diff --git a/anp_1d_regression.ipynb b/anp_1d_regression.ipynb
index d996499..8d17289 100644
--- a/anp_1d_regression.ipynb
+++ b/anp_1d_regression.ipynb
@@ -2,11 +2,11 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 199,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T07:34:58.547419Z",
-     "start_time": "2019-11-02T07:34:55.174078Z"
+     "end_time": "2020-02-01T03:19:32.799666Z",
+     "start_time": "2020-02-01T03:19:32.787679Z"
     }
    },
    "outputs": [
@@ -16,15 +16,6 @@
      "text": [
       "Populating the interactive namespace from numpy and matplotlib\n"
      ]
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "/home/wassname/.pyenv/versions/3.7.2/envs/jup3.7.2/lib/python3.7/site-packages/IPython/core/magics/pylab.py:160: UserWarning: pylab import has clobbered these variables: ['plt']\n",
-      "`%matplotlib` prevents importing * from pylab and numpy\n",
-      "  \"\\n`%matplotlib` prevents importing * from pylab and numpy\"\n"
-     ]
     }
    ],
    "source": [
@@ -32,8 +23,8 @@
     "import numpy as np\n",
     "import matplotlib as plt\n",
     "import collections\n",
-    "\n",
     "from tqdm.auto import tqdm\n",
+    "from tensorboardX import SummaryWriter\n",
     "\n",
     "import math\n",
     "%pylab inline"
@@ -41,11 +32,11 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 200,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T07:35:00.251788Z",
-     "start_time": "2019-11-02T07:34:58.550188Z"
+     "end_time": "2020-02-01T03:19:32.814367Z",
+     "start_time": "2020-02-01T03:19:32.804007Z"
     }
    },
    "outputs": [],
@@ -58,26 +49,763 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": null,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T07:35:00.285167Z",
-     "start_time": "2019-11-02T07:35:00.258227Z"
+     "end_time": "2020-02-01T02:13:20.324359Z",
+     "start_time": "2020-02-01T02:13:20.314382Z"
     }
    },
    "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
    "source": [
-    "from src.data.gp_curves import GPCurvesReader\n",
-    "from src.models.model import LatentModel"
+    "# Helpers"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Data"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 201,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T07:35:00.321327Z",
-     "start_time": "2019-11-02T07:35:00.305666Z"
+     "end_time": "2020-02-01T03:19:32.863577Z",
+     "start_time": "2020-02-01T03:19:32.817900Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# The (A)NP takes as input a `NPRegressionDescription` namedtuple with fields:\n",
+    "#   `query`: a tuple containing ((context_x, context_y), target_x)\n",
+    "#   `target_y`: a tensor containing the ground truth for the targets to be\n",
+    "#     predicted\n",
+    "#   `num_total_points`: A vector containing a scalar that describes the total\n",
+    "#     number of datapoints used (context + target)\n",
+    "#   `num_context_points`: A vector containing a scalar that describes the number\n",
+    "#     of datapoints used as context\n",
+    "# The GPCurvesReader returns the newly sampled data in this format at each\n",
+    "# iteration\n",
+    "\n",
+    "NPRegressionDescription = collections.namedtuple(\n",
+    "    \"NPRegressionDescription\",\n",
+    "    (\"query\", \"target_y\", \"num_total_points\", \"num_context_points\"))\n",
+    "\n",
+    "\n",
+    "class GPCurvesReader(object):\n",
+    "    \"\"\"Generates curves using a Gaussian Process (GP).\n",
+    "\n",
+    "  Supports vector inputs (x) and vector outputs (y). Kernel is\n",
+    "  mean-squared exponential, using the x-value l2 coordinate distance scaled by\n",
+    "  some factor chosen randomly in a range. Outputs are independent gaussian\n",
+    "  processes.\n",
+    "      \"\"\"\n",
+    "\n",
+    "    def __init__(self,\n",
+    "               batch_size,\n",
+    "               max_num_context,\n",
+    "               x_size=1,\n",
+    "               y_size=1,\n",
+    "               l1_scale=0.6,\n",
+    "               sigma_scale=1.0,\n",
+    "               random_kernel_parameters=True,\n",
+    "               testing=False):\n",
+    "        \"\"\"Creates a regression dataset of functions sampled from a GP.\n",
+    "\n",
+    "    Args:\n",
+    "      batch_size: An integer.\n",
+    "      max_num_context: The max number of observations in the context.\n",
+    "      x_size: Integer >= 1 for length of \"x values\" vector.\n",
+    "      y_size: Integer >= 1 for length of \"y values\" vector.\n",
+    "      l1_scale: Float; typical scale for kernel distance function.\n",
+    "      sigma_scale: Float; typical scale for variance.\n",
+    "      random_kernel_parameters: If `True`, the kernel parameters (l1 and sigma) \n",
+    "          will be sampled uniformly within [0.1, l1_scale] and [0.1, sigma_scale].\n",
+    "      testing: Boolean that indicates whether we are testing. If so there are\n",
+    "          more targets for visualization.\n",
+    "        \"\"\"\n",
+    "        self._batch_size = batch_size\n",
+    "        self._max_num_context = max_num_context\n",
+    "        self._x_size = x_size\n",
+    "        self._y_size = y_size\n",
+    "        self._l1_scale = l1_scale\n",
+    "        self._sigma_scale = sigma_scale\n",
+    "        self._random_kernel_parameters = random_kernel_parameters\n",
+    "        self._testing = testing\n",
+    "\n",
+    "    def _gaussian_kernel(self, xdata, l1, sigma_f, sigma_noise=2e-2):\n",
+    "        \"\"\"Applies the Gaussian kernel to generate curve data.\n",
+    "\n",
+    "    Args:\n",
+    "      xdata: Tensor of shape [B, num_total_points, x_size] with\n",
+    "          the values of the x-axis data.\n",
+    "      l1: Tensor of shape [B, y_size, x_size], the scale\n",
+    "          parameter of the Gaussian kernel.\n",
+    "      sigma_f: Tensor of shape [B, y_size], the magnitude\n",
+    "          of the std.\n",
+    "      sigma_noise: Float, std of the noise that we add for stability.\n",
+    "\n",
+    "    Returns:\n",
+    "      The kernel, a float tensor of shape\n",
+    "      [B, y_size, num_total_points, num_total_points].\n",
+    "        \"\"\"\n",
+    "        num_total_points = xdata.shape[1]\n",
+    "\n",
+    "    # Expand and take the difference\n",
+    "        xdata1 = xdata.unsqueeze(1)  # [B, 1, num_total_points, x_size]\n",
+    "        xdata2 = xdata.unsqueeze(2)  # [B, num_total_points, 1, x_size]\n",
+    "        diff = xdata1 - xdata2  # [B, num_total_points, num_total_points, x_size]\n",
+    "\n",
+    "    # [B, y_size, num_total_points, num_total_points, x_size]\n",
+    "        norm = (diff[:, None, :, :, :] / l1[:, :, None, None, :])**2\n",
+    "\n",
+    "        norm = torch.sum(norm, -1)  # [B, data_size, num_total_points, num_total_points]\n",
+    "\n",
+    "    # [B, y_size, num_total_points, num_total_points]\n",
+    "        kernel = ((sigma_f)**2)[:, :, None, None] * torch.exp(-0.5 * norm)\n",
+    "\n",
+    "    # Add some noise to the diagonal to make the cholesky work.\n",
+    "        kernel += (sigma_noise**2) * torch.eye(num_total_points)\n",
+    "\n",
+    "        return kernel\n",
+    "\n",
+    "    def generate_curves(self):\n",
+    "        \"\"\"Builds the op delivering the data.\n",
+    "\n",
+    "    Generated functions are `float32` with x values between -2 and 2.\n",
+    "    \n",
+    "    Returns:\n",
+    "      A `CNPRegressionDescription` namedtuple.\n",
+    "        \"\"\"\n",
+    "        num_context = int(np.random.rand()*(self._max_num_context - 3) + 3)\n",
+    "    # If we are testing we want to have more targets and have them evenly\n",
+    "    # distributed in order to plot the function.\n",
+    "        if self._testing:\n",
+    "            num_target = 400\n",
+    "            num_total_points = num_target\n",
+    "            x_values = torch.arange(-2, 2, 1.0/100).unsqueeze(0).repeat(self._batch_size, 1)\n",
+    "            x_values = x_values.unsqueeze(-1)\n",
+    "    # During training the number of target points and their x-positions are\n",
+    "    # selected at random\n",
+    "        else:\n",
+    "            num_target = int(np.random.rand()*(self._max_num_context - num_context))\n",
+    "            num_total_points = num_context + num_target\n",
+    "            x_values = torch.rand((self._batch_size, num_total_points, self._x_size))*4 - 2\n",
+    "            \n",
+    "\n",
+    "    # Set kernel parameters\n",
+    "    # Either choose a set of random parameters for the mini-batch\n",
+    "        if self._random_kernel_parameters:\n",
+    "            l1 = torch.rand((self._batch_size, self._y_size, self._x_size))*(self._l1_scale - 0.1) + 0.1\n",
+    "            sigma_f = torch.rand((self._batch_size, self._y_size))*(self._sigma_scale - 0.1) + 0.1\n",
+    "            \n",
+    "    # Or use the same fixed parameters for all mini-batches\n",
+    "        else:\n",
+    "            l1 = torch.ones((self._batch_size, self._y_size, self._x_size))*self._l1_scale\n",
+    "            sigma_f = torch.ones((self._batch_size, self._y_size))*self._sigma_scale\n",
+    "\n",
+    "    # Pass the x_values through the Gaussian kernel\n",
+    "    # [batch_size, y_size, num_total_points, num_total_points]\n",
+    "        kernel = self._gaussian_kernel(x_values, l1, sigma_f)\n",
+    "\n",
+    "    # Calculate Cholesky, using double precision for better stability:\n",
+    "        cholesky = torch.cholesky(kernel)\n",
+    "\n",
+    "    # Sample a curve\n",
+    "    # [batch_size, y_size, num_total_points, 1]\n",
+    "        y_values = torch.matmul(cholesky, torch.randn((self._batch_size, self._y_size, num_total_points, 1)))\n",
+    "\n",
+    "    # [batch_size, num_total_points, y_size]\n",
+    "        y_values = y_values.squeeze(3)\n",
+    "        y_values = y_values.permute(0, 2, 1)\n",
+    "\n",
+    "        if self._testing:\n",
+    "      # Select the targets\n",
+    "            target_x = x_values\n",
+    "            target_y = y_values\n",
+    "\n",
+    "      # Select the observations\n",
+    "            idx = torch.randperm(num_target)\n",
+    "            context_x = x_values[:, idx[:num_context]]\n",
+    "            context_y = y_values[:, idx[:num_context]]\n",
+    "\n",
+    "        else:\n",
+    "      # Select the targets which will consist of the context points as well as\n",
+    "      # some new target points\n",
+    "            target_x = x_values[:, :num_target + num_context, :]\n",
+    "            target_y = y_values[:, :num_target + num_context, :]\n",
+    "\n",
+    "      # Select the observations\n",
+    "            context_x = x_values[:, :num_context, :]\n",
+    "            context_y = y_values[:, :num_context, :]\n",
+    "\n",
+    "        query = ((context_x, context_y), target_x)\n",
+    "\n",
+    "        return NPRegressionDescription(\n",
+    "        query=query,\n",
+    "        target_y=target_y,\n",
+    "        num_total_points=target_x.shape[1],\n",
+    "        num_context_points=num_context)\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Modules"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 202,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:32.895332Z",
+     "start_time": "2020-02-01T03:19:32.868161Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "\n",
+    "class NPBlockRelu2d(nn.Module):\n",
+    "    \"\"\"Block for Neural Processes.\"\"\"\n",
+    "\n",
+    "    def __init__(self, in_channels, out_channels, dropout=0, norm=False, bias=False):\n",
+    "        super().__init__()\n",
+    "        self.linear = nn.Linear(in_channels, out_channels, bias=bias)\n",
+    "        self.act = nn.ReLU()\n",
+    "        self.dropout = nn.Dropout2d(dropout)\n",
+    "        self.norm = nn.BatchNorm2d(out_channels) if norm else False\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        # x.shape is (Batch, Sequence, Channels)\n",
+    "        # We pass a linear over it which operates on the Channels\n",
