad_fn=<DivBackward1>)\n",
            "tensor(0.0120, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(7.1523e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(1.9312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0035, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0148, grad_fn=<DivBackward1>)\n",
            "tensor(0.0114, grad_fn=<DivBackward1>)\n",
            "tensor(0.0185, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(2.0504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0039, grad_fn=<DivBackward1>)\n",
            "tensor(0.0040, grad_fn=<DivBackward1>)\n",
            "tensor(0.2828, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0061, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.0333e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.4199e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.0450e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.9789e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.0636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(9.7628e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.1178e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.8875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.5616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.7220e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.1020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0076, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0042, grad_fn=<DivBackward1>)\n",
            "tensor(0.0139, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.5285e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.1791e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.0848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2530e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.5830e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6808e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.7696e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4663e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.1961e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.2795e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(6.7828e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(1.9431e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.1352e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.9193e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.1948e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.8994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.0265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.2080e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.8214e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.3694e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.7022e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.0027e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.3616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.7113e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0064, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0035, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.7151e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.4583e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9193e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.8411e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0096, grad_fn=<DivBackward1>)\n",
            "tensor(2.8960, grad_fn=<DivBackward1>)\n",
            "tensor(0.0344, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(8.7022e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3365e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6703e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(5.0543e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0071, grad_fn=<DivBackward1>)\n",
            "tensor(2.4232, grad_fn=<DivBackward1>)\n",
            "tensor(0.1224, grad_fn=<DivBackward1>)\n",
            "tensor(0.0210, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0036, grad_fn=<DivBackward1>)\n",
            "tensor(0.0095, grad_fn=<DivBackward1>)\n",
            "tensor(5.4239e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.8954e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.1722e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.9960e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.4477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.6835e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.1178e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.0569e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4424e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.0424e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.4332e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.1059e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8146e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.4901e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.1948e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.8543e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.2357e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.8371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.9669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.9870e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.1577e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.4570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.0662, grad_fn=<DivBackward1>)\n",
            "tensor(2.5105, grad_fn=<DivBackward1>)\n",
            "tensor(0.0800, grad_fn=<DivBackward1>)\n",
            "tensor(0.0089, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(5.9722e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(5.5431e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(3.4093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.5565e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.8596e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.8211e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.2452e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.6492e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.6212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0099, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0050, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(3.9934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.8146e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.7245e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8794e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4424e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.0517e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.9709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0037, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0057, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0094, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.9312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.2769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4782e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0043, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(4.6133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.4604, grad_fn=<DivBackward1>)\n",
            "tensor(1.1779, grad_fn=<DivBackward1>)\n",
            "tensor(2.4129, grad_fn=<DivBackward1>)\n",
            "tensor(0.0865, grad_fn=<DivBackward1>)\n",
            "tensor(0.0117, grad_fn=<DivBackward1>)\n",
            "tensor(0.0125, grad_fn=<DivBackward1>)\n",
            "tensor(0.0167, grad_fn=<DivBackward1>)\n",
            "tensor(0.0106, grad_fn=<DivBackward1>)\n",
            "tensor(4.8876e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0091, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(7.6294e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.0769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.9284e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.0173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.5060e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8835e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.7493, grad_fn=<DivBackward1>)\n",