+    "        x = self.act(self.linear(x))\n",
+    "\n",
+    "        # Now we want to apply batchnorm and dropout to the channels. So we put it in shape\n",
+    "        # (Batch, Channels, Sequence, None) so we can use Dropout2d\n",
+    "        x = x.permute(0, 2, 1)[:, :, :, None]\n",
+    "\n",
+    "        if self.norm:\n",
+    "            x = self.norm(x)\n",
+    "\n",
+    "        x = self.dropout(x)\n",
+    "        return x[:, :, :, 0].permute(0, 2, 1)\n",
+    "\n",
+    "\n",
+    "class BatchMLP(nn.Module):\n",
+    "    \"\"\"Apply MLP to the final axis of a 3D tensor (reusing already defined MLPs).\n",
+    "\n",
+    "    Args:\n",
+    "        input: input tensor of shape [B,n,d_in].\n",
+    "        output_sizes: An iterable containing the output sizes of the MLP as defined \n",
+    "            in `basic.Linear`.\n",
+    "    Returns:\n",
+    "        tensor of shape [B,n,d_out] where d_out=output_size\n",
+    "    \"\"\"\n",
+    "\n",
+    "    def __init__(self, input_size, output_size, num_layers=2, dropout=0, norm=False):\n",
+    "        super().__init__()\n",
+    "        self.input_size = input_size\n",
+    "        self.output_size = output_size\n",
+    "        self.num_layers = num_layers\n",
+    "        \n",
+    "        \n",
+    "        self.initial = NPBlockRelu2d(input_size, output_size, dropout=dropout, norm=norm)\n",
+    "        self.encoder = nn.Sequential(* [\n",
+    "            NPBlockRelu2d(output_size, output_size, dropout=dropout, norm=norm) for _ in range(num_layers - 2)\n",
+    "        ])\n",
+    "        self.final = nn.Linear(output_size, output_size)\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        x = self.initial(x)\n",
+    "        x = self.encoder(x)\n",
+    "        return self.final(x)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:10:39.791917Z",
+     "start_time": "2020-02-01T03:10:39.775892Z"
+    }
+   },
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 203,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:32.954104Z",
+     "start_time": "2020-02-01T03:19:32.902132Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "class AttnLinear(nn.Module):\n",
+    "\n",
+    "    def __init__(self, in_channels, out_channels):\n",
+    "        super().__init__()\n",
+    "        self.linear = nn.Linear(in_channels, out_channels, bias=False)\n",
+    "        torch.nn.init.normal_(self.linear.weight, std=in_channels**-0.5)\n",
+    "\n",
+    "    def forward(self, x):\n",
+    "        x = self.linear(x)\n",
+    "        return x\n",
+    "\n",
+    "\n",
+    "class Attention(nn.Module):\n",
+    "    def __init__(self, hidden_dim, attention_type, attention_layers=2, n_heads=8, x_dim=1, rep='mlp', dropout=0):\n",
+    "        super().__init__()\n",
+    "        self._rep = rep\n",
+    "        \n",
+    "        if self._rep == 'mlp':\n",
+    "            self.batch_mlp_k = BatchMLP(x_dim, hidden_dim, attention_layers, dropout=dropout)\n",
+    "            self.batch_mlp_q = BatchMLP(x_dim, hidden_dim, attention_layers, dropout=dropout)\n",
+    "\n",
+    "        if attention_type == \"uniform\":\n",
+    "            self._attention_func = self._uniform_attention\n",
+    "        elif attention_type == \"laplace\":\n",
+    "            self._attention_func = self._laplace_attention\n",
+    "        elif attention_type == \"dot\":\n",
+    "            self._attention_func = self._dot_attention\n",
+    "        elif attention_type == \"multihead\":\n",
+    "            self._W_k = nn.ModuleList(\n",
+    "                [AttnLinear(hidden_dim, hidden_dim) for _ in range(n_heads)])\n",
+    "            self._W_v = nn.ModuleList(\n",
+    "                [AttnLinear(hidden_dim, hidden_dim) for _ in range(n_heads)])\n",
+    "            self._W_q = nn.ModuleList(\n",
+    "                [AttnLinear(hidden_dim, hidden_dim) for _ in range(n_heads)])\n",
+    "            self._W = AttnLinear(n_heads * hidden_dim, hidden_dim)\n",
+    "            self._attention_func = self._multihead_attention\n",
+    "            self.n_heads = n_heads\n",
+    "        elif attention_type ==\"ptmultihead\":\n",
+    "            self._attention_func = torch.nn.MultiheadAttention(\n",
+    "                hidden_dim, n_heads, bias=False, dropout=dropout\n",
+    "            )\n",
+    "        else:\n",
+    "            raise NotImplementedError\n",
+    "\n",
+    "    def forward(self, k, v, q):\n",
+    "        if self._rep == 'mlp':\n",
+    "            k = self.batch_mlp_k(k)\n",
+    "            q = self.batch_mlp_q(q)\n",
+    "        rep = self._attention_func(k, v, q)\n",
+    "        return rep\n",
+    "\n",
+    "    def _uniform_attention(self, k, v, q):\n",
+    "        total_points = q.shape[1]\n",
+    "        rep = torch.mean(v, dim=1, keepdim=True)\n",
+    "        rep = rep.repeat(1, total_points, 1)\n",
+    "        return rep\n",
+    "\n",
+    "    def _laplace_attention(self, k, v, q, scale=0.5):\n",
+    "        k_ = k.unsqueeze(1)\n",
+    "        v_ = v.unsqueeze(2)\n",
+    "        unnorm_weights = torch.abs((k_ - v_) * scale)\n",
+    "        unnorm_weights = unnorm_weights.sum(dim=-1)\n",
+    "        weights = torch.softmax(unnorm_weights, dim=-1)\n",
+    "        rep = torch.einsum('bik,bkj->bij', weights, v)\n",
+    "        return rep\n",
+    "\n",
+    "    def _dot_attention(self, k, v, q):\n",
+    "        scale = q.shape[-1]**0.5\n",
+    "        unnorm_weights = torch.einsum('bjk,bik->bij', k, q) / scale\n",
+    "        weights = torch.softmax(unnorm_weights, dim=-1)\n",
+    "\n",
+    "        rep = torch.einsum('bik,bkj->bij', weights, v)\n",
+    "        return rep\n",
+    "\n",
+    "    def _multihead_attention(self, k, v, q):\n",
+    "        outs = []\n",
+    "        for i in range(self.n_heads):\n",
+    "            k_ = self._W_k[i](k)\n",
+    "            v_ = self._W_v[i](v)\n",
+    "            q_ = self._W_q[i](q)\n",
+    "            out = self._dot_attention(k_, v_, q_)\n",
+    "            outs.append(out)\n",
+    "        outs = torch.stack(outs, dim=-1)\n",
+    "        outs = outs.view(outs.shape[0], outs.shape[1], -1)\n",
+    "        rep = self._W(outs)\n",
+    "        return rep\n",
+    "    \n",
+    "    def _pytorch_multihead_attention(self, k, v, q):\n",
+    "        # Pytorch multiheaded attention takes inputs if diff order and permutation\n",
+    "        q = q.permute(1, 0, 2)\n",
+    "        k = k.permute(1, 0, 2)\n",
+    "        v = v.permute(1, 0, 2)\n",
+    "        o = self._attention_func(q, k, v)[0]\n",
+    "        return o.permute(1, 0, 2)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 204,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:32.981923Z",
+     "start_time": "2020-02-01T03:19:32.957280Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "class LatentEncoder(nn.Module):\n",
+    "    def __init__(self,\n",
+    "                 input_dim,\n",
+    "                 hidden_dim=32,\n",
+    "                 latent_dim=32,\n",
+    "                 self_attention_type=\"dot\",\n",
+    "                 n_encoder_layers=3,\n",
+    "                 min_std=0.01,\n",
+    "                 use_lvar=False,\n",
+    "                 use_self_attn=False,      \n",
+    "                 attention_layers=2,\n",
+    "                ):\n",
+    "        super().__init__()\n",
+    "        self._input_layer = nn.Linear(input_dim, hidden_dim)\n",
+    "        self._encoder = nn.ModuleList([\n",
+    "            nn.Linear(hidden_dim, hidden_dim) for _ in range(n_encoder_layers)\n",
+    "        ])\n",
+    "        if use_self_attn:\n",
+    "            self._self_attention = Attention(hidden_dim, self_attention_type, attention_layers, rep='identity')\n",
+    "        self._penultimate_layer = nn.Linear(hidden_dim, hidden_dim)\n",
+    "        self._mean = nn.Linear(hidden_dim, latent_dim)\n",
+    "        self._log_var = nn.Linear(hidden_dim, latent_dim)\n",
+    "        self._min_std = min_std\n",
+    "        self._use_lvar = use_lvar\n",
+    "        self._use_self_attn = use_self_attn\n",
+    "\n",
+    "    def forward(self, x, y):\n",
+    "        encoder_input = torch.cat([x, y], dim=-1)\n",
+    "\n",
+    "        # Pass final axis through MLP\n",
+    "        encoded = self._input_layer(encoder_input)\n",
+    "        for layer in self._encoder:\n",
+    "            encoded = torch.relu(layer(encoded))\n",
+    "\n",
+    "        # Aggregator: take the mean over all points\n",
+    "        if self._use_self_attn:\n",
+    "            attention_output = self._self_attention(encoded, encoded, encoded)\n",
+    "            mean_repr = attention_output.mean(dim=1)\n",
+    "        else:\n",
+    "            mean_repr = encoded.mean(dim=1)\n",
+    "\n",
+    "        # Have further MLP layers that map to the parameters of the Gaussian latent\n",
+    "        mean_repr = torch.relu(self._penultimate_layer(mean_repr))\n",
+    "\n",
+    "        # Then apply further linear layers to output latent mu and log sigma\n",
+    "        mean = self._mean(mean_repr)\n",
+    "        log_var = self._log_var(mean_repr)\n",
+    "\n",
+    "        # Clip it in the log domain, so it can only approach self.min_std, this helps avoid mode collapase\n",
+    "        # 2 ways, a better but untested way using the more stable log domain, and the way from the deepmind repo\n",
+    "        if self._use_lvar:\n",
+    "            log_var = log_var + math.log(self._min_std)\n",
+    "            sigma = torch.exp(0.5 * log_var)\n",
+    "        else:\n",
+    "            sigma = self._min_std + (1 - self._min_std) * torch.sigmoid(log_var * 0.5)\n",
+    "        dist = torch.distributions.Normal(mean, sigma)\n",
+    "        return dist, log_var"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 205,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:32.998565Z",
+     "start_time": "2020-02-01T03:19:32.984808Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "class DeterministicEncoder(nn.Module):\n",
+    "    def __init__(\n",
+    "            self,\n",
+    "            input_dim,\n",
+    "            x_dim,\n",
+    "            hidden_dim=32,\n",
+    "            n_d_encoder_layers=3,\n",
+    "            self_attention_type=\"dot\",\n",
+    "            cross_attention_type=\"dot\",\n",
+    "            use_self_attn=False,\n",
+    "            attention_layers=2, ):\n",
+    "        super().__init__()\n",
+    "        self._use_self_attn = use_self_attn\n",
+    "        self._input_layer = nn.Linear(input_dim, hidden_dim)\n",
+    "        self._d_encoder = nn.ModuleList([\n",
+    "            nn.Linear(hidden_dim, hidden_dim)\n",
+    "            for _ in range(n_d_encoder_layers)\n",
+    "        ])\n",
+    "        if use_self_attn:\n",
+    "            self._self_attention = Attention(\n",
+    "                hidden_dim,\n",
+    "                self_attention_type,\n",
+    "                attention_layers,\n",
+    "                rep='identity')\n",
+    "        self._cross_attention = Attention(\n",
+    "            hidden_dim,\n",
+    "            cross_attention_type,\n",
+    "            x_dim=x_dim,\n",
+    "            attention_layers=attention_layers)\n",
+    "\n",
+    "    def forward(self, context_x, context_y, target_x):\n",
+    "        # Concatenate x and y along the filter axes\n",
+    "        d_encoder_input = torch.cat([context_x, context_y], dim=-1)\n",
+    "\n",
+    "        # Pass final axis through MLP\n",
+    "        d_encoded = self._input_layer(d_encoder_input)\n",
+    "        for layer in self._d_encoder:\n",
+    "            d_encoded = torch.relu(layer(d_encoded))\n",
+    "\n",
+    "        if self._use_self_attn:\n",
+    "            d_encoded = self._self_attention(d_encoded, d_encoded, d_encoded)\n",
+    "\n",
+    "        # Apply attention\n",
+    "        h = self._cross_attention(context_x, d_encoded, target_x)\n",
+    "\n",
+    "        return h"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 206,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:33.019374Z",
+     "start_time": "2020-02-01T03:19:33.002371Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "class Decoder(nn.Module):\n",
+    "    def __init__(self,\n",
+    "                 x_dim,\n",
+    "                 y_dim,\n",
+    "                 hidden_dim=32,\n",
+    "                 latent_dim=32,\n",
+    "                 n_decoder_layers=3,\n",
+    "                 use_deterministic_path=True,\n",
+    "                 min_std=0.01,\n",
+    "                 use_lvar=False):\n",
+    "        super(Decoder, self).__init__()\n",
+    "        self._target_transform = nn.Linear(x_dim, hidden_dim)\n",
+    "        if use_deterministic_path:\n",
+    "            hidden_dim_2 = 2 * hidden_dim + latent_dim\n",
+    "        else:\n",
+    "            hidden_dim_2 = hidden_dim + latent_dim\n",
+    "        self._decoder = nn.ModuleList([\n",
+    "            nn.Linear(hidden_dim_2, hidden_dim_2)\n",
+    "            for _ in range(n_decoder_layers)\n",
+    "        ])\n",
+    "        self._mean = nn.Linear(hidden_dim_2, y_dim)\n",
+    "        self._std = nn.Linear(hidden_dim_2, y_dim)\n",
+    "        self._use_deterministic_path = use_deterministic_path\n",
+    "        self._min_std = min_std\n",
+    "        self._use_lvar = use_lvar\n",
+    "\n",
+    "    def forward(self, r, z, target_x):\n",
+    "        # concatenate target_x and representation\n",
+    "        x = self._target_transform(target_x)\n",
+    "\n",
+    "        if self._use_deterministic_path:\n",
+    "            z = torch.cat([r, z], dim=-1)\n",
+    "\n",
+    "        representation = torch.cat([z, x], dim=-1)\n",
+    "        \n",
+    "        # Pass final axis through MLP\n",
+    "        for layer in self._decoder:\n",
+    "            representation = torch.relu(layer(representation))\n",
+    "\n",
+    "        # Get the mean and the variance\n",
+    "        mean = self._mean(representation)\n",
+    "        log_sigma = self._std(representation)\n",
+    "\n",
+    "        # Bound the variance\n",
+    "        if self._use_lvar:\n",
+    "            log_sigma = log_sigma + math.log(self._min_std)\n",
+    "            sigma = torch.exp(log_sigma)\n",
+    "        else:\n",
+    "            sigma = self._min_std + (1 - self._min_std) * F.softplus(log_sigma)\n",
+    "\n",
+    "        dist = torch.distributions.Normal(mean, sigma)\n",
+    "        return dist, log_sigma"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 207,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:33.033796Z",
+     "start_time": "2020-02-01T03:19:33.021744Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "def kl_loss_var(prior_mu, log_var_prior, post_mu, log_var_post):\n",
+    "    \"\"\"\n",
+    "    Analytical KLD for two gaussians, taking in log_variance instead of scale ( given variance=scale**2) for more stable gradients\n",
+    "    \n",
+    "    For version using scale see https://github.com/pytorch/pytorch/blob/master/torch/distributions/kl.py#L398\n",
+    "    \"\"\"\n",
+    "\n",
+    "    var_ratio_log = log_var_post - log_var_prior\n",
+    "    kl_div = (\n",
+    "        (var_ratio_log.exp() +\n",
+    "         (post_mu - prior_mu)**2) / log_var_prior.exp() - 1.0 - var_ratio_log)\n",
+    "    kl_div = 0.5 * kl_div\n",
+    "    return kl_div\n",
+    "\n",
+    "def log_prob_sigma(value, loc, log_scale):\n",
+    "    \"\"\"A slightly more stable (not confirmed yet) log prob taking in log_var instead of scale.\n",
+    "    modified from https://github.com/pytorch/pytorch/blob/2431eac7c011afe42d4c22b8b3f46dedae65e7c0/torch/distributions/normal.py#L65\n",
+    "    \"\"\"\n",
+    "    var = torch.exp(log_scale * 2)\n",
+    "    return (\n",
+    "        -((value - loc) ** 2) / (2 * var) - log_scale - math.log(math.sqrt(2 * math.pi))\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 208,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:33.064975Z",
+     "start_time": "2020-02-01T03:19:33.036431Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "class LatentModel(nn.Module):\n",
+    "    def __init__(self,\n",
+    "                 x_dim,\n",
+    "                 y_dim,\n",
+    "                 hidden_dim=32,\n",
+    "                 latent_dim=32,\n",
+    "                 latent_enc_self_attn_type=\"dot\",\n",
+    "                 det_enc_self_attn_type=\"dot\",\n",
+    "                 det_enc_cross_attn_type=\"dot\",\n",
+    "                 n_latent_encoder_layers=3,\n",
+    "                 n_det_encoder_layers=3,\n",
+    "                 n_decoder_layers=3,\n",
+    "                 use_deterministic_path=True,\n",
+    "                 min_std=0.01,\n",
+    "                 use_lvar=False,\n",
+    "                 attention_layers=2,\n",
+    "                ):\n",
+    "\n",
+    "        super(LatentModel, self).__init__()\n",
+    "\n",
+    "        self._latent_encoder = LatentEncoder(\n",
+    "            x_dim + y_dim,\n",
+    "            hidden_dim=hidden_dim,\n",
+    "            latent_dim=latent_dim,\n",
+    "            self_attention_type=latent_enc_self_attn_type,\n",
+    "            n_encoder_layers=n_latent_encoder_layers,\n",
+    "            attention_layers=attention_layers)\n",
+    "\n",
+    "        self._deterministic_encoder = DeterministicEncoder(\n",
+    "            input_dim=x_dim + y_dim,\n",
+    "            x_dim=x_dim,\n",
+    "            hidden_dim=hidden_dim,\n",
+    "            self_attention_type=det_enc_self_attn_type,\n",
+    "            cross_attention_type=det_enc_cross_attn_type,\n",
+    "            n_d_encoder_layers=n_det_encoder_layers,\n",
+    "            attention_layers=attention_layers)\n",
+    "\n",
+    "        self._decoder = Decoder(\n",
+    "            x_dim,\n",
+    "            y_dim,\n",
+    "            hidden_dim=hidden_dim,\n",
+    "            latent_dim=latent_dim,\n",
+    "            min_std=min_std,\n",
+    "            use_lvar=use_lvar,\n",
+    "            n_decoder_layers=n_decoder_layers,\n",
+    "            use_deterministic_path=use_deterministic_path,\n",
+    "            \n",
+    "        )\n",
+    "        self._use_deterministic_path = use_deterministic_path\n",
+    "        self._use_lvar = use_lvar\n",
+    "\n",
+    "    def forward(self, context_x, context_y, target_x, target_y=None):\n",
+    "        num_targets = target_x.size(1)\n",
+    "\n",
+    "        dist_prior, log_var_prior = self._latent_encoder(context_x, context_y)\n",
+    "\n",
+    "        if target_y is not None:\n",
+    "            dist_post, log_var_post = self._latent_encoder(target_x,\n",
+    "                                                             target_y)\n",
+    "            z = dist_post.loc\n",
+    "        else:\n",
+    "            z = dist_prior.loc\n",
+    "\n",
+    "        z = z.unsqueeze(1).repeat(1, num_targets, 1)  # [B, T_target, H]\n",
+    "\n",
+    "        if self._use_deterministic_path:\n",
+    "            r = self._deterministic_encoder(context_x, context_y,\n",
+    "                                            target_x)  # [B, T_target, H]\n",
+    "        else:\n",
+    "            r = None\n",
+    "\n",
+    "        dist, log_sigma = self._decoder(r, z, target_x)\n",
+    "        if target_y is not None:\n",
+    "            if self._use_lvar:\n",
+    "                log_p = log_prob_sigma(target_y, dist.loc, log_sigma).mean(-1) # [B, T_target, Y].mean(-1)\n",
+    "                kl_loss = kl_loss_var(dist_prior.loc, log_var_prior,\n",
+    "                                      dist_post.loc, log_var_post).mean(-1) # [B, R].mean(-1)\n",
+    "            else:\n",
+    "                log_p = dist.log_prob(target_y).mean(-1)\n",
+    "                kl_loss = torch.distributions.kl_divergence(\n",
+    "                    dist_post, dist_prior).mean(-1)\n",
+    "            kl_loss = kl_loss[:, None].expand(log_p.shape)\n",
+    "            mse_loss = F.mse_loss(dist.loc, target_y)\n",
+    "            loss = (kl_loss - log_p).mean()\n",
+    "\n",
+    "        else:\n",
+    "            log_p = None\n",
+    "            mse_loss = None\n",
+    "            kl_loss = None\n",
+    "            loss = None\n",
+    "\n",
+    "        y_pred = dist.rsample() if self.training else dist.loc\n",
+    "        return y_pred, kl_loss, loss, mse_loss, dist.scale"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Util"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 209,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:33.080375Z",
+     "start_time": "2020-02-01T03:19:33.067520Z"
     }
    },
    "outputs": [],
@@ -99,7 +827,7 @@
     "    std: An array of shape [B,num_targets,1] that contains the\n",
     "        predicted std dev of the y values at the target points in target_x.\n",
     "      \"\"\"\n",
-    "  # Plot everything\n",
+    "    # Plot everything\n",
     "    plt.plot(target_x[0], pred_y[0], 'b', linewidth=2)\n",
     "    plt.plot(target_x[0], target_y[0], 'k:', linewidth=2)\n",
     "    plt.plot(context_x[0], context_y[0], 'ko', markersize=10)\n",
@@ -107,29 +835,91 @@
     "          target_x[0, :, 0],\n",
     "          pred_y[0, :, 0] - std[0, :, 0],\n",
     "          pred_y[0, :, 0] + std[0, :, 0],\n",
-    "          alpha=0.5,\n",
+    "          alpha=0.2,\n",
     "          facecolor='#65c9f7',\n",
     "          interpolate=True)\n",
     "\n",
-    "  # Make the plot pretty\n",
+    "    # Make the plot pretty\n",
     "    plt.yticks([-2, 0, 2], fontsize=16)\n",
     "    plt.xticks([-2, 0, 2], fontsize=16)\n",
     "    plt.ylim([-2, 2])\n",
     "    plt.grid('off')\n",
-    "    ax = plt.gca()\n",
-    "    plt.show()"
+    "    ax = plt.gca()"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 210,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T07:35:00.335641Z",
-     "start_time": "2019-11-02T07:35:00.323713Z"
+     "end_time": "2020-02-01T03:19:33.099668Z",
+     "start_time": "2020-02-01T03:19:33.083014Z"
     }
    },
    "outputs": [],
+   "source": [
+    "def test(model, dataset_test, writer=None, plot=False, global_step=None):\n",
+    "    model.eval()\n",
+    "    with torch.no_grad():\n",
+    "        data_test = dataset_test.generate_curves()\n",
+    "\n",
+    "        (context_x, context_y), target_x = data_test.query\n",
+    "        target_y = data_test.target_y\n",
+    "\n",
+    "        context_x = context_x.cuda()\n",
+    "        context_y = context_y.cuda()\n",
+    "        target_x = target_x.cuda()\n",
+    "        target_y = target_y.cuda()\n",
+    "\n",
+    "        y_pred, kl, loss, mse_loss, y_std = model(context_x, context_y, target_x, target_y)\n",
+    "\n",
+    "        if writer is not None:\n",
+    "            writer.add_scalar('val/loss', loss, global_step=global_step)\n",
+    "            writer.add_scalar('val/y_std', y_std.mean(), global_step=global_step)\n",
+    "            writer.add_scalar('val/mse_loss', mse_loss, global_step=global_step)\n",
+    "            writer.add_scalar('val/kl', kl.mean(), global_step=global_step)\n",
+    "        \n",
+    "        if plot:\n",
+    "            fig = plt.figure()\n",
+    "            plt.title(f\"Iter {n_iter}\")            \n",
+    "            plot_functions(target_x.detach().cpu().numpy(),\n",
+    "                           target_y.detach().cpu().numpy(),\n",
+    "                           context_x.detach().cpu().numpy(),\n",
+    "                           context_y.detach().cpu().numpy(),\n",
+    "                           y_pred.detach().cpu().numpy(),\n",
+    "                           y_std.detach().cpu().numpy())\n",
+    "            \n",
+    "            writer.add_figure('test', fig, global_step=global_step, close=False)\n",
+    "            plt.show()\n",
+    "            \n",
+    "    return y_pred, kl, loss, y_std"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Init"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 211,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:33.119679Z",