            "tensor(0.8057, grad_fn=<DivBackward1>)\n",
            "tensor(0.1020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0255, grad_fn=<DivBackward1>)\n",
            "tensor(0.0544, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.6504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.9087e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5643e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.1791e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.9669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.2570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0045, grad_fn=<DivBackward1>)\n",
            "tensor(0.0056, grad_fn=<DivBackward1>)\n",
            "tensor(0.3258, grad_fn=<DivBackward1>)\n",
            "tensor(0.1882, grad_fn=<DivBackward1>)\n",
            "tensor(0.2839, grad_fn=<DivBackward1>)\n",
            "tensor(0.0406, grad_fn=<DivBackward1>)\n",
            "tensor(0.0158, grad_fn=<DivBackward1>)\n",
            "tensor(0.0055, grad_fn=<DivBackward1>)\n",
            "tensor(0.0039, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.4212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.0769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.5391e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.8649e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.6967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.2438e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.6555e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.8876e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.4795e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(7.5102e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.0056, grad_fn=<DivBackward1>)\n",
            "tensor(0.0078, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0079, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0097, grad_fn=<DivBackward1>)\n",
            "tensor(0.0427, grad_fn=<DivBackward1>)\n",
            "tensor(0.1061, grad_fn=<DivBackward1>)\n",
            "tensor(0.1042, grad_fn=<DivBackward1>)\n",
            "tensor(0.1010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0969, grad_fn=<DivBackward1>)\n",
            "tensor(0.0922, grad_fn=<DivBackward1>)\n",
            "tensor(0.0870, grad_fn=<DivBackward1>)\n",
            "tensor(0.0815, grad_fn=<DivBackward1>)\n",
            "tensor(0.0761, grad_fn=<DivBackward1>)\n",
            "tensor(0.0707, grad_fn=<DivBackward1>)\n",
            "tensor(0.0655, grad_fn=<DivBackward1>)\n",
            "tensor(0.0606, grad_fn=<DivBackward1>)\n",
            "tensor(0.0560, grad_fn=<DivBackward1>)\n",
            "tensor(5.7198, grad_fn=<DivBackward1>)\n",
            "tensor(5.4603, grad_fn=<DivBackward1>)\n",
            "tensor(0.2084, grad_fn=<DivBackward1>)\n",
            "tensor(0.1691, grad_fn=<DivBackward1>)\n",
            "tensor(0.0412, grad_fn=<DivBackward1>)\n",
            "tensor(0.0240, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0167, grad_fn=<DivBackward1>)\n",
            "tensor(9.1191e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0057, grad_fn=<DivBackward1>)\n",
            "tensor(0.0092, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.5736e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.9641e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.0861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.1668e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.8214e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.4175e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.0212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0092, grad_fn=<DivBackward1>)\n",
            "tensor(2.0861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(6.9377e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.1007e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.6848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.5616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.6278e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.5272e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(7.7486e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.5630e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(2.1219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0192, grad_fn=<DivBackward1>)\n",
            "tensor(0.0059, grad_fn=<DivBackward1>)\n",
            "tensor(0.0035, grad_fn=<DivBackward1>)\n",
            "tensor(1.7047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0049, grad_fn=<DivBackward1>)\n",
            "tensor(0.0061, grad_fn=<DivBackward1>)\n",
            "tensor(0.0122, grad_fn=<DivBackward1>)\n",
            "tensor(0.0131, grad_fn=<DivBackward1>)\n",
            "tensor(0.0460, grad_fn=<DivBackward1>)\n",
            "tensor(2.6141, grad_fn=<DivBackward1>)\n",
            "tensor(0.0210, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(0.0059, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(7.7125e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9722e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5524e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(5.7457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1682e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.6661e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.2953e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.9999e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3246e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.0333e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.9339e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.3337e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.8662e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.2875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.4741e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0039, grad_fn=<DivBackward1>)\n",
            "tensor(6.3894e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.7179e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3246e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.0808e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.6757e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.3259e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.1272e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1444e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0084, grad_fn=<DivBackward1>)\n",
            "tensor(7.4265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.4638e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.8729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(4.4107e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.8412e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.2053e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.7643e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.5830e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.0821e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.2769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.4067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.3630e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(4.9351e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0490e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.6861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8954e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.6729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.7351e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.0994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.5814e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5005e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.8967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0059, grad_fn=<DivBackward1>)\n",
            "tensor(0.0137, grad_fn=<DivBackward1>)\n",
            "tensor(0.0286, grad_fn=<DivBackward1>)\n",