+     "start_time": "2020-02-01T03:19:33.101969Z"
+    }
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "runs/Feb01_11-19-33_mjcdesktop-anp_1d_reg\n"
+     ]
+    }
+   ],
    "source": [
     "\n",
     "MAX_CONTEXT_POINTS = 50 \n",
@@ -139,46 +929,48 @@
     "    batch_size=16, max_num_context=MAX_CONTEXT_POINTS, random_kernel_parameters=random_kernel_parameters)\n",
     "\n",
     "dataset_test = GPCurvesReader(\n",
-    "    batch_size=1, max_num_context=MAX_CONTEXT_POINTS, testing=True, random_kernel_parameters=random_kernel_parameters)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 6,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2019-11-02T07:35:08.171125Z",
-     "start_time": "2019-11-02T07:35:00.338388Z"
-    }
-   },
-   "outputs": [],
-   "source": [
+    "    batch_size=1, max_num_context=MAX_CONTEXT_POINTS, testing=True, random_kernel_parameters=random_kernel_parameters)\n",
     "\n",
-    "model = LatentModel(1, 1, 64, latent_enc_self_attn_type=\"multihead\", det_enc_self_attn_type=\"multihead\",\n",
-    "                   det_enc_cross_attn_type=\"multihead\").cuda()"
+    "writer = SummaryWriter(comment='-anp_1d_reg')\n",
+    "print(writer.logdir)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 212,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T07:35:08.184936Z",
-     "start_time": "2019-11-02T07:35:08.177059Z"
+     "end_time": "2020-02-01T03:19:33.175871Z",
+     "start_time": "2020-02-01T03:19:33.122446Z"
     }
    },
    "outputs": [],
    "source": [
-    "optim = torch.optim.AdamW(model.parameters(), lr=1e-5)"
+    "hparams = dict(x_dim=1, y_dim=1, hidden_dim=128, latent_dim=128, latent_enc_self_attn_type=\"multihead\", det_enc_self_attn_type=\"multihead\",\n",
+    "                   det_enc_cross_attn_type=\"multihead\")\n",
+    "model = LatentModel(**hparams).cuda()\n",
+    "\n",
+    "optim = torch.optim.Adam(model.parameters(), lr=1e-4)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T01:23:02.658135Z",
+     "start_time": "2020-02-01T01:23:02.653762Z"
+    }
+   },
+   "source": [
+    "# Train"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": null,
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2019-11-02T09:48:40.661394Z",
-     "start_time": "2019-11-02T07:35:08.187579Z"
+     "start_time": "2020-02-01T03:19:32.800Z"
     },
     "scrolled": true
    },
@@ -186,12 +978,12 @@
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "0bb37d13873f46dbbb1433a8d4d8ca8e",
+       "model_id": "6c16408b24384de3bb64ec531aabba61",
        "version_major": 2,
        "version_minor": 0
       },
       "text/plain": [
-       "HBox(children=(IntProgress(value=0, max=100000), HTML(value='')))"
+       "HBox(children=(FloatProgress(value=0.0, max=100000.0), HTML(value='')))"
       ]
      },
      "metadata": {},
@@ -201,12 +993,12 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "train: 0   12\n"
+      "train: 0 loss= 0.9087 val_loss= 2.267\n"
      ]
     },
     {
      "data": {
-      "image/png": "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\n",
+      "image/png": "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\n",
       "text/plain": [
        "<Figure size 432x288 with 1 Axes>"
       ]
@@ -215,1401 +1007,91 @@
       "needs_background": "light"
      },
      "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 100 6.391\n",
-      "train: 200 5.626\n",
-      "train: 300 6.16\n",
-      "train: 400 5.766\n",
-      "train: 500 5.624\n",
-      "train: 600 5.689\n",
-      "train: 700 5.117\n",
-      "train: 800 3.194\n",
-      "train: 900 4.035\n",
-      "train: 1000 3.541\n",
-      "train: 1100 4.288\n",
-      "train: 1200 4.116\n",
-      "train: 1300 3.648\n",
-      "train: 1400 2.625\n",
-      "train: 1500 3.166\n",
-      "train: 1600 3.01\n",
-      "train: 1700 3.108\n",
-      "train: 1800 3.864\n",
-      "train: 1900 4.774\n",
-      "train: 2000 2.515\n",
-      "train: 2100 2.655\n",
-      "train: 2200 3.437\n",
-      "train: 2300 2.293\n",
-      "train: 2400 2.701\n",
-      "train: 2500 3.267\n",
-      "train: 2600 2.44\n",
-      "train: 2700 2.663\n",
-      "train: 2800 2.313\n",
-      "train: 2900 2.96\n",
-      "train: 3000 3.034\n",
-      "train: 3100 2.297\n",
-      "train: 3200 2.276\n",
-      "train: 3300 2.389\n",
-      "train: 3400 2.222\n",
-      "train: 3500 2.562\n",
-      "train: 3600 1.77\n",
-      "train: 3700 3.312\n",
-      "train: 3800 2.641\n",
-      "train: 3900 1.478\n",
-      "train: 4000 1.908\n",
-      "train: 4100 1.738\n",
-      "train: 4200 2.144\n",
-      "train: 4300 2.049\n",
-      "train: 4400 1.978\n",
-      "train: 4500 2.141\n",
-      "train: 4600 2.234\n",
-      "train: 4700 1.66\n",
-      "train: 4800 2.124\n",
-      "train: 4900 1.431\n",
-      "train: 5000 1.551\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 5100 1.679\n",
-      "train: 5200 1.883\n",
-      "train: 5300 2.049\n",
-      "train: 5400 1.543\n",
-      "train: 5500 2.868\n",
-      "train: 5600 1.683\n",
-      "train: 5700 1.899\n",
-      "train: 5800 1.303\n",
-      "train: 5900 1.385\n",
-      "train: 6000 1.294\n",
-      "train: 6100 1.185\n",
-      "train: 6200 2.53\n",
-      "train: 6300 1.166\n",
-      "train: 6400 1.804\n",
-      "train: 6500 1.36\n",
-      "train: 6600 0.9277\n",
-      "train: 6700 1.271\n",
-      "train: 6800 1.094\n",
-      "train: 6900 1.327\n",
-      "train: 7000 1.138\n",
-      "train: 7100 1.075\n",
-      "train: 7200 1.835\n",
-      "train: 7300 1.104\n",
-      "train: 7400 1.488\n",
-      "train: 7500 1.139\n",
-      "train: 7600 0.6968\n",
-      "train: 7700 0.7172\n",
-      "train: 7800  1.5\n",
-      "train: 7900 2.028\n",
-      "train: 8000 1.62\n",
-      "train: 8100 0.5501\n",
-      "train: 8200 0.7213\n",
-      "train: 8300 0.7354\n",
-      "train: 8400 0.5944\n",
-      "train: 8500 0.6943\n",
-      "train: 8600 1.779\n",
-      "train: 8700 1.152\n",
-      "train: 8800 0.519\n",
-      "train: 8900 0.8859\n",
-      "train: 9000 0.4783\n",
-      "train: 9100 0.7806\n",
-      "train: 9200 0.6132\n",
-      "train: 9300 0.5164\n",
-      "train: 9400 0.7371\n",
-      "train: 9500 1.511\n",
-      "train: 9600 0.567\n",
-      "train: 9700 0.5146\n",
-      "train: 9800 0.6316\n",
-      "train: 9900 0.5202\n",
-      "train: 10000 1.142\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 10100 0.8833\n",
-      "train: 10200 0.5416\n",
-      "train: 10300 0.4089\n",
-      "train: 10400 0.4403\n",
-      "train: 10500 2.297\n",
-      "train: 10600 0.4174\n",
-      "train: 10700 1.498\n",
-      "train: 10800 0.6495\n",
-      "train: 10900 0.2989\n",
-      "train: 11000 0.6395\n",
-      "train: 11100 0.44\n",
-      "train: 11200 0.5389\n",
-      "train: 11300 0.2969\n",
-      "train: 11400 0.5898\n",
-      "train: 11500 0.2404\n",
-      "train: 11600 0.8724\n",
-      "train: 11700 0.6155\n",
-      "train: 11800 1.239\n",
-      "train: 11900 0.7685\n",
-      "train: 12000 0.7193\n",
-      "train: 12100 0.2667\n",
-      "train: 12200 0.4916\n",
-      "train: 12300 0.4796\n",
-      "train: 12400 0.1291\n",
-      "train: 12500 0.5619\n",
-      "train: 12600 0.06822\n",
-      "train: 12700 0.1349\n",
-      "train: 12800 0.1807\n",
-      "train: 12900 -0.01337\n",
-      "train: 13000 0.1332\n",
-      "train: 13100 0.1858\n",
-      "train: 13200 0.2523\n",
-      "train: 13300 0.6207\n",
-      "train: 13400 0.1929\n",
-      "train: 13500 0.3042\n",
-      "train: 13600 0.07761\n",
-      "train: 13700 0.1647\n",
-      "train: 13800 0.4122\n",
-      "train: 13900 0.3124\n",
-      "train: 14000 0.2281\n",
-      "train: 14100 0.6712\n",
-      "train: 14200 0.6139\n",
-      "train: 14300 0.4546\n",
-      "train: 14400 0.9218\n",
-      "train: 14500 0.01121\n",
-      "train: 14600 0.07928\n",
-      "train: 14700 -0.1427\n",
-      "train: 14800 -0.1767\n",
-      "train: 14900 0.3025\n",
-      "train: 15000 0.18\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 15100 0.06924\n",
-      "train: 15200 0.104\n",
-      "train: 15300 -0.05432\n",
-      "train: 15400 0.0878\n",
-      "train: 15500 0.3096\n",
-      "train: 15600 -0.101\n",
-      "train: 15700 -0.1815\n",
-      "train: 15800 0.01411\n",
-      "train: 15900 -0.0145\n",
-      "train: 16000 0.4367\n",
-      "train: 16100 -0.1469\n",
-      "train: 16200 0.2669\n",
-      "train: 16300 0.3939\n",
-      "train: 16400 0.001367\n",
-      "train: 16500 0.06631\n",
-      "train: 16600 0.007301\n",
-      "train: 16700 0.05867\n",
-      "train: 16800 0.259\n",
-      "train: 16900 -0.06433\n",
-      "train: 17000 -0.1931\n",
-      "train: 17100 0.3358\n",
-      "train: 17200 -0.05535\n",
-      "train: 17300 0.08028\n",
-      "train: 17400 -0.332\n",
-      "train: 17500 -0.1409\n",
-      "train: 17600 -0.07089\n",
-      "train: 17700 -0.2534\n",
-      "train: 17800 0.8266\n",
-      "train: 17900 -0.2207\n",
-      "train: 18000 0.08128\n",
-      "train: 18100 -0.1946\n",
-      "train: 18200 -0.2114\n",
-      "train: 18300 0.5667\n",
-      "train: 18400 0.03189\n",
-      "train: 18500 -0.2915\n",
-      "train: 18600 0.02664\n",
-      "train: 18700 0.1672\n",
-      "train: 18800 -0.1621\n",
-      "train: 18900 -0.2582\n",
-      "train: 19000 -0.1913\n",
-      "train: 19100 -0.1661\n",
-      "train: 19200 0.2957\n",
-      "train: 19300 -0.3063\n",
-      "train: 19400 -0.1645\n",
-      "train: 19500 0.2155\n",
-      "train: 19600 0.4337\n",
-      "train: 19700 0.7424\n",
-      "train: 19800 -0.07544\n",
-      "train: 19900 -0.02267\n",
-      "train: 20000 0.07467\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 20100 0.06119\n",
-      "train: 20200 -0.199\n",
-      "train: 20300 -0.1302\n",
-      "train: 20400 -0.2106\n",
-      "train: 20500 -0.2188\n",
-      "train: 20600 0.4964\n",
-      "train: 20700 -0.1303\n",
-      "train: 20800 1.103\n",
-      "train: 20900 0.09169\n",
-      "train: 21000 -0.3041\n",
-      "train: 21100 -0.4382\n",
-      "train: 21200 -0.181\n",
-      "train: 21300 -0.4443\n",
-      "train: 21400 -0.3059\n",
-      "train: 21500 0.05672\n",
-      "train: 21600 -0.1732\n",
-      "train: 21700 0.02088\n",
-      "train: 21800 0.2825\n",
-      "train: 21900 -0.2488\n",
-      "train: 22000 -0.2143\n",
-      "train: 22100 -0.3213\n",
-      "train: 22200 -0.2591\n",
-      "train: 22300 -0.005762\n",
-      "train: 22400 0.7372\n",
-      "train: 22500 -0.3272\n",
-      "train: 22600 -0.01348\n",
-      "train: 22700 -0.1986\n",
-      "train: 22800 -0.2316\n",
-      "train: 22900 0.2392\n",
-      "train: 23000 -0.07332\n",
-      "train: 23100 -0.3443\n",
-      "train: 23200 -0.1578\n",
-      "train: 23300 -0.2344\n",
-      "train: 23400 0.1857\n",
-      "train: 23500 -0.3545\n",
-      "train: 23600 -0.1258\n",
-      "train: 23700 -0.4249\n",
-      "train: 23800 -0.2055\n",
-      "train: 23900 -0.266\n",
-      "train: 24000 -0.3299\n",
-      "train: 24100 -0.3036\n",
-      "train: 24200 -0.2779\n",
-      "train: 24300 0.4617\n",
-      "train: 24400 -0.2216\n",
-      "train: 24500 -0.009392\n",
-      "train: 24600 -0.1579\n",
-      "train: 24700 -0.2255\n",
-      "train: 24800 -0.4099\n",
-      "train: 24900 0.2578\n",