            "tensor(0.0072, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.1989e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.5974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.0265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7524e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.4809e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0121, grad_fn=<DivBackward1>)\n",
            "tensor(0.0146, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(8.9761e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.0067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.3126e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3947e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.5192e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.8027e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.2212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5405e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2173e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.3192e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(5.5312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9007e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.4252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.8398e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6808e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.2159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.1457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1761e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.4438e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(5.0186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0091, grad_fn=<DivBackward1>)\n",
            "tensor(4.0708, grad_fn=<DivBackward1>)\n",
            "tensor(0.0105, grad_fn=<DivBackward1>)\n",
            "tensor(0.0519, grad_fn=<DivBackward1>)\n",
            "tensor(8.3085e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(8.9522e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1100e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8678e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.6757e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(3.8742e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.0994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.7138e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.2928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.3417e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0088, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.5102e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.3047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0048, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0166, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.4226e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7404e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.2728e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.7949e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.9390e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.9165e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.8279e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.6492e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.7138e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.6610e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4901e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.9522e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.7245e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.8214e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.8120e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.6623e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1696e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8198e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0243, grad_fn=<DivBackward1>)\n",
            "tensor(0.0096, grad_fn=<DivBackward1>)\n",
            "tensor(0.0236, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0095, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.8848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.3590e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.2769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.6757e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2530e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.0599e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.1709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0073, grad_fn=<DivBackward1>)\n",
            "tensor(0.0021, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(6.4371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.9045e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.6120e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.1791e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.9669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.6954e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.1590e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8678e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3736e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.9509e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(0.0883, grad_fn=<DivBackward1>)\n",
            "tensor(0.0268, grad_fn=<DivBackward1>)\n",
            "tensor(0.0127, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(3.8981e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.0797e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3497e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.5139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.3909e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0074, grad_fn=<DivBackward1>)\n",
            "tensor(0.0321, grad_fn=<DivBackward1>)\n",
            "tensor(0.0137, grad_fn=<DivBackward1>)\n",
            "tensor(0.0423, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(3.8265e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.0504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.2238e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5020e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.5565e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.0531e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(0.0312, grad_fn=<DivBackward1>)\n",
            "tensor(0.1715, grad_fn=<DivBackward1>)\n",
            "tensor(0.5471, grad_fn=<DivBackward1>)\n",
            "tensor(0.5089, grad_fn=<DivBackward1>)\n",
            "tensor(0.4518, grad_fn=<DivBackward1>)\n",
            "tensor(0.3857, grad_fn=<DivBackward1>)\n",
            "tensor(0.3188, grad_fn=<DivBackward1>)\n",
            "tensor(3.8293, grad_fn=<DivBackward1>)\n",
            "tensor(4.2871, grad_fn=<DivBackward1>)\n",
            "tensor(2.2482, grad_fn=<DivBackward1>)\n",
            "tensor(0.0809, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.4332e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3126e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(2.6345e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0421, grad_fn=<DivBackward1>)\n",
            "tensor(0.0074, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.2318e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.6397e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.0357e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(9.8581e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0621, grad_fn=<DivBackward1>)\n",
            "tensor(4.0295, grad_fn=<DivBackward1>)\n",
            "tensor(4.6788, grad_fn=<DivBackward1>)\n",
            "tensor(0.0494, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.0158, grad_fn=<DivBackward1>)\n",
            "tensor(0.0018, grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0102, grad_fn=<DivBackward1>)\n",
            "tensor(0.0313, grad_fn=<DivBackward1>)\n",
            "tensor(0.4621, grad_fn=<DivBackward1>)\n",
            "tensor(0.4452, grad_fn=<DivBackward1>)\n",
            "tensor(0.4143, grad_fn=<DivBackward1>)\n",
            "tensor(0.3742, grad_fn=<DivBackward1>)\n",
            "tensor(0.3296, grad_fn=<DivBackward1>)\n",
            "tensor(0.2845, grad_fn=<DivBackward1>)\n",
            "tensor(0.2418, grad_fn=<DivBackward1>)\n",
            "tensor(5.0572, grad_fn=<DivBackward1>)\n",
            "tensor(4.4311, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0114, grad_fn=<DivBackward1>)\n",
            "tensor(0.0070, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.6955e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0086, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(0.2577, grad_fn=<DivBackward1>)\n",
            "tensor(0.2542, grad_fn=<DivBackward1>)\n",
            "tensor(0.2454, grad_fn=<DivBackward1>)\n",
            "tensor(0.2323, grad_fn=<DivBackward1>)\n",
            "tensor(0.2162, grad_fn=<DivBackward1>)\n",
            "tensor(3.4677, grad_fn=<DivBackward1>)\n",
            "tensor(4.9268, grad_fn=<DivBackward1>)\n",
            "tensor(2.8793, grad_fn=<DivBackward1>)\n",
            "tensor(0.0147, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.8662e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.0080e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.2199e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.4767e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.0093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.5630e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8913e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.4638e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.2119e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.1722e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8358e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.1629e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.8954e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.0531e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.9563e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.2953e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0142, grad_fn=<DivBackward1>)\n",
            "tensor(2.3549, grad_fn=<DivBackward1>)\n",
            "tensor(2.6700, grad_fn=<DivBackward1>)\n",
            "tensor(2.8884, grad_fn=<DivBackward1>)\n",
            "tensor(0.0147, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0106, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.7193e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.5205e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.8700e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.2755e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0259, grad_fn=<DivBackward1>)\n",
            "tensor(0.0752, grad_fn=<DivBackward1>)\n",
            "tensor(3.7369, grad_fn=<DivBackward1>)\n",
            "tensor(0.0108, grad_fn=<DivBackward1>)\n",
            "tensor(0.0122, grad_fn=<DivBackward1>)\n",
            "tensor(0.0061, grad_fn=<DivBackward1>)\n",
            "tensor(2.4795e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3365e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.0080e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.6159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.6278e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.8939e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.9802e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(6.0795e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(9.1549e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.6848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.6517e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.1791e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.3736e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.6757e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.2398e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.4199e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.5775e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.9509e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.9033e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.6040e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.2053e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.0437e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.7751e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.8954e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.6001e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.8716e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.8675e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.0490e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.8146e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.5565e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.0848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7563e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.4026e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.3669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0942, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.1866, grad_fn=<DivBackward1>)\n",
            "tensor(0.0047, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0116, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0036, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.6053e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(6.4133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4067e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.9509e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5736e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(0.6704, grad_fn=<DivBackward1>)\n",
            "tensor(0.6865, grad_fn=<DivBackward1>)\n",
            "tensor(0.0225, grad_fn=<DivBackward1>)\n",
            "tensor(1.1400, grad_fn=<DivBackward1>)\n",
            "tensor(0.0070, grad_fn=<DivBackward1>)\n",
            "tensor(1.6451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.6555e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.9407e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.6252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.7486e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.0994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0077, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.2981, grad_fn=<DivBackward1>)\n",
            "tensor(1.0064, grad_fn=<DivBackward1>)\n",
            "tensor(0.9431, grad_fn=<DivBackward1>)\n",
            "tensor(0.8467, grad_fn=<DivBackward1>)\n",
            "tensor(0.7324, grad_fn=<DivBackward1>)\n",
            "tensor(0.6138, grad_fn=<DivBackward1>)\n",
            "tensor(1.4622, grad_fn=<DivBackward1>)\n",
            "tensor(4.6245, grad_fn=<DivBackward1>)\n",
            "tensor(3.0096, grad_fn=<DivBackward1>)\n",
            "tensor(0.0205, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.8755e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(1.6093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(3.5166e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.4345e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.2159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.7895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.5060e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.7022e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.0159e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.9457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.0821e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.0543e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(9.2741e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.6212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.7762e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(5.9841e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(7.6294e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.1934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.7431e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.8700e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.0599e