-      "train: 25000 -0.3274\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXsAAAENCAYAAADuRcXXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nO3dd3gU1f7H8feZ2U1PCIQQOoEAShFEqoAICiJWEFQUsKCiclVA7HqVC/Zrx3qt/FBBjQW7qICKCEpRepGO9ISSstns7pzfH5NACCmbZDcJ2e/refIAOzszJ8vuZ86eOUVprRFCCFGzGVVdACGEEMEnYS+EECFAwl4IIUKAhL0QQoQACXshhAgBEvZCCBECJOyFyKOUSlZKaaWUo6rLIkSgSdgLUU5KqTuVUiuVUhlKqc1KqTsLbd+ilHIppTLzfmYX2j5BKbVbKXVYKfWWUiq8wLZkpdRcpVS2UmqtUqq/v/sKURQJeyHKTwFXAbWBc4FblFLDCz3nQq11TN7POUd2VGogcA9wNtAMaAH8p8B+M4BlQAJwP5CqlEr0c18hjiNhL6otpVRDpdTHSql9eTXn2wpsm6SUSlVKfZBXs16qlOpYYHsbpdQ8pdRBpdQqpdRFBbZFKqWeVkptVUodUkrNV0pFFjj1CKXUNqXUfqXU/cWVT2v9pNZ6qdbaq7VeB8wCevn5610NvKm1XqW1PgBMAa7JK19r4DTgIa21S2v9MbACGFravkIUR8JeVEtKKQP4AvgLaIRdix2fV6vNdzHwEVAHeB/4TCnlVEo58/adDdQDbgXeU0qdlLffU0BnoGfevncBVoHj9gZOyjvng0qpNn6UVwFnAKsKbXov72I1u+DFCGiX97vl+wtIUkol5G3bpLXOKLS9nR/7ClEkCXtRXXUFErXWk7XWuVrrTcDrQMFmkiVa61SttQd4BogAeuT9xACP5+07B/gSuCLvIjIaGKe1/kdr7dNaL9Bauwsc9z95Neq/sIO0YEgXZxL25+ntAo+NAJKxm1rmAt8ppeLztsUAhwo8N//vsUVsy98e68e+QhRJeh2I6qoZ0FApdbDAYybwS4F/b8//i9baUkrtABrmb9NaF6ytb8X+hlAX+6KwsYRz7y7w92zscC2WUuoW7Lb7MwpeNLTWvxZ42mNKqauxa/9fAJlAXIHt+X/PKGJb/vb8mn5J+wpRJKnZi+pqO7BZax1f4CdWa31egec0yf9LXo29MbAz76dJ3mP5mgL/APuBHCAlEIVUSo0m72ap1npHKU/X2Dd1wW7uKfiNoSOwR2udlrethVIqttD2VX7sK0SRJOxFdfU7kKGUujvvhqqplGqvlOpa4DmdlVKX5PWLHw+4gYXAIuwa+V15bfh9gQuBmXm1/beAZ/JuAJtKqdPL03VRKTUCeBQYkNfMVHBbU6VUL6VUmFIqIq9bZl0gv7b/f8B1Sqm2eU07DwDvAGit1wN/Ag/l7TsE6AB8XNq+QhRHwl5US1prH3ABcCqwGbtG/gZQq8DTZgGXAweAUcAlWmuP1joXO9wH5e33MnCV1npt3n53YPdu+QNIB56gfJ+Fh7G7Rv5RoC/9q3nbYoFX8sr2D3bXzEH5tW+t9bfAk9ht+duwm5keKnDs4UCXvP0fB4Zprff5ua8Qx1HBWrxEKTUMuAL7DVsP+035CfBooV4GQpSZUmoS0FJrPbKqyyLEiSCYNfs7AB9wH3at5hXgZuD7Qm2pQgghgiyYvXEuzP/amecnpVQ6MA3oC8wJ4rmFEEIUELSwLxT0+f7I+7NRsM4rQoPWelJVl0GIE0llN6ecmffnmko+rxBChLSg3aA97kRKNcKe2OkvrfWAYp4zBhgDEBkZ2blJkyZFPa1UlmVhGHJbQASHvL9EsFXkPbZ+/fr9WuvEwo9XStgrpWKAedijG7v5MfiELl266MWLF5frfPPmzaNv377l2leI0sj7SwRbRd5jSqklWusuhR8P+nQJebMJfoE9DeuZ/gS9EEKIwApq2OfNPpiK3dd+gNZ6RTDPJ4QQomhBC/u8vvTvAWcBF2itFwbrXEIIIUoWzJr9S8ClwCNAllKqR4FtO6Q5RwghKk8wuxQMyvvzfuC3Qj/XB/G8QgghCgnmoKrkYB1bCCFE2UhnYSGECAES9kIIEQIk7IUQIgRI2AshRAiQsBdCiBAgYS+EECFAwl4IIUKAhL0QQoQACXshhAgBEvZCCBECJOyFECIESNgLIUQIkLAXQogQIGEvhBAhQMJeCCFCgIS9EEKEAAl7IYQIARL2QggRAiTshRAiBEjYCyFECJCwF0KIECBhL4QQIUDCXgghQoCEvRBChAAJeyGECAES9kIIEQIk7IUQIgRI2AshRAiQsBdCiBAgYS+EECFAwl4IIUKAhL0QQoQACXshhAgBEvZCCBECJOyFECIESNgLIUQIkLAXQogQIGEvhBAhQMJeCCFCgIS9EEKEAAl7IYQIARL2QggRAiTshRAiBEjYCyFECJCwF0KIECBhL4QQIUDCvobavHkzLperqoshhKgmJOxroKeeeooWLVrwySefsH//fnJycqq6SEKIKiZhXwNt3LgRgIyMDJo1a8aQIUOquERCiKomYV8DvfLKK6xYsYJrrrmG7Oxsvv32W/bs2VPVxRJCVCEJ+xpCa83atWvRWgPQvn17IiIiWLBgAVlZWSQlJVVxCYUQVUnCvoZ4+eWXGTFiBB999BFut/vI46effjpRUVFVWDIhRHUgYV9DbN26laVLlzJ58mSUUsdt9/l8TJkyhY8//rgKSieEqGoS9jXEk08+icfjYenSpYSFhR23PTU1lQcffJAJEyZgWVYVlFAIUZUcVV0AETgOR/H/nZdeeik//PAD55xzDpZlYRhynRcilEjY1wBpaWnEx8djmmaxzzEMg9dff70SSyWEqE6kencCc7vdbNu2jdGjR9O4cWMWLlxY1UUKGEtr3D5NhkdzMFdzINf+M8OjsfJ6HAkh/Cc1+xPY4sWL6d2795F/N2/evNR9Vq5cycyZMxkyZAidO3cOZvH84rHsIE9zaw56INur8Vjg0RzpRlqYaSgSwxWNIhX1IxVmETekhRDHkrCvAMuySEtLIzExsUrOv3fvXhISEsjIyCA1NdWvvvTTpk3jqaeeIisrq0rC3tKa/W7NLpcd8Fm+4kO9OD5Ls9ul2e0Cp6FoHKVoEWMQ7ZDQF6I4EvYVsGLFCgYPHszkyZMZNWpUpZ9/yJAhDB48mJycHCIjI/3aZ/jw4WRmZjJ06NAgl+4or6XZk6PZnaPZm6PxWIFrhvFYms2Zmq1ZmkaRipRYgzinhL4QhUnYV0BsbCxdunThX//6F/379+fDDz/kyy+/5Oabb2bAgAHExsYGvQxKKb+DHqBz586VVqPP8Wk2ZlpsywpswBfF0prt2Zrt2RZ1wxUpMQb1IlSRYw6ECEVyg7YCmjdvTv369bnmmmto0KABWVlZ/PDDDwwdOpS33367QsfWWuPTGq+lOezR7HJZ7Mmx2O+22OOy+OG3xazctJ1Mj/286sLts8u6JN3Hj7t9bMywgh70he13axal+Zi318cul4wpEAKkZl8u69atY+HChVx00UVMnToVn88HcEzf9XU79rI92yIhTBFVTFuyy6fJ9Ng3Jg97NC6fxmtBbt5Pcb1O3C4XNw25GK0tHk39kSYtWhPlAEPZV29TgaEUDgWxTog0FeEmKCDHB5lezWepH7J0wXxG3/0QcbXrEGkqwgx73zBDEWFChMmRm58eS+PT4M27eeq1wKc1uRa4LTvkcyxweavPhSfDo/kjzUejKE3HeAOHIbV8Ebok7Mth2rRpPPbYY4wdO5aXXnqJLMtgR6aPU6++gw+H3oxhmDjDw1mWbl8EIh127xFDARpyLDuIssoZjAf27qJ+0+ZkZRyiXuNkfFqT4Sn8rLxjF7N+yfRXp7J28W+073cep/U95+jza6B/si0yvZruCSYRpgS+CE0S9n7akW2xz61xKkg6qQM9+55Nx4GXMHuXlxzf0aAMjzx20rEdf6+lccuT2RbAGm/9Zi147JO5FRoJO2D4aLr2P5+GzVsGrFzV2aFczS/7fPRIMImVG7giBEnYl8Lt0/x10GJ3gbbf5mcP5e6z7d4sBYO+IK01T9w0nEXffc47S7ZRKyHw3TMrMuVB/8uvDmBJTgwur+bXfT461zFIjJDbVSK01Lh3vNfr5dJLL2Xr1q0VPtYul8VPe31Hgv6fTevZ8OdivyYSU0rhzs7G4XCyedVfFS5LPrfLRfqeXQE7HoDXc1wbUI2Va2kWplmsOOjDW8qNY6/Xe8x00UKcyGpc2D/44IPs37+fMWPGlHmwTr4Mj+a3/T7+SPMdU3P/4YN3uOPCnnz4/CN+HeemR6cyffkuTu3Tv1zlKMpf83/k2i7NeOa2itfMfV4vj4+5jKEtolm+YF6Fj3ei0FqzOdO+kO/NsS/cGzduZOzYscTFxWEYBnFxcbRq1YpGjRqxd+/eKi6xEBVX45pxxo0bx9q1a7n55pvZsGED6enp9OjRo8R98ofsp+dq0t2atNyiR3VGx9aiQXJLWnbwr596UpPk8vwKJYqpVZs+g4eT1LT0qRFKYzoceHJzObnL6Wxfv5oOPftWvIAnkCyvZuF+Hxvnf80D112Ox+PBk/ctJyMjg4yMDACSkpJYvXo1bdq0qcriClEhNS7sk5KSuO222zjzzDOZPn06t956K2lpaWCYeDVY2u534rEgPdcedp+e69/kWsNuuZtht9xd5m8MntxcXrj9Ojr0PosBw68t529ma9utF/M+eY9+Q0dU6Dj57njpXTIPHSCxYZOAHO9Es2vLRu4dfTluV3axzzEMg6+//lrCXpzQalwzjqU1Xg1bszSTHn6Uw4cP8795y/l6p5fZu7z8sNvLj7u9/LzXy8qDPva7rVKDPuNA+jHt2mUZlam1ZtLI8/l51gfs2LAWn9db5t9Ja82s/z1H5sEDAIx9/GUaNm9V5uMUJTI6JmSDHmDW/54r9Z6FZVksWLCgkkokCiuqiW3s2LFs3Lixqot2QqlxYe+x7K/nyw/6GP3wVJ6fvYQGLStWI3vlvn/xwOUDSNu9s8z7KqUYOOJ6Tj2jPwOuuBazhAVGirN28W+8NeUu7hnat9z3IUqzbslC7r9sAMPbJDC4aTjD2yTw6n23smtLzf5Azfv0fXze0m9Q//DDD6SmptKpUyfuvvtuCZ5K8s0339ChQwfeeOMNMjIy0FqTkZHBG2+8QYcOHfjmm2+quognDBWs8KioLl266MWLF5d5P7dP8+2ceag2vUt/sh9yc3K4pnMTXJkZ/O+3DVVSC97w52JmPDuFFu06MvKuyQE//uI53zDl6sEUHlhlOpw4nE7ufm0mnfude8w2r8fD7YO6ce6oGznv6psCXqbKMrhpuF8XUKUUzZs3Z9OmTRiGcUyPLKfTidPpJDU1lUGDBgWzuCHB4/Hw2GOPsW7dOj799FNcrmJGBgJRUVEsX76clJSUSixh8M2bN4++ffuWa1+l1BKtdZfjHq/pYZ9xIJ0fP5xG5qED5Q7K7MwM1i1ZSKczB5Rrf7CbAvZu30JOdhbJbU4p1zG01gGf2GvXlo2MO6dziW3W4ZFRPD97CYfS9uF2ZfPR1MdZsWAeYLdnz1x7AE+um/VLF3Fav4EBLV+wDW+TgCszo9TnxcTEYFkW2dnFv041NXgqW3Z2Nu+88w733HPPkZvkxXE6nYwZM4YXX3yxkkp3PLfPvu932GP/PffIegx5U5jk/Zgc+++Cn2R3XouE2wen1jZZ8dtPAQ/7GneDtrCc7EzefvhuEuo34rLb7iMsIqLMx4iKiS0x6LWGvdth40pF2m6FMwwiozVtu2kSG9nPWf7rXB66chCtTu3KU1/86td5tdZ4c3NxhocDZbtX4C9/2qy9Hg+fvvo0f/0yh93bNgEQFh5Bv0tHcfalV+HOyeaWfh3wej289ftmIqKiA17OYOk75Epmz3irxKYch9NJcnIy69atK/FYHo+HZ599tkqDpyaIioriqquuYuLEiaU+1+PxMH369KC95pbWHMi155PyWHZAa8BtabK9dkBneO3xL+XJlqIEa2LDGh/2iY2a0rnfuVx6271HQtNf2zesoXZifQxHbX6YabBzM0TFQlQMRMbAvn/sgN+0QpF56PggVkpz+nmaS262aNWxC7G1E2h2cnu8Hg8Op/PI83Zt2cis/z3HvE/fJycrk4joGHpfMIzlC+ZiGA7ueOldWnY4rcKvRVH8abP2eT3M++R9rr7vMV5/cDxaa7oPvIgbJj1z5DWt36wFDmcYB/ftoX6zFkEpazBcPGY8c1Knl/gamA4nmzdvPtItszjBDp5QEhMT4/eAtszMzICfX2vNpkzN35kW7mJGyedzZWVyx4U9ufaBJ+hyVvVtxqvxYQ9w35sfHxOuJSkYvPlf753hg/G4nwKK/3peK0GTcoomqanG51Uc2AvLflYs+MpgwVcGbbvV4YyLN9PspCjWLNbUSdJYPlj527e8NXk4Xq8Hy2eHiSszgzmp7wJ20M5JnR60sM/J8u+D4nZlc84Vo0lucwrN23UkKubYufonz/j2hKrR52uQnMLdr83ksesvxZN7bLgUvGcx5eqL/TpeMIInlNx2222kpKRw/fXXExMTU2ozDtgXhkDK8WmWHbDYl+Pf9NhbVi8nNr4OX7z5Iqf1Hcii7z5n4befMf65t6vVegohEfYFg37TymUc3L8vb6bHYy2Z+y1P3Dgcr8dzTE3P4/4KpWZz9mUfUL/ZILIzwZUBtepCyimalPaaOvWh8P9r2m744k2Db6YbrP7dYPXvBQNSAxuByylqasr88ysVxcoF47j9PJOkptAwWVOviSa+LjRtrUlqBhWYIoeI6Bi/2qzzZ/Js173oG98RUdHkZGexftnvuF3ZdO1/fvkLVck69zuXqT/+yUdTH2fB15+Qk5VFZHQMfYdeyUXXj6NBcorfr1OggyeU/Pnnn0ydOpXIyEiGDRvGyJEjeeONN0r8RuV0OgO6SlyaW7M43Vdqbb4gwzTZuWkDk977Cp/Xy//+PZ70PTvpP/xaTjn9zICVraJCIuzzrV2ykHuH9qNVxy7Hhf2uLRt54sbhxdyo9KC1h18+v9zuypns3w24hPpwzf0Ww26xWP6rYucmxT8bFSsXvcq+Hf9G64NAybUHrT1sXfc88CIbVxy/PSxC0+wkTbM2kNxG07S1pmFzTUID/y4C/rRZG6aDzn7ceN25aT3/Hj6QRimtT5iw37VlI19Pe5WUU07ltqdf57anXy/yef68ToEOnlDTsWNHZs2axc6dO2nUqBETJ05k2rRppYb9hAkTAnL+HdkWfx2wSmwzP/abfyZh4eH0unAYUz6YTbOT2gEwfMIDeD0emrfteNw++c20fYdcycVjxvudJYFQ43vjFJSTncXYM9vT+axzufHhF3A4nezcvJF3n3yOBV+/hbZKbpM1HU4GjriOGx9+oczlyrdk7rc8et3QMk0+FhEVx+QZaezaCru3KPb9o0jbDdvW2TeEi+II09RvCk1P0rTvoWne1v5GEBEJznD7xzDK1huntDemx+3muu4tSG7bgUtuvoOOvc/y62uszwsZB+FwOmSkK9z5X3TydlUKHGHQsLmmTtLx36AqYtHsL3j0uqF0PmsQD06bVezz/HmdpDdO4H3zzTcMGzbsmKksILDdXbXWrMuwWH+45IpXcd/887/1Fu6inJ1xmGmP3c/3M94COGafkro196hrsnrhz9L1sjRFhb3PBxnp4M4Bt8uH5TPZs10x56NvWfjd5aA9gH/hGxUTx4w1+8tcLvAvMIqilMFn23KK3JZ5ELauVWxZq9iyRrF9A+zeqjiwt/REjE/UND1JExH1NUvnXY62PH6/IYvz+/df8sjoSwCYtnQ78YlJ+HywcxPs2qLYvVWxeyvs3qbYs01xcB9F3twuTnQtTac+mr6XWHQ6U+Pw71ZMsbatX83X016lXffenHHRZSU+98iHPdd9ZHUysLufRkRESD/7CihpbYaNGzfy7LPPMn36dDIzM4mJiWHUqFFMmDChQhdWrTVpuZr1hzX73SUHfXkqRSsX/sz9l5Y8CWJRFalghX2NasY5fBje/wB+nt+MA5ZB2m47UHZtBY87P1Dy0yG/vbxswevy84ZmUfzp5liUyOji24Fj4qFdD027HsdetF1ZsHsrbPhTsfoPg23rIH2PIjcHct3263Fwn/0DFwDLgWdRajpaZ+IIi6FdtxFcdMNtdOzt/weq24AL6HL2eRxKy+Sb6S42rzZZuVCRVUKgK6WJiYe4OlCrjias4Prpeb9Wjgt2bLB7Pc3/QjH/C4NaCZozh1j0PE/T+jSNafpdTAC2rVvFnRefQWKjJtz0SOnf1jr3O5fnZy9h1pP38NO8ubgyM3A4HIwePZo777xTavTlZFkW3bt3p2/fvjz00EPH3fdISUnhxRdfPKaXk9aaRYsW8dxzzzF16lS/zuOxNLtzNIdy7SU009za77Z5f7sof/7G80e++c///COUUiUO2iu8TzDVqJr9rl3QsGHR22JrayKi7OYLh8Ni7z9dyclaVuZyVaRm7+8AnsLOu/qmgL8ZLMvuOrptnWLrWsWqRYrVvytyso8PZdOhiYqFiCgIj4LwCPu1zP93RCSER2mUgu0bFJtXHd8VNbGRpnFLTf2mmvrNoH4zTf1mdrNMdC38Cur88Qy/fG4w9xODHRuOniOujqbzWZpOfSw6nmHfwC7Jz5/N5KTTunPbgNOo27Axz3y96LhVxootx5r5qDa9MZTi1NoGjaNq3KwjlWrRokX06NGDRo0asWnTJsLCwkrdx+VyERUVhdPpxOVyYZbwBvJamjWHLbZla3ylrGFQHH8/uwXzoTz7gNTs/VKvHowYpcnJ3UZC+8bUSdLUa6xp0NzuG1/QkGZlX1DEdDjpO/TKcpfP326OhV10/bhyn7M4hgFJTSCpiaZrf82wW+wLwJ6tHGkSyv/ZvVWRcQAyDuTvXXqzS1wdzSk9NZ3OtOjQS5MUgFkmlIKkpjDsFouh/7LY8Kfi51mKP7432L1NMTdVMTfVDt7mbTWn9rHoerambXd9TDv/3h1beW7CaBzOMF74fmm5xwVY2u6il+2F1nES+OXVrVs3li5dyq5du/wKeoDIyEhGjx5NQkICbrebqKiiL9R7cyyWH7TIruCyoP5+dgt+8y/PPsFUo8LeNOHNt+HbOZtQbYqp4uexynGFdzidFQpef7vvFdTl7PMq7Y69YUCD5tCgueb0QUdfn9wcyMk++uPOVuS4wF3wMZfCmwsNW2hatNPE17P4e/liTurULShlVQpad9K07qS57iGL7RtgyRyDP39RrF6k2LxasXm1yaevQrOTNFfd56PLWfbvpC2LHucOJiwiosIDwLIzM5j83Atk7/uHD996DYdRffpVnyiUUnTq1IlOnTqVab8333yz2G0HczXrM45dTrQi/P3sFmxyLc8+wVSjwr4sImP8D96CNyorErz+dN8rKCw8gusnPV3u8wVKWIT9E1cn/5GiLpRHH/N5vUw8rwdb1qzg5Z9WBmw65uIoBU1bQ9PWFkNusi9OaxYrls5T/PSpwdZ1iilXOxh8o49Rd1skNW3OXa+879fykqVxOMP46MUnyM1x8eFNExjS9WQiTQn8quDTmp0uzZZMiwO5gW2e9uuzq5xExoxk/EAHhw+AZY0E3qCkzh+G4eSkziOY/4XC6wXLCxsjoXFC2Ub7+yNkw97f4M3vbpk/uKYi/BmaD3a/dmdYWIUvLlXFdDhIbtuBzEMH2b1lU9DDvrCwCOjYW9Oxt2bkXRazXjd4778Gn71msvp3xcSpPuo3q9iC7fmcYWFcMfFBGrVoTUzjlvyy10fXBJPaYRL4/vjll1949dVXGTJkCMOGDSvTvl6vl23btuHyWqik5mzL1uSWYTBUaf5eDnM/NtiyRnFo/x1YvumU2GtPO0nbfXuB7tATgWkl7mNZTpb9NJFlPx0bxU884d/9o7II2bD3J3j97V/ur/yh+UX11T1yzohIzr786oBcXKrSTY9MJTwyqsqHizvDYNi/LNp11zz1L5P1yw5w24DN3DC5Nf0vjwxIn/1Lbjo6YVeOT/PrPh/taxkkx0g7fmn+/vtvFixYQP369f0Oe0trMr3w2tvvcdeN19DrgmHc9cr7AStTTja8dr/JnNSC/38tgeuBV/L+XfCz68QwnZw76gO69W9GXB0PsbVBqaYsnTeT1x8ajq/AdCj5DDOSU3p8SFRccwzTwnSA6YAGMYq6dXMD9vvkC9mwLyl4CzfbKKVoGKmo5VTkWpocH3g1RJgQbSocBoQZ4FD2VKWHPZp9bruLV2H53fc+f+N55n38Pq6szOOG5tcE1W2enDZdNM9/5+WhkV/x919X8+KdQ5n/+YfcMNlH45aBPZel7cVz0nI17WsZhEuzTrEuu+wyJk6cyMUXHzv3kNtnf4a8FmT77OmDc3wajwa3z36NHU3aEJ+YRHRcrYCVJ203/GeUg61rFWERmoEjLDr308TEWzw+5kv27/TQrvv5bFzxC25XUZ/dYz/zA0ecS4dex37ewyMj6TFoMMPHP5C3j++YfezeOFkB+53y1aiul1D2xUt2bdlYYvAmhCva1TKJL8fX8t0uixUHLVwB/Gp5ovF5vXg9uX53awy2Fb/9zH/H3kx2xnA87smYDs25oywuH2dRK8G/Y+R3vSxo/bI/mPPR/9Hq1K6cfdlVRx4PNxUtYwwaRykJ/QI2b95Mw4YNcTgcZGRkEB8fT05ewO9yafbmlL4udKDXd9i1GR4c4WDvdkWjFM3dr3ppdrK9zbIs5n/+IasWzWfMw8+X2NWzLFJfepLZ77/Jo6k/UrdBY9wuF2c2iQlK18uQ/57ZIDmFGx9+gRlr9vPZthxmrNnPjQ+/QMuWLTmtjkmvREe5gh6gfqRB3yQzZL/Of/Lq0wxvk8BX77xc1UU54pTT+/D24uVM/eFWBo7wYfngq7dNbuzt4KOpxtGpGsron03r+Gb6ayye8/Uxj7t9mlWHfPyw28dfB3whfeEvaNiwYdSpU4ffl/5JuiOO+fu8fL/bx7J0H7tdpa8LDfNAbq0AACAASURBVIFd32HrWrhnqB30rU61ePzTo0EP9v2dPoOHc/NjLwYs6AHWL13EwX17cGVm8PW0V7n8pHg+eHdawI5fUGimUDGUUtQOU5xa26RfkhmQwTJOQ9Eh3uSMeg5axBhEhFDtLi4+gdwcF3t3bCvzvh4/5zIvD9M0aZAcz9jHLZ77zkvnfhauTMW7T5pMGORg3bKy/x+17daba+5/nIuuu63I7T6t2ZplMWe3j8XpPrZnWWR5ddDWFK5uvJZmX47FlkyLFem5HMh04c7N5Z/arVh50Ee6u/yvhSc3lx0bS15YpiSbVsL9lzk4uE/RobfF5Bk+4mof3b54zjdsW7eq3McvyWXj7uek07rjcDgJC49Aa82s1A+Dcq6Qb8YBiA9TNI4yaBipgh7Gltbsd2t2u+ylzDK81NgPfHZmBj6Ph9jadUp/cgFb1qzggcsGcNm4+7jwulsDUoP785cfWfDVx5xx8eVFTju7/FfFa/822bFBYRiawTdaDJ9gER55/LGKasYpL1MpYpyQGK5ICFNEOxThpl1JqAnS3JrNWRZ7XBqf1liWxdJ533HK6X0BXeHmvX82ref+SwdgmCav/bKmzAsUrVqoeOR6k6xDii5nW9z9qo+wAgtO+Xw+ru+eQvqenTz91cKgrSsB9nrXOzevZ/gZnWQEbXlFmAqHOnb9xzDD/nDVj7Q/YJXFUIp6EYp6eW8or6VxW5Br2ReqLC8c9Gj25Whyyzm0u7oovMCJP2Y+O4UZz0zBdDj48u2X6Nr/fJb99D0DR1yP6fD/7Vp4WlnD4cDn8eCMiCgy7Dv00jz7tZf3nzb47DWDT14xWfidwZ0ve2nRrsy/ht98WnMoFw7lav4u8LhpKGIckBShaBRpEOs8ccJfa7vdfUOmdVwnhVWLfuGZW6+iy9nncfsLFW+uqN8shfDIKHulND9H34LdPj/zeZOfPlForeg+0OLOl304Cx0iJyuTrgPO5++/lpByStkGfZVVWEQEyW06BO34NS7sHQZEmYoOCSbhpiLaUb1rSQ7D7s1j9105Wk5Law557JpRltf+Sc/Fr7bM6ujwgTTiapd8BzQ74zAznpkCwH/e/4bmbTty39B+bF23irg6CTRs3pJadZNIqF/y6OiipqL1eTwow2D2u29w2pnnFDmLZ1iEvf7A6YM0U+802b5ecc8QB+Of89HzvNJf95zsLH7+bCaOsDDOGlaxee191tGLwPrDFrXCFE2jDBpUwrfP8kpza/5xWex2aXKKuTfRMLklrswMfvp0Bjc+/EKFe9KYpsm/p32GaZilfgPUGn75XPHduwYrF9pNtA6nZshNPoZPsI7MoFrU/PNnDrmS3Vs3ndC95WpcMw7AvHnzyv0VqDrzWHaN/5An70KQW/6JnSqLZVncc0lfNiz7nXeWbqdWQmKJz//tm0/JPHSQ/pdfg1KK1JeeZP4XH+HzeNi2fjWj//0kF48ZX+z+gZqfPzcHXr7XPDLXznUP+bjoenvEbXHNOEvnzeY/oy6gbsMm/O/XdWX6JuIvpRTxTkiMUMQ77XtMZe3lY2lNrgUun71gdrbX7t7o9tmjkSNNiHYowgz7z1pO+xtpYR7LbpLck2O/L/29+bzou89p1/0MYuJrl/7kANn3D7xwh8ny+fb/Z1iE5oyLNJeP85HU9OjzipuzvjzTfZfXCTkRmlKqCfAsMAC72voDMF5rXfY7dgKnoWgYpciv11rabvfP76qWVcHJnoLBMAwio2MJi4xix9/rSg370wcNOebfg8dMIDq2FofS9mJ97sMRVnKbbHmmoi1KWASMe8ZH8smadx4xePM/Zt6UysW/xqf26c9Zw0bRrkcfLMsicH02jtJacyCXI9MB5Id/rFMRbthhbWm7FuvR4NPgtSA3r7nQY9khXRamoYjOuwBEOexjH/ZAem7p3SPzpe/Zxa9ffcyA4dfSfeBFZf69S5OdcZj1f/5BWHgEbbv1OmbbumWKR0abHNqviK2tGXGHRZ/BFtFxxx6jpNXqfF47/J+4cXhAB1pWpqDV7JVSUcBfgBt4AHu0wcNAFNBBa13iqAGp2ZddrqU5kGv/HMqFQ57iv05Xpl2b/6Z2UoMSB1r9/v2XJDZqQrOTT6nQNAblnVa2JLNeN3hrsonDqXlouo9T6vwSsBu0oeLHD/+PFyZeD8ArP68K+BQav3z+IU/9ayTdBlzA/W99cuTxX79SPDfOJNet6HiGxe0v+Iqd/vrV+24tdQqVQKxWV5pg1eyD2fXyBqAFMFhr/ZnWehZwEdAMuDGI5w1ZYYYiKcLg5DiT7nVNBtQ36VHXpEl01Y7ibNC8ZYlBb1kWL9x+PeMHdmXPts0VOlcwppW9+AaLC6/z4fUoHrvBZMum6jU6+ERw5pAr6DN4OJ3OPIekJs0DfvxmJ7enfY8+JLe1b3BqDakvGjx5k4Nct+KcKy0enFZ80APM+/T9Uuet8nk9zPu45KkZHIbdtOY0VJVPF1JQMJtxLgIWaq2PdDLQWm9WSv0KXAw8E8RzC+yv+AV7/rh8mkyP3ePnQK5mbxlW6gkEy7LYvn41zU5uf8zjOVmZdO1/Pju3/F3qlMM52Vm4MjOoXa9+kduDNa3s6Act0vcofv3SYPL9HXmiI9RrXPRzN636k+Xz59LrwmEkNgzARP4nOHuKgCgmTv2/oJ2jaeu2PPLRDwBsWQNvTDJZscBAKc1V91kMudEqdR6kslQUHIZ9TyPChBiHff8k1mn3oCpYscq/P5Ljsz9/bh/k5E254vaB27K3a8Ck8MQJgRXMsG8HFLWC8yrg0iCeVxQj0lREmopEIBn7jXggF3a6LLZl2f2ggyU7M4M7L+zJzk0b+Pe0zzmt7zlHtkXFxjHu2eLnJs/382czeeGOG4iOi+fpr36jboPj09af2UzLswiNYcD4Z30c3AerFoUzaaRmykwvCUVccz547hEWfjuLmPg69L/86jKdp6bJOJDONV2a0qRVG5749KegTptxOB3ef8rgu/cMLEsRG6+55b8+epxb+vt615aNGKbDr+nHY2NjOK+hf9FpKEWEaV8U4v1Y9AeCN+4mmGFfBzhQxOPpQJG34ZVSY4AxAElJScybN69cJ87MzCz3vqEqCruvv8ciKKEfCZzerSsrI8LY88cP6KSyf+jb1Y2kUaPGdDu9J9b639EHtxz3nIvO7s2cD6eV+KF1mCYX9uuFXjO/TOd3Avfe7eD+CR3ZujGOu8/38eCjf9Gk6bE39Lp3aEMMHhqarjKfo6aJAfr07cec2d+R+p9bufLqawN+Dk+u4qvPG/Phe83IzspEKZPzLz7I8FGbiY3zoteUvP+S3xfx5JT/YPm8pZ7LNE369esX9HwJRoYF8wZtLvCM1vqeQo8/DNyjtS7xQiM3aKtOpsdemHl3jsXBIPXtX7/sD/7ZtI5/Nq6nVkIip583pMiaemE+r7fULo353ec8uW4s39EvxoHqPnfwt0U8+nhP1i01iK6lufd1H6ecXvU3wquDovqod+1/Ps6wcC4ffz9JTZIDdq6sw/DzLHsQ3O6tCrgWeIer7n2HoWP9++bmT1fdgqKioli+fHnQF5evSIZVRdfLAxRdgy+uxi+qiRinoqVT0TLWwJvXw+ewx+6TfdBj/72iF4CfZ83kizenYhgGjrBwmrRq41fYFwz69D27iIqNO+7mb7Cnka4V72HKTB9P3wKLZhs8dKXJzY/5GDD86GuitWbZT9/jcecEpathdVRUH3VXZga/fvkxDqeTXhcMrXDYZ2fC4h8U8780WDpP4XHbTSONW2mat2nIHz/GUCfJ/xXI/OmqC/b9r8jISFJTU4Me9MESzLBfhd1uX1hbYHUQzysCyGEoEiMUiQXmC/Fa9uIRLp99oynba/8719KEGYpIE6IcdltlpGn/GZ7X78un4b9PPsEXb04lNjaWjIwMRowaxW1DB6KVwpPXH9ybd9NKa/uGVkbeaOL8JqafP5vJs+OvxfL5ePyTebTp2hMAt8vFpJHn06pjF8ZMKbkvfUWER8Ld//Pxf49qPvufyYt3Oti+3sfV91uYpj2R2/Jf5/DFm1N5+quFJLc5JSjlqC6C2UfdlQWLf7Rvji+Zo8jNC3ilNB16WfS/3KL3hZqsjFu55b93lmktBX964AA4nc5KqdEHUzDD/nPgKaVUC631JgClVDLQC7inhP1ENecwFPFh/t9wKuiVqVP593330rBhQ77++mtatGhBbGzBOXSKP6bXsgePbcmy15GNrZ3A7c+/Q4v2R+csWbFgLqt/n48r83DQu72ZJlz7b4vGrTSv3mcy63WTHRsVd7zoY/2fv/Lpq8/QbcAFRJZjjqATTaAGs+X7eznM/8JeEnDVIkVuztGAb9vNDvfTB1nUSTq6T2nTcRTF3x44Xq/3hA56CG6bfTT2oCoXRwdVTQFisQdVlfgqS5t9zeTxeFi0aBFdu3YlvIwzFBZ0MFezdPt+0rwm4dGxHE7fT0x8HQzDYNlP32NZvqANay9quoSVvykev9Ek44CiaWvNA297jxmGX9MFYjCbKwt+maX47j2Dv5cfOwTo5C4WvS/Q9DzPIqFByefQWpOTneVX91p/yx0XF8ehQ4dKfV6gnFBt9lrrLKXUWdjTJUzHrrL9iD1dgv8jWkSN4nQ66d274qNP48MUZ6UkYml7bpdzR45g9coVPPf+Zww6dyA+bTcrZXvLPj1AebQ/XfPfz708PNrBtvWKOy50cO/rPtp2C40btxUZzLZpJXz3nsFPnxm4Mu0afHQtzVnD7LWDW52qqVtKwOdbt3QRj4+5nCat2zD5/W9KfG6kqbhixEimv/UGnhK+lTidTkaNqtjEdtVBUOfGyZsDZ2gwzyFCl9aah6dMYf369ezavpWD6Wmc3/lkEhLMY55zyAOrD1nsd/t/4648GiTDk595+e9Yk2U/Gfx7uMnwCf9Qv9nPtO3ak4QGjfw6TtbhQ3jcOcQnJrHyt5/JOnyQU/sMIDyyiMn1q4HtG9bYk/L40UoQGR1D5iFYsUDx58+KP3822L3taHNbm64WA0dY9DxPF7mWQGkSGzUlfc9OEuo3LHHZwoZRBh3iDVrcOZGZ06eVGvYTJkwoe2GqmRo3xbEIHUopXnvtNXbu3MmmTZswDIOEhITjnhMfBj0TTbZlKVYfsoK6TkB0HPz7HR9vP6z54k2Td5+8HfiIof96havuua7U/X/+bCZvTbmL+MQkJk3/khnPTmHlbz8B0H3gRdz7+kfVagg+wJrFvxEdW4vMQyV3slOGk7DIkYzq6MDyHf0dYuM1Z15iMfBKi6YnVawsdZIa8GjqHNp07Vnk6xTtsFeiSwi3t6WkpJCamsqwYcPweDzHhL7T6cTpdJ7QPXAKkrAXJ7SHHnoIh8NBfHw8tWuXPGVu02iDehGK5QctdruCV8s3HXD9JIs2XTTPjT+HXPchPn6pIRv+NBk61qLjGbrIofs7Nq7j6VvtxcoP7N3Nwu8+p/NZ5x4J+5zsLPbv2oEzLJz4uvWCVv6yOueK0SS3OYX7L+1Pbk7xi/hqy8nBfbdjOqBtN4tOfTSn9tGkdNAEcFlX2nUvupkwxqnoWdc8bj2Ad999l3HjxpGens6MGTPIzMwkJiaGUaNGMWHChBoR9CDz2YsQtTfHYs3h41dS8kdZliU8uB9mvQ5fvb0Wt2slcAUpp2gGj/HR+Sx9zDS7u7duYtLIC2h6UlvOGjaKVh27sHXtSiJj46jftDkznpnMd++9wVX3PsLQsXeWudyBZlnWMTOUFjcXvD322MmpfT7k/GsG0v50TVTZpiYqF5/PB1pjOhxEOxS9Eo8P+uzsbGJjY1FKcfjwYaKigjedQ1mcUDdohajO6kUYJIYrdudoNmVapLmDU+mJrwtXTszlk5ft2Rhj6wxg44q6PH2rA8PUnHSaplMfzWn9NCmntOCJz+xafP68/wXb+ftffi0/fPAO/2xcX+RI1b5DruTiMeMrZa51n9fL07eOIqlJMiPunIzD6TwymO3th1/g99nvYffDiKHZSSO45anbaH1qCnanvOB7Y9JEvnhzKg9N/5IeZw+kRxE1erCb+d599122b99ebYI+WCTsRchSStEgUtEg0uBgrmbFQd+RRUECyRkWRrcBF7B8wTweSz3AV9N+Y8uq81i3zGDNHxZr/pjH+0/3p3Y9Tdtu9Uk5RdOygyalvSYm/uhxWnfqysvzVrJp9V+MO6fzcSNVZ894izmp0ytlNaX0PTtZ/OM3KMNg4IgbqN+sBds3wPTHW7No9svAyyS30Yx7xkuL9qUeLuCcYWEYpkl4eARd65jFrjMdGRnJFVdcUcmlqxoS9kJgd+XsnWiyOUuz7rAV8O6at0/9PzIOpPH4mMvYuGIZz89eTL3GHZj1+hxmPjuI6LjnOLD3Nn79UvHrl0f3a9La7oJ41jCL+ES7d9Ezt16Nx51z3DkqczWlxEZNuemRF6jbqCmZh1rwwkSTuakKy1KER2ouGWsxdKx13ALeleWi68fhcIbR5+QmR27GFrZx40YaNGhQ42v0+STshcijlKJFjKJRpOLvTHva50CFfmR0DJF5k4Jt37CWpie1x+fNZfn8KdRKqMu9b3QiOtbDhr8Ufy+3f7asVmxfr5j2qMm7Txp07a/x5D6Hx+0u8VxlGalaXlmHwe26hrenGGxaaYepYWoGjvQxfPyxI1urQt2kBjzz2MNkbFvPU089xXXXXXfcDfyRI0eycuVKvv32W3r16lXMkWoOCXshCgk3Fe1qmbSMsWv5W7N1wOYYv/TWe2nX/QwMw0A5ndz8+Es0aJaCM280cdOTNGdfZp/L64FlPym+n2Hwx4+Khd8awPuU1u6dv5pScWGvNbgy4dB+OJim2L8Ttq1X/LNRkZMNnlwwlL0Ob3ik/WdYhCY8AhxOWP/nVtYuSSY3x745Gxuv6TfMYtBVFg0DvwhVmdUKU3SqbRLnVJw1dixz586lcePGDB8+HJ/Ph9Ya0zTp2bMnf/31Fx06dKjqIlcKCXshihFuKjrUNqkfabEkPTBNOw6nkw69+gH2N4mmrduW8Fzo2l/Ttb+P9D0w92OD/3vMv5Gq2VmZfDTVIPMQHEpTHE6zewYdSlMcSuPIbJFltw9IAVrT/vRVDLwSepyrCYsobb/gM5SiZayidayBkde39YorrqBx48Y0a9aM5cuXc+GFF/Laa69x7rnn8uijj9KvX79CczPVXBL2QpSiXoRBj7qKhft9lTL1QlHqJMHQsRYfTfVv2UV0DO8+WXzn9YgoTa26UCtBUzsRmrTSNGmtialljxPQFrhzIDcHct2Qm6Nwu+CfTWv59Yu6JLdtwiMfVp9u2zFOuzZfO+zYi9gNN9zADTfcAMCgQYPYtm2b3SUTCA8P54ILLqj0slYVCXsh/FA7TNEtwWRhmo/S1zMKHn+WXUQ5adpqBF0H+IiKhfiE/GCHWnU1tRIgosz3JPODvQf/emJHqaNlK4uhFE2jFW3jDBxGyd9WZsyYwSOPPMLZZ59dSaWrXiTshfBTQriiax2D38oxtXOgXDxmPHNSp5cY9uERTu578zYaJAdnlLBhGOWaTjigZVCKxlGKVrFGsd0qC4uPj+e///1vkEtWfRmlP0UIka9ehEG0A8wqmp+mQXIKd782k/DIKEyHs8jnjH/u7aB0u9y2bhWe3NyAH7cslFI0ijLol2Ryau3i+8+L40nYC1FGDgU96pqEFTEiszLkj1QdOOI6omLiUMogKiaOJq3bct7VN5PYuCmv3ncrw9skMLhpOMPbJPDqfbeya8vGcp9Ta82kURcyumsye7ZvCdwvUwZJEQZ9Ek06lzBIShRPmnGEKIeEcEWfRJM/0n3lml+nohokp3Djwy8c171yydxvuW9oP3Jzjg66qsjo2vxpGeZ+8t6ROes/eeVpBlfStAxG3ijnljEGtcIk4CtCavZClFOUQ9GrrknDqOrxMcpfB7Zg0OfzeT24Xdk8ceNwv2v4S+Z+y7hzOjN7xlvHLE7y/Yy3GHdOZ5bM/TZgZS8s2qHoVMdkYAO7Ji9BX3HV410qxAnKYSg61zZoHWdU+TzzZVkHtjQFFxAvfDO4PBcOf0WaivbxJv2STJpEGThL6WEj/CdhL0QFKaU4Oc6kS53Su/8F07xP3y+5SyZHR9eWJpAXDn/EhylOq2Nydn2TFjFHB0WJwJGwFyJAGkQa9KxrEl5FN24rsg5sYYG8cBTHVIrGUQZn1HPQp56DxlES8sEkN2iFCKD82TMX7veR5a3cG7cR0f6Nro2MLn3lkEBeOAoyDUVSuKJhpKJehKrSb0KhRmr2QgRYtMMO/MJD94Ot75Ari+17n88wTfoOvbLUY0X4cUEA/y4cALFOu5lmYH2TLgn2TW0J+solYS9EEISbip6JJo0rsafOxWPG43CWHPaWZXHR9eNKPZY/Fw7T4Sz1wmEoRZtaJmfWs18LCfiqI2EvRJCYyq7NdqljEuMMfsiVNLrWdDgxTJP2Pfr41T/enwuHw+ks8cIRmbfua6tYaYuvDiTshQiyhlEGfeuZdIgP/s3b4kbXDhxxHS/PW8nDH8z26zgNklO469UZRW4zHU7CI6O4+7WZxV444pyKM6qgKUsUT27QClEJDKVIjlE0ilJsz9ZszbLI8ATnBm5xo2sLys3JISyi5Eno23U/g9PPG8Jf8+eApXFlZRIZHUPfoVdy0fXjig36hHBF1wSTMGmyqVYk7IWoRE7DXvqwRYy9yPkul8UulyazknruHNy3h0kjzyfj4AHeWPh3iQPBIqNjuOe1D/B5vZiO0qNCKUXzaEXbWtJsUx1JM44QVSQ+zL55eVZ9B70SHTSPMYLeth+XkEja7p3s37mdK9rW5Y1JE0vdx5+gj3EouicYtI83JeirKQl7IaqBhHDFKfEmZyU5GNjAQdcEk2bRRsCbQgzD4JEPv+eJz37GlZnBF29OLfJ5W9YsZ+2ShVhWyXPiRzoUHeJN+iaZ1IuQOKnOpBlHiGom3LRnemwQCe1qaXZka/7OtMgOUFNP05PaobXmhsnPsvr3XxneJoGcrEwiomPoO+RKLh4zng9feIxfv/yY6x56iouuv+24Y0Q7FC1jDZpEKanJnyAk7IWoxhyGfWO3abRiW5ZmU5ZFZgBu7C6d9x3/99j9eD2eI9MiuDIz+O79N5k94y2andyexIaNuWDIMOqGG4DGYSiiTKgXoUgMV1U+8ZsoGwl7IU4A+b15kmMM0t2af1wWe3J0uWr7BWe0LMzy2Svsblu7gtWrV9OqVbMKl11UDxL2Qpxg6oQr6oSbnAIc9mh2ZFukuTVuCywNpgKHARGGIsphTxscYUJYXpP6PVOexyplkjOtNc8//zwvvvhi8H8hUSkk7IU4gcU5FW1rmWXaJ3XGe3hKmb7Y5/Mxffp0CfsaRG6fCxFiMjP9m6nS3+eJE4OEvRAhJibGv5kq/X2eODFI2AsRYkaOHImzlEnOnE4no0aNqqQSicogYS9EiJk4caJfYT9hwoRKKpGoDBL2QoSYlJQUUlNTiYqKOi70nU4nUVFRpKamkpJS+lTI4sQhYS9ECBo0aBDLly9nzJgxxMXFYRgGcXFxjBkzhuXLlzNo0KCqLqIIMOl6KUSISklJ4cUXX5TulSFCavZCCBECJOyFECIESNgLIUQIkLAXQogQIGEvhBAhQMJeCCFCgIS9EEKEAAl7IYQIARL2QggRAiTshRAiBEjYCyFECJCwF0KIECBhL4QQIUDCXgghQoCEvRBChAAJeyGECAES9kIIEQIk7IUQIgRI2AshRAiQsBdCiBAgYS+EECFAwl4IIUKAhL0QQoQACXshhAgBEvZCCBECJOyFECIESNgLIUQIkLAXQogQIGEvhBAhQMJeCCFCgIS9EEKEAAl7IYQIARL2QggRAiTshRAiBEjYCyFECJCwF0KIECBhL4QQIUDCXgghQoCEvRBChAAJeyGECAES9kIIEQIk7IUQIgRI2AshRAiQsBdCiBAgYS+EECFAwl4IIUKAhL0QQoQACXshhAgBEvZCCBECJOyFECIESNgLIUQIkLAXQogQIGEvhBAhQMJeCCFCgIS9EEKEAAl7IYQIAUEJe6VUa6XU80qp5UqpTKXULqXU50qpjsE4nxBCiJIFq2Z/DtAPmAZcCIwFEoGFSqnOQTqnEEKIYjiCdNyZwEtaa53/gFJqDrAFGAdcFaTzCiGEKEJQwl5rvb+Ixw4ppdYDjYJxTiGEEMWrtBu0Sqk6QHtgTWWdUwghhC1YzThFmQoo4LninqCUGgOMyftnplJqXTnPVRc47tuFEAEi7y8RbBV5jzUr6kFVoFm9WEqp/sD3fpzkJ6113yL2vxd4FLhOa/2WH8epEKXUYq11l2CfR4QmeX+JYAvGe8zfmv0CoI0fz8su/IBS6ibsoH+gMoJeCCHE8fwKe611NrC2rAdXSo0CXgae1lo/Utb9hRBCBEbQbtAqpYYAbwNvaK3vCNZ5ivG/Sj6fCC3y/hLBFvD3mF9t9mU+qFJ9gNnAKuBWwCqw2a21XhbwkwohhChWsHrjnAWEA6cBvxbathVIDtJ5hRBCFCEoNXshhBDVS42e9VImZBOBopRqopRKVUodUkodVkp9opRqWtXlEjWDUmqYUupjpdRWpZRLKbVOKfWYUio2YOeoyTV7pdQt2IO0pgFLgXjgLuBUoLfWekkVFk+cIJRSUcBfgBt4ANDAw0AU0EFrnVWFxRM1gFJqIbANmAXsADoBk7B7QfbUWlvF7+3nOWp42NcF0gpNyFYLe0K2L7TWMiGbKJVSahzwDHCS1vrvvMeaAxuAu7TWz1Rl+cSJTymVqLXeV+ixq7ArqmdrredU9Bw1uhlHa71fF7qaaa0PATIhmyiLi4CF+UEPoLXejN354OIqK5WoMQoHfZ4/8v4MSFbV6LAvikzIJsqhHbCyiMdXAW0ruSwidJyZ92dAsirkwh4/JmQTopA6fhOBIAAAAVZJREFUwIEiHk8HaldyWUQIUEo1AiYDP2itFwfimCdU2Cul+iultB8/84rZ/17gSuCWgl/JhRCiulBKxWDfqPUC1wbquJU5xXEgyIRsoiocoOgafHE1fiHKRSkVCXwBtADO1FrvCNSxT6iwlwnZRBVZhd1uX1hbYHUll0XUUEopJ5AKdAEGaK1XBPL4J1QzTnlU8YRsomb4HOihlGqR/4BSKhnolbdNiApRShnAe9hTzQzWWi8M+DlqeD97mZBNVJhSKhp7UJWLo4OqpgCx2IOqMquweKIGUEq9AtwEPAJ8WWjzjkA059T0sJ8EPFTM5q1a6+TKK404keVNjfAsMAC7N9ePwHit9ZaqLJeoGZRSWyhmOUHgP1rrSRU+R00OeyGEELYa32YvhBBCwl4IIUKChL0QQoQACXshhAgBEvZCCBECJOyFECIESNgLIUQIkLAXQogQ8P9QhZ1SPrLPvAAAAABJRU5ErkJggg==\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 25100 -0.2749\n",
-      "train: 25200 0.7059\n",
-      "train: 25300 -0.159\n",
-      "train: 25400 -0.4899\n",
-      "train: 25500 -0.3573\n",
-      "train: 25600 -0.3288\n",
-      "train: 25700 -0.4397\n",
-      "train: 25800 -0.3074\n",
-      "train: 25900 -0.2532\n",
-      "train: 26000 -0.4401\n",
-      "train: 26100 -0.1473\n",
-      "train: 26200 -0.04771\n",
-      "train: 26300 -0.2409\n",
-      "train: 26400 -0.1282\n",
-      "train: 26500 -0.1763\n",
-      "train: 26600 -0.4274\n",
-      "train: 26700 -0.3214\n",
-      "train: 26800 -0.2783\n",
-      "train: 26900 -0.2924\n",
-      "train: 27000 -0.4839\n",
-      "train: 27100 -0.1878\n",
-      "train: 27200 -0.3737\n",
-      "train: 27300 0.01148\n",
-      "train: 27400 -0.3291\n",
-      "train: 27500 -0.008919\n",
-      "train: 27600 -0.4141\n",
-      "train: 27700 0.02007\n",
-      "train: 27800 -0.4151\n",
-      "train: 27900 -0.4876\n",
-      "train: 28000 -0.2893\n",
-      "train: 28100 0.03273\n",
-      "train: 28200 -0.3545\n",
-      "train: 28300 -0.08176\n",
-      "train: 28400 -0.5277\n",
-      "train: 28500 -0.3849\n",
-      "train: 28600 -0.1808\n",
-      "train: 28700 0.6243\n",
-      "train: 28800 -0.1382\n",
-      "train: 28900 -0.4157\n",
-      "train: 29000 -0.5043\n",
-      "train: 29100 -0.2007\n",
-      "train: 29200 -0.2879\n",
-      "train: 29300 -0.03593\n",
-      "train: 29400 -0.4114\n",
-      "train: 29500 -0.09357\n",
-      "train: 29600 -0.1297\n",
-      "train: 29700 -0.3818\n",
-      "train: 29800 -0.3324\n",
-      "train: 29900 -0.271\n",
-      "train: 30000 0.7307\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 30100 0.016\n",
-      "train: 30200 -0.3577\n",
-      "train: 30300 -0.1435\n",
-      "train: 30400 -0.09736\n",
-      "train: 30500 -0.3209\n",
-      "train: 30600 -0.4411\n",
-      "train: 30700 -0.3613\n",
-      "train: 30800 -0.3102\n",
-      "train: 30900 0.003127\n",
-      "train: 31000 -0.3805\n",
-      "train: 31100 -0.5039\n",
-      "train: 31200 -0.2611\n",
-      "train: 31300 0.4362\n",
-      "train: 31400 -0.1305\n",
-      "train: 31500 -0.06019\n",
-      "train: 31600 -0.09598\n",
-      "train: 31700 -0.1576\n",
-      "train: 31800 -0.07163\n",
-      "train: 31900 -0.3561\n",
-      "train: 32000 -0.365\n",
-      "train: 32100 -0.5078\n",
-      "train: 32200 -0.1704\n",
-      "train: 32300 -0.4464\n",
-      "train: 32400 -0.3407\n",