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.4928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.0875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.8820e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.8239e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.8876e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.1828e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.0333e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.4345e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.7312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.8994e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.7285e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.2251e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.8053e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.3643e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.6570e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.3549e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.3430e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(5.4596e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0900, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(0.0040, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(9.0599e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0066, grad_fn=<DivBackward1>)\n",
            "tensor(0.0080, grad_fn=<DivBackward1>)\n",
            "tensor(0.0377, grad_fn=<DivBackward1>)\n",
            "tensor(0.0198, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.5868e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.4543e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.8292e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.1232e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(5.2452e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.0927e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0045, grad_fn=<DivBackward1>)\n",
            "tensor(0.0114, grad_fn=<DivBackward1>)\n",
            "tensor(0.2255, grad_fn=<DivBackward1>)\n",
            "tensor(0.0083, grad_fn=<DivBackward1>)\n",
            "tensor(2.1179, grad_fn=<DivBackward1>)\n",
            "tensor(0.0137, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.1338e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.6742e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.9206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.4571e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.5946e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(7.1761e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.2808e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.7934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(0.0035, grad_fn=<DivBackward1>)\n",
            "tensor(5.9604e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3603e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0130, grad_fn=<DivBackward1>)\n",
            "tensor(0.0040, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.3140e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.3907e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.6436e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.4013e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5005e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(5.7220e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.8981e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.9165e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.1802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.6345e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.6294e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.4424e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(7.4980e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.8014e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.9403e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(5.9126e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(9.0953e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.5325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0055, grad_fn=<DivBackward1>)\n",
            "tensor(0.0648, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.8636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0252e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.3272e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.7404e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.9484e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.1723e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0180, grad_fn=<DivBackward1>)\n",
            "tensor(0.0091, grad_fn=<DivBackward1>)\n",
            "tensor(0.0150, grad_fn=<DivBackward1>)\n",
            "tensor(0.0125, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.6464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.4239e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.0093e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.3616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(6.7949e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.2689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.7060e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0025, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.7312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.0875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0490e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0440, grad_fn=<DivBackward1>)\n",
            "tensor(0.0058, grad_fn=<DivBackward1>)\n",
            "tensor(0.0029, grad_fn=<DivBackward1>)\n",
            "tensor(0.0041, grad_fn=<DivBackward1>)\n",
            "tensor(4.5895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.7669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(5.4836e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.2040e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.9073e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.4226e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2530e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.8211e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.5895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(3.3140e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.5565e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.9870e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.1974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.4373e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.5047e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.3987e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.0437e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.0599e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.4477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(7.9390e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(7.7960e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.3379e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.4596e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.6332e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.8147e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0140, grad_fn=<DivBackward1>)\n",
            "tensor(3.0159, grad_fn=<DivBackward1>)\n",
            "tensor(0.0073, grad_fn=<DivBackward1>)\n",
            "tensor(0.0089, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(6.6874e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.0490e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(1.9193e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0040, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0109, grad_fn=<DivBackward1>)\n",
            "tensor(0.0105, grad_fn=<DivBackward1>)\n",
            "tensor(0.0193, grad_fn=<DivBackward1>)\n",
            "tensor(0.0055, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(1.8239e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0046, grad_fn=<DivBackward1>)\n",
            "tensor(0.2603, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0044, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.4052e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.8104e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.6674e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8596e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.0623e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.0530e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.6559e-06, grad_fn=<DivBackward1>)\n",
            "tensor(5.0068e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.0953e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.4464e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.2755e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.4107e-06, grad_fn=<DivBackward1>)\n",
            "tensor(8.2254e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.8848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.9457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.0967e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0059, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0038, grad_fn=<DivBackward1>)\n",
            "tensor(0.0109, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0019, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(6.3181e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.0861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(6.1989e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.2254e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.5139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.0599e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.9219e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.0266e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(3.7312e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.5537e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1260e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(7.1165e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(6.7709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(1.8000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.2025e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.8596e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.1815e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.8517e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.5378e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.9669e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(3.6358e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.1525e-06, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.2025e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.9141e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.1062e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.4186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.2557e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(7.3549e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0055, grad_fn=<DivBackward1>)\n",
            "tensor(0.0036, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0032, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.4861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.8504e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.1934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9960e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0057, grad_fn=<DivBackward1>)\n",
            "tensor(2.8568, grad_fn=<DivBackward1>)\n",
            "tensor(0.0311, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(1.1206e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.8000e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2888e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(4.7683e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0039, grad_fn=<DivBackward1>)\n",
            "tensor(0.0024, grad_fn=<DivBackward1>)\n",
            "tensor(0.0076, grad_fn=<DivBackward1>)\n",
            "tensor(2.4092, grad_fn=<DivBackward1>)\n",
            "tensor(0.1239, grad_fn=<DivBackward1>)\n",
            "tensor(0.0165, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0036, grad_fn=<DivBackward1>)\n",
            "tensor(0.0087, grad_fn=<DivBackward1>)\n",
            "tensor(4.9590e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0026, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(9.8581e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.8610e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.8835e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.3140e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.8835e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.1458e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.3417e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.9709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.4332e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(7.4026e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.6423e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.6212e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.3217e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(5.6980e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(3.0279e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(9.3933e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0017, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.7895e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(8.2132e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.4809e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3113e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.3590e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1325e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(9.3813e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4663e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.9921e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.4371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(5.6028e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.0650e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.5153e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.8358e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.4373e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.5153e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.0806, grad_fn=<DivBackward1>)\n",
            "tensor(2.4933, grad_fn=<DivBackward1>)\n",
            "tensor(0.0764, grad_fn=<DivBackward1>)\n",
            "tensor(0.0087, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(4.7683e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(4.2915e-06, grad_fn=<DivBackward1>)\n",
            "tensor(7.9509e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.6423e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(5.4716e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(2.6106e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7684e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.8371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(1.5616e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.5337e-05, grad_fn=<DivBackward1>)\n",
            "tensor(5.1260e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.4543e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0081, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0049, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(2.8371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.5763e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.0398e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(6.9139e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.2782e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.6729e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.4861e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.3232e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(3.2544e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(4.7802e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0031, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0054, grad_fn=<DivBackward1>)\n",
            "tensor(0.0013, grad_fn=<DivBackward1>)\n",
            "tensor(0.0077, grad_fn=<DivBackward1>)\n",
            "tensor(0.0010, grad_fn=<DivBackward1>)\n",
            "tensor(2.7418e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(1.5497e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(1.3709e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.1457e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.5974e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(1.6689e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0036, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.3947e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(5.4477e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(0.4432, grad_fn=<DivBackward1>)\n",
            "tensor(1.1745, grad_fn=<DivBackward1>)\n",
            "tensor(2.4154, grad_fn=<DivBackward1>)\n",
            "tensor(0.0786, grad_fn=<DivBackward1>)\n",
            "tensor(0.0075, grad_fn=<DivBackward1>)\n",
            "tensor(0.0205, grad_fn=<DivBackward1>)\n",
            "tensor(0.0135, grad_fn=<DivBackward1>)\n",
            "tensor(0.0090, grad_fn=<DivBackward1>)\n",
            "tensor(5.2452e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0022, grad_fn=<DivBackward1>)\n",
            "tensor(0.0016, grad_fn=<DivBackward1>)\n",
            "tensor(0.0082, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(0.0033, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0028, grad_fn=<DivBackward1>)\n",
            "tensor(6.7949e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.7881e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0020, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(7.1526e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.6001e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.0729e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.2941e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(3.3855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.0769e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.7762e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.7281, grad_fn=<DivBackward1>)\n",
            "tensor(0.8412, grad_fn=<DivBackward1>)\n",
            "tensor(0.0917, grad_fn=<DivBackward1>)\n",
            "tensor(0.0236, grad_fn=<DivBackward1>)\n",
            "tensor(0.0565, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(4.8875e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.4107e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(2.8133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.9934e-05, grad_fn=<DivBackward1>)\n",
            "tensor(8.4638e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.2650e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.6451e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.4305e-06, grad_fn=<DivBackward1>)\n",
            "tensor(4.6133e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.2636e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(0.0037, grad_fn=<DivBackward1>)\n",
            "tensor(0.0052, grad_fn=<DivBackward1>)\n",
            "tensor(0.3403, grad_fn=<DivBackward1>)\n",
            "tensor(0.1941, grad_fn=<DivBackward1>)\n",
            "tensor(0.2417, grad_fn=<DivBackward1>)\n",
            "tensor(0.0351, grad_fn=<DivBackward1>)\n",
            "tensor(0.0141, grad_fn=<DivBackward1>)\n",
            "tensor(0.0048, grad_fn=<DivBackward1>)\n",
            "tensor(0.0037, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(5.9605e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(2.8848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(3.5405e-05, grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.4795e-05, grad_fn=<DivBackward1>)\n",
            "tensor(4.7683e-05, grad_fn=<DivBackward1>)\n",
            "tensor(2.6226e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(8.9641e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(9.5367e-07, grad_fn=<DivBackward1>)\n",
            "tensor(4.6848e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.9908e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(8.8688e-05, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(4.5299e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(2.5749e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(5.7220e-06, grad_fn=<DivBackward1>)\n",
            "tensor(2.5034e-06, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0015, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0040, grad_fn=<DivBackward1>)\n",
            "tensor(0.0061, grad_fn=<DivBackward1>)\n",
            "tensor(0.0034, grad_fn=<DivBackward1>)\n",
            "tensor(0.0072, grad_fn=<DivBackward1>)\n",
            "tensor(0.0009, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0044, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0095, grad_fn=<DivBackward1>)\n",
            "tensor(0.0339, grad_fn=<DivBackward1>)\n",
            "tensor(0.1067, grad_fn=<DivBackward1>)\n",
            "tensor(0.1049, grad_fn=<DivBackward1>)\n",
            "tensor(0.1018, grad_fn=<DivBackward1>)\n",
            "tensor(0.0978, grad_fn=<DivBackward1>)\n",
            "tensor(0.0931, grad_fn=<DivBackward1>)\n",
            "tensor(0.0879, grad_fn=<DivBackward1>)\n",
            "tensor(0.0825, grad_fn=<DivBackward1>)\n",
            "tensor(0.0771, grad_fn=<DivBackward1>)\n",
            "tensor(0.0717, grad_fn=<DivBackward1>)\n",
            "tensor(0.0665, grad_fn=<DivBackward1>)\n",
            "tensor(0.0615, grad_fn=<DivBackward1>)\n",
            "tensor(0.0569, grad_fn=<DivBackward1>)\n",
            "tensor(5.6830, grad_fn=<DivBackward1>)\n",
            "tensor(5.4255, grad_fn=<DivBackward1>)\n",
            "tensor(0.1922, grad_fn=<DivBackward1>)\n",
            "tensor(0.1470, grad_fn=<DivBackward1>)\n",
            "tensor(0.0370, grad_fn=<DivBackward1>)\n",
            "tensor(0.0238, grad_fn=<DivBackward1>)\n",
            "tensor(0.0027, grad_fn=<DivBackward1>)\n",
            "tensor(0.0023, grad_fn=<DivBackward1>)\n",
            "tensor(0.0150, grad_fn=<DivBackward1>)\n",
            "tensor(7.7841e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(0.0057, grad_fn=<DivBackward1>)\n",
            "tensor(0.0078, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.6928e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(8.2847e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0005, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(-0., grad_fn=<DivBackward1>)\n",
            "tensor(2.3246e-05, grad_fn=<DivBackward1>)\n",
            "tensor(7.8794e-05, grad_fn=<DivBackward1>)\n",
            "tensor(1.1921e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0006, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.8943e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(9.1791e-06, grad_fn=<DivBackward1>)\n",
            "tensor(6.4371e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0001, grad_fn=<DivBackward1>)\n",
            "tensor(0.0008, grad_fn=<DivBackward1>)\n",
            "tensor(0.0030, grad_fn=<DivBackward1>)\n",
            "tensor(0.0083, grad_fn=<DivBackward1>)\n",
            "tensor(1.5855e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0014, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n",
            "tensor(8.3446e-07, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(6.8305e-05, grad_fn=<DivBackward1>)\n",
            "tensor(6.0797e-06, grad_fn=<DivBackward1>)\n",
            "tensor(0.0003, grad_fn=<DivBackward1>)\n",
            "tensor(0.0004, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0007, grad_fn=<DivBackward1>)\n",
            "tensor(0.0012, grad_fn=<DivBackward1>)\n",
            "tensor(2.3842e-07, grad_fn=<DivBackward1>)\n",
            "tensor(3.2186e-05, grad_fn=<DivBackward1>)\n",
            "tensor(3.7193e-05, grad_fn=<DivBackward1>)\n",
            "tensor(0.0002, grad_fn=<DivBackward1>)\n",
            "tensor(0.0011, grad_fn=<DivBackward1>)\n"
          ]
        }
      ]
    }
  ]
}