-      "train: 32500 -0.4767\n",
-      "train: 32600 -0.4733\n",
-      "train: 32700 -0.2781\n",
-      "train: 32800 -0.5611\n",
-      "train: 32900 -0.4024\n",
-      "train: 33000 -0.3068\n",
-      "train: 33100 -0.5028\n",
-      "train: 33200 -0.2751\n",
-      "train: 33300 -0.239\n",
-      "train: 33400 -0.4221\n",
-      "train: 33500 -0.3042\n",
-      "train: 33600 -0.4042\n",
-      "train: 33700 -0.5841\n",
-      "train: 33800 -0.115\n",
-      "train: 33900 -0.4258\n",
-      "train: 34000 -0.1761\n",
-      "train: 34100 -0.4703\n",
-      "train: 34200 -0.2885\n",
-      "train: 34300 -0.2225\n",
-      "train: 34400 -0.4583\n",
-      "train: 34500 -0.2198\n",
-      "train: 34600 -0.4471\n",
-      "train: 34700 -0.5334\n",
-      "train: 34800 -0.4232\n",
-      "train: 34900 -0.1241\n",
-      "train: 35000 -0.387\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 35100 -0.2951\n",
-      "train: 35200 -0.0906\n",
-      "train: 35300 -0.04057\n",
-      "train: 35400 -0.1684\n",
-      "train: 35500 -0.415\n",
-      "train: 35600 -0.1741\n",
-      "train: 35700 -0.3815\n",
-      "train: 35800 -0.4587\n",
-      "train: 35900 -0.2565\n",
-      "train: 36000 -0.3454\n",
-      "train: 36100 -0.3148\n",
-      "train: 36200 -0.3147\n",
-      "train: 36300 -0.008909\n",
-      "train: 36400 -0.4153\n",
-      "train: 36500 -0.3658\n",
-      "train: 36600 -0.1066\n",
-      "train: 36700 -0.4122\n",
-      "train: 36800 -0.3988\n",
-      "train: 36900 1.394\n",
-      "train: 37000 -0.4366\n",
-      "train: 37100 -0.3834\n",
-      "train: 37200 -0.2065\n",
-      "train: 37300 0.5142\n",
-      "train: 37400 -0.3017\n",
-      "train: 37500 -0.5638\n",
-      "train: 37600 -0.2884\n",
-      "train: 37700 0.8149\n",
-      "train: 37800 -0.4126\n",
-      "train: 37900 -0.5088\n",
-      "train: 38000 -0.2031\n",
-      "train: 38100 -0.4316\n",
-      "train: 38200 -0.4564\n",
-      "train: 38300 -0.1997\n",
-      "train: 38400 -0.4695\n",
-      "train: 38500 -0.1389\n",
-      "train: 38600 -0.1985\n",
-      "train: 38700 -0.4592\n",
-      "train: 38800 -0.4399\n",
-      "train: 38900 -0.4231\n",
-      "train: 39000 -0.4825\n",
-      "train: 39100 -0.1456\n",
-      "train: 39200 -0.3884\n",
-      "train: 39300 -0.1665\n",
-      "train: 39400 -0.4494\n",
-      "train: 39500 0.178\n",
-      "train: 39600 -0.536\n",
-      "train: 39700 -0.3568\n",
-      "train: 39800 -0.1335\n",
-      "train: 39900 -0.5706\n",
-      "train: 40000 -0.2678\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 40100 -0.3361\n",
-      "train: 40200 -0.2769\n",
-      "train: 40300 -0.1495\n",
-      "train: 40400 -0.4157\n",
-      "train: 40500 -0.509\n",
-      "train: 40600 -0.5947\n",
-      "train: 40700 0.1806\n",
-      "train: 40800 -0.5638\n",
-      "train: 40900 -0.4924\n",
-      "train: 41000 -0.2861\n",
-      "train: 41100 -0.4502\n",
-      "train: 41200 -0.2923\n",
-      "train: 41300 -0.1556\n",
-      "train: 41400 -0.4992\n",
-      "train: 41500 -0.4752\n",
-      "train: 41600 -0.5355\n",
-      "train: 41700 -0.06264\n",
-      "train: 41800 -0.1835\n",
-      "train: 41900 -0.605\n",
-      "train: 42000 -0.4835\n",
-      "train: 42100 -0.4136\n",
-      "train: 42200 -0.4599\n",
-      "train: 42300 -0.2532\n",
-      "train: 42400 -0.5888\n",
-      "train: 42500 -0.4319\n",
-      "train: 42600 0.1936\n",
-      "train: 42700 -0.4948\n",
-      "train: 42800 -0.4585\n",
-      "train: 42900 0.05243\n",
-      "train: 43000 0.6354\n",
-      "train: 43100 -0.5109\n",
-      "train: 43200 -0.5011\n",
-      "train: 43300 -0.1594\n",
-      "train: 43400 -0.4082\n",
-      "train: 43500 -0.5146\n",
-      "train: 43600 -0.5954\n",
-      "train: 43700 -0.4624\n",
-      "train: 43800 -0.3886\n",
-      "train: 43900 -0.3067\n",
-      "train: 44000 -0.4036\n",
-      "train: 44100 -0.4288\n",
-      "train: 44200 -0.3456\n",
-      "train: 44300 -0.4572\n",
-      "train: 44400 -0.1831\n",
-      "train: 44500 -0.4771\n",
-      "train: 44600 -0.07303\n",
-      "train: 44700 -0.2803\n",
-      "train: 44800 -0.3427\n",
-      "train: 44900 -0.4726\n",
-      "train: 45000 -0.4157\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 45100 -0.3842\n",
-      "train: 45200 -0.3435\n",
-      "train: 45300 -0.2018\n",
-      "train: 45400 -0.3381\n",
-      "train: 45500 -0.2645\n",
-      "train: 45600 -0.4835\n",
-      "train: 45700 -0.3632\n",
-      "train: 45800 -0.4354\n",
-      "train: 45900 -0.6229\n",
-      "train: 46000 -0.5243\n",
-      "train: 46100 -0.551\n",
-      "train: 46200 0.1835\n",
-      "train: 46300 -0.4622\n",
-      "train: 46400 -0.3765\n",
-      "train: 46500 -0.5116\n",
-      "train: 46600 -0.3575\n",
-      "train: 46700 0.2924\n",
-      "train: 46800 -0.2678\n",
-      "train: 46900 -0.5704\n",
-      "train: 47000 -0.3927\n",
-      "train: 47100 -0.3198\n",
-      "train: 47200 -0.3481\n",
-      "train: 47300 -0.543\n",
-      "train: 47400 -0.285\n",
-      "train: 47500 -0.1016\n",
-      "train: 47600 0.1696\n",
-      "train: 47700 -0.1291\n",
-      "train: 47800 -0.4109\n",
-      "train: 47900 0.6475\n",
-      "train: 48000 -0.2291\n",
-      "train: 48100 -0.5583\n",
-      "train: 48200 -0.6285\n",
-      "train: 48300 -0.5422\n",
-      "train: 48400 -0.5883\n",
-      "train: 48500 -0.5173\n",
-      "train: 48600 -0.2797\n",
-      "train: 48700 -0.33\n",
-      "train: 48800 -0.4838\n",
-      "train: 48900 -0.1947\n",
-      "train: 49000 -0.4829\n",
-      "train: 49100 -0.4886\n",
-      "train: 49200 -0.398\n",
-      "train: 49300 -0.4105\n",
-      "train: 49400 -0.6567\n",
-      "train: 49500 -0.2483\n",
-      "train: 49600 -0.248\n",
-      "train: 49700 -0.5554\n",
-      "train: 49800 -0.401\n",
-      "train: 49900 -0.2328\n",
-      "train: 50000 -0.4117\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 50100 -0.2059\n",
-      "train: 50200 -0.4191\n",
-      "train: 50300 -0.6203\n",
-      "train: 50400 -0.5944\n",
-      "train: 50500 -0.3157\n",
-      "train: 50600 -0.5358\n",
-      "train: 50700 -0.5212\n",
-      "train: 50800 -0.628\n",
-      "train: 50900 -0.5914\n",
-      "train: 51000 -0.4556\n",
-      "train: 51100 -0.6906\n",
-      "train: 51200 -0.5757\n",
-      "train: 51300 -0.6216\n",
-      "train: 51400 -0.3698\n",
-      "train: 51500 -0.4812\n",
-      "train: 51600 -0.4582\n",
-      "train: 51700 -0.3907\n",
-      "train: 51800 -0.3347\n",
-      "train: 51900 -0.5924\n",
-      "train: 52000 -0.05161\n",
-      "train: 52100 -0.5887\n",
-      "train: 52200 0.06346\n",
-      "train: 52300 -0.3497\n",
-      "train: 52400 -0.4621\n",
-      "train: 52500 -0.4636\n",
-      "train: 52600 -0.199\n",
-      "train: 52700 -0.5693\n",
-      "train: 52800 -0.1125\n",
-      "train: 52900 -0.4188\n",
-      "train: 53000 -0.4789\n",
-      "train: 53100 -0.2192\n",
-      "train: 53200 -0.3205\n",
-      "train: 53300 -0.5039\n",
-      "train: 53400 -0.5589\n",
-      "train: 53500 -0.256\n",
-      "train: 53600 -0.5859\n",
-      "train: 53700 -0.5084\n",
-      "train: 53800 -0.6056\n",
-      "train: 53900 -0.2234\n",
-      "train: 54000 -0.4973\n",
-      "train: 54100 -0.4658\n",
-      "train: 54200 -0.2071\n",
-      "train: 54300 -0.1337\n",
-      "train: 54400 -0.5018\n",
-      "train: 54500 -0.5083\n",
-      "train: 54600 -0.399\n",
-      "train: 54700 -0.3569\n",
-      "train: 54800 -0.3479\n",
-      "train: 54900 -0.5855\n",
-      "train: 55000 -0.5745\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 55100 -0.3495\n",
-      "train: 55200 -0.4135\n",
-      "train: 55300 -0.5178\n",
-      "train: 55400 0.0407\n",
-      "train: 55500 -0.5934\n",
-      "train: 55600 0.4012\n",
-      "train: 55700 -0.5306\n",
-      "train: 55800 -0.2641\n",
-      "train: 55900 -0.5301\n",
-      "train: 56000 0.3507\n",
-      "train: 56100 -0.5479\n",
-      "train: 56200 -0.5012\n",
-      "train: 56300 -0.5158\n",
-      "train: 56400 -0.3515\n",
-      "train: 56500 -0.2489\n",
-      "train: 56600 -0.4201\n",
-      "train: 56700 -0.5697\n",
-      "train: 56800 -0.5252\n",
-      "train: 56900 -0.4077\n",
-      "train: 57000 -0.3253\n",
-      "train: 57100 -0.6657\n",
-      "train: 57200 -0.2893\n",
-      "train: 57300 -0.4924\n",
-      "train: 57400 -0.2679\n",
-      "train: 57500 -0.5879\n",
-      "train: 57600 -0.485\n",
-      "train: 57700 -0.1754\n",
-      "train: 57800 -0.4477\n",
-      "train: 57900 0.5399\n",
-      "train: 58000 -0.2076\n",
-      "train: 58100 -0.2377\n",
-      "train: 58200 -0.2971\n",
-      "train: 58300 -0.1129\n",
-      "train: 58400 -0.5077\n",
-      "train: 58500 -0.5757\n",
-      "train: 58600 0.08109\n",
-      "train: 58700 -0.5244\n",
-      "train: 58800 -0.3019\n",
-      "train: 58900 -0.1113\n",
-      "train: 59000 -0.3951\n",
-      "train: 59100 -0.4641\n",
-      "train: 59200 -0.6042\n",
-      "train: 59300 -0.5076\n",
-      "train: 59400 -0.5421\n",
-      "train: 59500 -0.1061\n",
-      "train: 59600 -0.2968\n",
-      "train: 59700 0.2121\n",
-      "train: 59800 -0.526\n",
-      "train: 59900 -0.4611\n",
-      "train: 60000 -0.4645\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 60100 -0.4997\n",
-      "train: 60200 -0.2832\n",
-      "train: 60300 -0.4246\n",
-      "train: 60400 -0.683\n",
-      "train: 60500 -0.4127\n",
-      "train: 60600 -0.5115\n",
-      "train: 60700 -0.4926\n",
-      "train: 60800 -0.6959\n",
-      "train: 60900 -0.6024\n",
-      "train: 61000 -0.4754\n",
-      "train: 61100 -0.2929\n",
-      "train: 61200 -0.05929\n",
-      "train: 61300 -0.5446\n",
-      "train: 61400 -0.1851\n",
-      "train: 61500 -0.5031\n",
-      "train: 61600 -0.5556\n",
-      "train: 61700 -0.3432\n",
-      "train: 61800 0.3903\n",
-      "train: 61900 -0.5995\n",
-      "train: 62000 -0.4801\n",
-      "train: 62100 -0.5703\n",
-      "train: 62200 -0.4849\n",
-      "train: 62300 -0.6401\n",
-      "train: 62400 -0.4575\n",
-      "train: 62500 -0.3785\n",
-      "train: 62600 -0.6659\n",
-      "train: 62700 -0.481\n",
-      "train: 62800 0.2743\n",
-      "train: 62900 -0.5887\n",
-      "train: 63000 -0.2506\n",
-      "train: 63100 -0.3805\n",
-      "train: 63200 -0.4799\n",
-      "train: 63300 -0.5453\n",
-      "train: 63400 -0.4022\n",
-      "train: 63500 -0.5959\n",
-      "train: 63600 -0.5643\n",
-      "train: 63700 -0.5686\n",
-      "train: 63800 -0.6008\n",
-      "train: 63900 -0.2137\n",
-      "train: 64000 -0.486\n",
-      "train: 64100 -0.59\n",
-      "train: 64200 -0.4985\n",
-      "train: 64300 -0.6159\n",
-      "train: 64400 -0.5215\n",
-      "train: 64500 -0.5433\n",
-      "train: 64600 -0.3497\n",
-      "train: 64700 -0.5788\n",
-      "train: 64800 -0.554\n",
-      "train: 64900 -0.2378\n",
-      "train: 65000 -0.5698\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 65100 -0.6555\n",
-      "train: 65200 -0.6415\n",
-      "train: 65300 -0.6141\n",
-      "train: 65400 -0.6939\n",
-      "train: 65500 -0.5913\n",
-      "train: 65600 0.3291\n",
-      "train: 65700 -0.3898\n",
-      "train: 65800 -0.3637\n",
-      "train: 65900 -0.5646\n",
-      "train: 66000 -0.2308\n",
-      "train: 66100 -0.245\n",
-      "train: 66200 -0.6031\n",
-      "train: 66300 -0.4759\n",
-      "train: 66400 -0.293\n",
-      "train: 66500 -0.5612\n",
-      "train: 66600 -0.4972\n",
-      "train: 66700 -0.1064\n",
-      "train: 66800 -0.624\n",
-      "train: 66900 -0.5692\n",
-      "train: 67000 0.6284\n",
-      "train: 67100 -0.4339\n",
-      "train: 67200 -0.2647\n",
-      "train: 67300 -0.4461\n",
-      "train: 67400 -0.5603\n",
-      "train: 67500 0.2487\n",
-      "train: 67600 -0.2918\n",
-      "train: 67700 -0.3734\n",
-      "train: 67800 0.005057\n",
-      "train: 67900 -0.2184\n",
-      "train: 68000 -0.6369\n",
-      "train: 68100 -0.6717\n",
-      "train: 68200 -0.2967\n",
-      "train: 68300 -0.5906\n",
-      "train: 68400 -0.3586\n",
-      "train: 68500 -0.3405\n",
-      "train: 68600 -0.559\n",
-      "train: 68700 -0.4297\n",
-      "train: 68800 -0.384\n",
-      "train: 68900 -0.6318\n",
-      "train: 69000 -0.5671\n",
-      "train: 69100 -0.7246\n",
-      "train: 69200 -0.5696\n",
-      "train: 69300 -0.6761\n",
-      "train: 69400 -0.3284\n",
-      "train: 69500 -0.4868\n",
-      "train: 69600 -0.3016\n",
-      "train: 69700 -0.5472\n",
-      "train: 69800 -0.3524\n",
-      "train: 69900 -0.4159\n",
-      "train: 70000 -0.3449\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 70100 -0.6308\n",
-      "train: 70200 -0.6324\n",
-      "train: 70300 -0.4466\n",
-      "train: 70400 -0.3397\n",
-      "train: 70500 -0.4859\n",
-      "train: 70600 -0.4712\n",
-      "train: 70700 -0.5666\n",
-      "train: 70800 -0.4675\n",
-      "train: 70900 -0.494\n",
-      "train: 71000 -0.5291\n",
-      "train: 71100 -0.5492\n",
-      "train: 71200 -0.2206\n",
-      "train: 71300 -0.3999\n",
-      "train: 71400 -0.4293\n",
-      "train: 71500 -0.428\n",
-      "train: 71600 -0.3946\n",
-      "train: 71700 -0.722\n",
-      "train: 71800 -0.4989\n",
-      "train: 71900 -0.5032\n",
-      "train: 72000 -0.3629\n",
-      "train: 72100 -0.5504\n",
-      "train: 72200 -0.4529\n",
-      "train: 72300 -0.6499\n",
-      "train: 72400 -0.6617\n",
-      "train: 72500 -0.5534\n",
-      "train: 72600 -0.5168\n",
-      "train: 72700 -0.5097\n",
-      "train: 72800 -0.4929\n",
-      "train: 72900 -0.2385\n",
-      "train: 73000 -0.5748\n",
-      "train: 73100 -0.52\n",
-      "train: 73200 0.6264\n",
-      "train: 73300 -0.6384\n",
-      "train: 73400 -0.6059\n",
-      "train: 73500 -0.6357\n",
-      "train: 73600 -0.1385\n",
-      "train: 73700 -0.6341\n",
-      "train: 73800 -0.5348\n",
-      "train: 73900 -0.6438\n",
-      "train: 74000 -0.7087\n",
-      "train: 74100 -0.3484\n",
-      "train: 74200 -0.4897\n",
-      "train: 74300 -0.6807\n",
-      "train: 74400 -0.2819\n",
-      "train: 74500 -0.4799\n",
-      "train: 74600 -0.5518\n",
-      "train: 74700 -0.5643\n",
-      "train: 74800 -0.5721\n",
-      "train: 74900 -0.4933\n",
-      "train: 75000 -0.6345\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXsAAAENCAYAAADuRcXXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nOzdd3gU1f7H8feZ3U1vhA4CgdCr9KYCIk0EFFARROUi6EURuBbE689+VWxYsGNHQAURAUFEBBXpgiBFkN6FBNKz2d05vz8mgQApm7JJYL+v59mHMDvl7Gbz2TNnzpyjtNYIIYS4tBmlXQAhhBC+J2EvhBB+QMJeCCH8gIS9EEL4AQl7IYTwAxL2QgjhByTshciklIpRSmmllL20yyJEcZOwF6KQlFKLlFLJ2R4ZSqkt2Z7fp5RKy/b8kvO2n6CUOqaUSlRKfaiUCsz2XIxS6ielVKpSaodS6hpvtxUiJxL2QhSS1rqP1jos6wH8Bnx13mr9sq3TM2uhUqoX8DDQHagF1AGezLbdTGAjUB74LzBbKVXRy22FuICEvSizlFLVlFJzlFInlFJ7lVL3ZXvuCaXUbKXUF0qpJKXU70qpFtmeb6SUWq6UOq2U2qqU6p/tuWCl1MtKqf1KqQSl1K9KqeBshx6mlDqglDqplPqvl2WNAa4EPvXy5d0OfKC13qq1PgU8DdyRua/6QCvgca11mtZ6DrAFGJTftkLkRsJelElKKQOYD/wBVMeqxY7PrNVmGYBVk44GZgDfKKUcSilH5rZLgErAWOBzpVSDzO1eAloDnTK3fQgws+33CqBB5jEfU0o18qLItwG/aK33nbf888wvqyXZv4yAJpmvLcsfQGWlVPnM5/ZorZPOe76JF9sKkSMJe1FWtQUqaq2f0lpnaK33AO8DQ7Kts0FrPVtr7QJeAYKADpmPMOD5zG2XAQuAWzK/RP4FjNNaH9Zae7TWv2mtndn2+2RmjfoPrCDNHtK5uQ34+Lxlw4AYrKaWn4DvlVJRmc+FAQnZ1s36OTyH57KeD/diWyFyJL0ORFlVC6imlDqdbZkN+CXb/w9m/aC1NpVSh4BqWc9prbPX1vdjnSFUwPpS2J3HsY9l+zkVK1xzpZS6AqgCzM6+XGu9Mtt/n1NK3Y7V1DMfSAYisj2f9XNSDs9lPZ9V089rWyFyJDV7UVYdBPZqraOyPcK11tdmW6dG1g+ZNfbLgCOZjxqZy7LUBA4DJ4F0ILYYy3o78LXWOjmf9TSgMn/eyrlnDC2A41rruMzn6iilws97fqsX2wqRIwl7UVatBZKUUhMzL6jalFJNlVJts63TWik1MLNf/HjACawG1mDVyB/KbMPvCvQDZmXW9j8EXsm8AGxTSnUsbNfFzAu7N3FeE45SqqZSqrNSKkApFaSUehDrrCKrtv8pMFIp1TizaefRrH1orXcCm4DHM7e9AWgOzMlvWyFyI2EvyiSttQe4Drgc2ItVI58GRGZbbR5wM3AKGA4M1Fq7tNYZWOHeJ3O7t4DbtNY7Mrd7AKt3yzogHphM4f8WrgdOY7XJZxcOvJ1ZtsNAb6BPVu1ba70YeCFzuwNYzUyPZ9t+CNAmc/vngcFa6xNebivEBZSvJi9RSg0GbsH6wFbC+lB+DTx7Xi8DIQpMKfUEUFdrfWtpl0WIi4Eva/YPAB7gEaxazdvAv4EfzmtLFUII4WO+7I3TL+u0M9MKpVQ88AnQFVjmw2MLIYTIxmdhf17QZ1mX+W91Xx1X+Aet9ROlXQYhLiYl3ZzSJfPf7SV8XCGE8Gs+u0B7wYGUqo41sNMfWuseuawzGhgNEBwc3LpGjRo5rZYv0zQxDLksIHxDPl/C14ryGdu5c+dJrXXF85eXSNgrpcKA5Vh3N7bTWh/Kb5s2bdro9evXF+p4y5cvp2vXroXaVoj8yOdL+FpRPmNKqQ1a6zbnL/f5cAmZN53MxxqGtYs3QS+EEKJ4+TTsM0cfnI3V176H1npLPpsIIYTwAZ+FfWZf+s+Bq4HrtNarfXUsIYQQefNlzf5N4Ebgf0CKUqpDtucOSXOOEEKUHF92KeiT+e9/gVXnPe704XGFEEKcx5c3VcX4at9CCCEKRjoLCyGEH5CwF0IIPyBhL4QQfkDCXggh/ICEvRBC+AEJeyGE8AMS9kII4Qck7IUQwg9I2AshhB+QsBdCCD8gYS+EEH5Awl4IIfyAhL0QQvgBCXshhPADEvZCCOEHJOyFEMIPSNgLIYQfkLAXQgg/IGEvhBB+QMJeCCH8gIS9EEL4AQl7IYTwAxL2QgjhByTshRDCD0jYCyGEH5CwF0IIPyBhL4QQfkDCXggh/ICEvRBC+AEJeyGE8AMS9kII4Qck7IUQwg9I2AshhB+QsBdCCD8gYS+EEH5Awl4IIfyAhL0QQvgBCXshhPADEvZCCOEHJOyFEMIPSNgLIYQfkLAXQgg/IGEvhBB+QMJeCCH8gIS9EEL4AQl7IYTwAxL2QgjhByTsxTl2797NmDFjiIiIwDAMIiIiGDNmDLt37y7togkhikDCXpyxaNEimjdvzrRp00hKSkJrTVJSEtOmTaN58+YsWrSotIsohCgkCXsBWDX6wYMHk5qaisvlOuc5l8tFamoqgwcPlhq+EBcpCXsBwMsvv3xByJ/P5XIxZcqUEiqREKI4SdgLAKZPn+5V2H/22WclVCIhRHGSsBcAJCcnF+t6QoiyRcJeABAWFlas6wkhyhYJewHArbfeit1uz3Mdh8PB8OHDS6hEQojiJGEvALj//vtxOBx5ruNwOJgwYUIJlUjkxW1qnB7rYWpd2sURF4G8q3LCb8TGxjJnzhwGDx6My+U652KtYRgEBgYye/ZsYmNjS7GU/s3p0RxI1RxJM0l0gc4MeaUUITYIsyvCHBBkKILtgAanCaYGmwK7AQ4DAg1FiB0CDFW6L0iUKAl7cUafPn3YvHkzU6ZM4bPPPiM5OZmwsDCGDx/OhAkTJOhLyUmnyf4UzbE0jSeHWrzWmhQ3pLg1x9O936/DUATZrPAPtEGwDULtyvrSsEOgTb4MfM1tapwmJLs1iS6Id2piw33zvkvYCwASExPZuXMnDRs2ZOrUqUydOrW0i+TXTK05nKrZnWyS6PJNM43L1LhMSCLn/duNzDMGhyI6QFEuQBHhAJuSL4HCSHFrDqVqTjpNUtzWWZfO4cu7dpjNJ8eXsBcArFy5kmuvvZZu3bqxbNmyc55bu3Ytq1ev5l//+pf0xvExU2sOpGh2JZmkeUq3Ld5tahJNSHRpjqRay+yGokaIIjbMIMQuoQ+Qnp5OcnIyFSpUuOA5rTUnnJp9KZrj6TrHcC8pEvYCALfbTdOmTWnZsuUFz/Xs2ZOEhAQ6duxI27ZtS6F0/uFIqsn2RJMUd9m94Oo2NXuTNftTNNWDFXXDDcId/h36d9xxB/Pnz2fbtm3UqlULU2tOZcCRNJNjabrUv7SzSNgLAPr160e/fv1yfG7ixIk8+eSTeDyeEimL06OJz9CcyrDaM20ACrIiJetPx8C66Bhss5ob7IbCoSAo8+eLxekMzZ8JHuKdZSMUvGFqzcFUzaE0K/QbRvhfTT/JpYlPdfLFF18A8P2aP1j/9kd0Gjic6Go1S7l0F5KwF/maNGkSPXr04I8//iAgIIBWrVoV274zTE2iS5PkggSXFfLJxdBGHWRThDsUFQIVlYMUEWWw9un0aHYkmhxILb7T+6P7djPvvVdZPncG6SnJBIWG0fWGoQwYPZ6qMcV/gV1rqx36SJqmVqiifrhxyV7Y/SfdZM22v/nwjVf4fvbZ97fzdYNYuWAOd908AIBdu3bRoff1dOxzfSmX+FwS9uLMEAh5tcfPmzePZ555hscff7xIYa+11XZ5KE0T79Sk++gUN91j7ftEOmxPsHqZVAlW1AgxykTwH0k12ZxgklGMr3/DT4uZfNcQ3C4XHrfVdTYtOYklMz/khy8+4qG3Z9K+Z85nb0Vlaqt551CqpnaYIibUIOgSCX2nR/PHaZOF332X4/u7evG3WOed1u9y+dczWP71DGZujyMkLLz0Cn4eualKsGjRIsqVK8cDDzyQ6zodOnTg9ttvp0WLFoU6hlUDNPnpuIe1cR6OpJo+C/qcpLg1u5NMlh93s+IfN3uTTTLMkm82cZuajfEe1sd7ijXoj+7bzeS7huBMSz0TRFk8bhfujAyeH30zR/f5dohql6nZmWiy9JiH1Sc9HEgxcZaRNuuCOp2h2ZrgYdlxDxu378rz/QWNIzCQt1ZspfN1g2nS/kpSkxJKp+C5kJq9IC4uDrfbTXp67p20+/btS9++fQu1/2SXZtPpstMmnZCh2ZLhYWuConxgVrdCiApQPr3RKMWtWRfn8UlXynnvvYo7n1FLQfPttNe465nXi/345zO15p90zT/pYCirOa1KkKJysCK4jNT4nR5NXIbGpiAks0yJLs3RNKs5MXtlxJv31/SYLPjwDR56e4ZPy11YEvaCu+++mxEjRuB0Oot93/tTTLYmmLhLoRadH1NrTqRbTT1ZwuxWIFULNohyWHenFoeTTpP18cXbbJPd8rkzLqhxns/0eFg+Z0aJhP05x80W/CpBEeWACoGKyADri7Ykm3tS3Nb1hePpJqcycu7nnhNv3l+P23Xm/fW43Sye/j4bli3ivx9+jS2fcadKQumXQJQJgYGBBAYG5rlOamoqO3bsoHnz5vkOmpbVznkszSzOYvpcsluTnGQ1+QTZrBuJQu1Wm3+EQxGZ9/BBF3Cbmp1JJnuSfTuGTXqKd0NPp6Uk4UxLIzA42GdlyYvO7JZ4KuPsexEZoKgQYIW/1uDRoJQ1xEOY3bqRyyjgl67WZ+9MTciw/j3t0iRkFO534P37a61n2GzM/+ANju77m8T4k5SrVKVQxy1OEvbCa02bNmXv3r1s27aNRo0a5bre6QyruaKs9C8urHSPdUqfnU0ptBu2nvYQFWB9AYTazw0jp8cavuCkU7M/pWRujgoKDSMtOSnf9bSGJTM/oN+/7vV5mbyVkJF3CNuUFfjRAYroQEW43RreIaunZ5rHan5JcEGc0yTZDRkmxfrl6u37GxxqdXJQSnHnEy+xcuHXZSLoQcJeALfccgsej4epU6dSqVKlXNdr3rw5QUFBJCTkfuHppNNkXZyJqww22xQHj7a6Se5OPnvGopQiILOrg6uYQ8ZbXW8YypKZH+bZ1GAYNkzTQ1SF3H/HZZEn29nA7mwVbKUUipJ5v715f212B10HDT3z/zbdr6VN92t9XjZvSW8cP6e1Zv78+Xz11Vf5NuPMnTuXbdu20aFDhxyf/yfdZM3JSzfoc6N16Q83PGD0+PznIwgMZOCYB7my/01FPt7Rfbt555GxDGlUnutrBjKkUXneeWSsz3v7ZKd1yb3fA0aPx57PEOB2h4P+d467YLnWmuTTp3xVNK9J2AsWLlzIzJkziYiIyHO9vC5Wnki3avQ5jcoofK9qTCxNOlwFWDX47Gx2B4HBIUx8dxa3T/pfkY+14afFjOvZmiUzPyQtOQmt9Zn+/ON6tmbDT4uLfIyypmpMLBPfnUVgcEie7+/5N66dOHyAEW1iGNerTamOiwMS9n5PKUWXLl0YMmSI1z1Pzh82Ic6pWRsvQV/aBo15EABlKILDwlHKICQsgl7DRvLakg207tYbAI/bzZ+rfmbN998W+Bj59ed3pqUy+a4hJVrDLyl1mvRmwKiNXFZ3NDZ7BGBgGBFUqDaKqwZs5PiBa9nwk2Lfdjiy13qkJl+Gy+kmPdXJuqX/sGuT4tQ/1rWTkiZt9sJrbrebtm3bsmvXLuLj4wkICCDOqVkT58HjZ003ZVHTjlfx+ZbjJJ2Oz3NohJ0b1/Lfm66hSs06tOvZD9M0sdm8G1bXm/7mblcGk+8awqvfrytQ+cuyX75VvD3JRkpifeCtzAeYJhw/AD8cyGvr9cBl/O9fZ+vWVWpp2vc06dBb06C1xsu3v0gk7IvRV199xa+//sqLL76Iw+Hg5MmTVKxYsbSLlSutNaNHj6ZFixaMGjUq3zZ7u91OUlISKSkp7Nmzh0q1G7A2zlMm+9D7k0N/7+DXBbNp1KYTLa64mrCocnmuX79Vey6r15BOfW7g8xcfZ93Shby0YBWOgIB8j+Vdf3M3e7f9wdG9f1O1dl3cLle+7d1lVWoyvP9/NpbNtoK6dpMFdOrbhuq1KxBd2aqhnzwCxw8q63EATp1QuJyZ3UcdYLPVwGYHR6CJy2mtc2y/Yt77Nua9D1EVNe17mbTupqlQVRNdFdLSir/R5ZILe7epyTCtf0t65MM333yTFStWEBERwYoVK4iPj2fTpk35XjgrLXv37mXatGlER0czZswYr7ZZtGgRVapUwRUQxpo4j99djC2Ltq37jZkvP0XXgUNpccXV+a5vs9l4Y+kmtGlyX4+WHPr7L1YtmstVA27Od1tv+5sD7Ny0juTEBB4f2puug25l9FNTvN62tJ0+AT/MMlj4scGpfxQBQZqhD5xkxks38dVrmmlrdhNZPntFLv+/g6w2e9NU/LVBseZ7xarFBscPKL6fbuP76WfXffbZKPr0Kd7XVDZTqAg8GtI8mqXHPNQLN6gTportLsj83HrrrbhcLtq2bcv06dM5duwYGzduLLNjwJcrV45p06aRlJSEYXhXk6hXrx6nMzRrTkrQlxUxDZsx8N/3U7dFm3OWe9xwdB8c3q1wpllNDqYJdgcEBtsIDLZx/V0f4wiAWg2bc/oEhJcDWx6p4HV/87BwutxwC4f37CQlMYGFH71Z5sNea9i2VjH/A4M1SxSmx8qN2GYm41/1EBAYz9bV3TA95nlBn783HhjN0i8+5p1ftlM1JpbG7TSN22nueNRk33ZYtchg+zpF0mmFzQXh4e5if32XXNhnyTCtQYyOpitaRNlKZIKFkSNHcueddwJWn/SqVavm2zRSmsqVK8fIkSMLtE2yS0uNvgw5eRTijrXDsLVn7Q+KH7+AxFOQmqQ4fhDcGfl97s/tRquUJrwcRJSHqPKayAoQWV5TNQZiGmtadxvAb9/NxMxjbgOb3UG3QcMAqFKzDs98+QMxDZsV8ZUWr4x0WP29YsMyg5REyHDC/h2K0yes98uwadr1MLn2dpPLr9JY9cU6PPrRN16MQXSh5IRTRERXIDQi8pzlSkHtxlC78dn7NjpUsLFtdWJRXl6OLrmwz8jIYOG8uTQlgphGzYl3an4+4aFppEGtUN90PlqyZAm//vor/fr1O1OLj4mJ8cmxSlOqW/Ptlv18OPkJ7A4HY55/q7SL5Jd2bFAs+0rxxy8Gxw7kHeYVq2tq1NOERoAywDDAlWGFnTMt66FISfyLhLgZZKTfR2J8BRLj4dCunPZ9mvyaLLL3N7fZ7TTr2IXUpEQS4k4QFhXNzo1riW3akoCgoEK+A4V3bD8s+Mhg+RyDpNMXvr5ylTQ9hpj0uc0kunLO+yjM9YfhE5/mh1kfExyWd/dmX7rkwn7SxAd5/803uLtafWIaNQfAY2r+OOXhpFPTIsoo9rb8b775hrfffpuIiIgLmmyyRpMsi3O3zp8/H7vdzpVXXplv+VLcmt9OenBh48cvPyE0Mop/P/dmiTWRCTi0Gz573sbqxWcrLQGBr1G5Vjjtew6nWm1FRLTVFBMSpqlQHUK8/NhNGfc0y7+ewfhX69Lyqls4fRK2rPqNbWt/okGrSezatJuj+6uxb1tHPO7FQCDgBs7Wcm12B3aHI8f+5l9NfZ6v33qJa2//N8u//pzet47i9keeK/qb4gWXE9YvUyz/2mDtEoVpZjXPaLoNMqlcU2N3QKXLNNVjIaePdGpSIgkn/6FSjZhCDWp2Wd2GjHj0+aK+lCK55MJ+0OCbmP3VbKKrVDuzzDRNDMPgcKpJgkvTJtpWrBNYDBo0iIiICLp3737O8nnz5nHvvfcyfPhwnn322WI7XnEZO3Ys+/fvZ/fu3XmGfVbQp7k1URUrc8/kt6lSq04JltS/nT4JM182WDLTwPQoAoM1fUeYdOqjmTL+bQ7u/Isxz9WhcbvOhT7GlQNuxpXhJCDQQVRFCA5LZUKf7pgeD78tfBKAR6bN5ud5W7DZn+fI3v7s2vQa8BmQjGELo1XXYfzrsfuoVvvCbp+XxTYEILJ8BVKTEvl53pc+DfvEeNj0i2LjcoM1PyhSEqy/d5tdc/VAk74jPNRt7v3+Nv78Ay/cfQvte/bjkQ/mFKpM+7ZvYdWiuVxWt2Gx3MVcUJdc2Hfq3Jm3PvoUe9MuLJnxAV+//TKD7nmQHkNGAFab8y//eGgaVXzNOt27d78g6AGioqI4dOgQP/74Y7Ecp7j16NGD/fv3U7lyLuerWE03WUEP1k1YPYcWrJ1fFI7bBQs/Npg1xSA1SWEYmp5DTYZM8FC+ilWJufb2u/lz9c/Ub9muSMdqc3Uf2lx9tvtHQFDwBe3yaSlJPPjWp2f+v2/7FOZ/+Dor5ipcTsW6pXDsgOa6ESbdBpkEZhtYs1mnLjw/dwX1L29LpRoxtOrSs0jlzUliPCz90mDlfMXuLQqtz1boajfRdL3B5MoBJuULMS6Zy5lOhWo1qB5bv9DlO7BzG7OmPEPHa28olbBXpX0Lb27atGmj169fX+DtnB7N4mXLUY2u4LtP3uHdR++j66BhTHj1owvWrR5i0DzKwOGjLpput5utW7fSrFkzr3u7lCVJLs3qkxf/6JXFTW//FdXoCp8eY+tqxVuP2M60m7fuZnLHox5qFj5rCsQ0TXZuXEv9lu1IjDvB5pU/0b7XgByHRk6Mh+8/N/juE4P441Z5wyI17XpqmnUyadZRU7F6zseZ9/5r1G3emibtC/d+ag27Nim++9Tg1/nWlw6APcDq7dKyi6bN1WaxvW9ZrQSFcWTvLn788lPqX96W9r3657qedYH2Z7p27Vqo4yilNmit21yw/FIO+9Mn/+H4gb3Ubd4613a2IJuiTpjBZSGFm0Thscceo0OHDvTu3fuiDPTcJLk0q056cpw68Ojev9n0y49UvKzmObVBf+HLsHemwWeTDeZ/YN1SWa22ZuTjHtp0L5t/p9m5XfDbQsX8Dw12bjz3b6FyTU3d5pqYhppajTQ162lsjn94ccxAQiOieGL6gnz3r7V1Yfn0Sfh7s2Ln74o/Vhrs3Wr93SqladVN02uoSYsrNUEhPnmZPuersL/kmnGyi6pQKd/hXNM9mm0JHv5KVNQJU9QL9/4Crtaa119/naeffpqjR49SpUrZGLfaG2lpabhcLsLDwy+4yBrv1KzNY47UHb+v5p3/jqXzdYP9Mux9Zft6xev/sXFkr8Jm1wy+1+TGe00cOfTenf7CY+zYsJqbxj1C805dS7ysObE74KrrNVdd72Hfdg9//GKwZZVi6xrF8QPWY+U5mW4AawkOu5aX7jGoWhtSEuDYAUXyachIV2Q4rQuszjRIScq5K2l4OasHTa9hJlVqFe9rSk9NYfbUydw49mECg0OIcFhTWVpDK0OKx/p7uRjGhbqkwz47Z1oqgcG5f9V7tGZXkuZAqqZOmEFMqMq3ecfj8TBmzBgOHTqUa7v30aNHuffee0lKSmLJkiVFeg3FaebMmYwcOZLbb7+djz/++MzyQ6kmf5w28xzrJrZpS665+Q4at/NtU4a/yEiHz18ymPeegdaKWg0046a4ic2ja/qmn5ey64/1DL7noZIraAHENIKYRiYDRoPHA/t3wL7tiv3bFfu2Kw7tVpw+UQW3K4205CB+8XJMNkegJiwSajXUNGiladhK07SjJsCLXpxrvv+WtUsXMvbFd71+HWuXzOerN57n8N9/MWv2bGqEXHj27jI1x9I0B1JN4vKZZ9mZlsbmlctITjhFt0G3el2O4nDJh73H7eb9x//Dr/O/4u0VWwkvF53n+k6PZnuCh11JiurBilqhBlEBOYe+3W7Pt5dNVFQU8+fPx+12k5SURHh4eKFfS3FKSEggNDSUBg0aANaX3fYEkz3J+U8jWLNBE8a+9J5Py5eanMTOjWup26xVvmO9XMz274CX7rFzYKd1AXbQGA9DJuRcm8/u/z6Zx7a1K2nYpmPJFLQIbDao0wTqNNFk76OvNSTG29j82w6+fmsS4eWuoHW38VSpqYkoD4FBGkcgBARBQCCEhHPORd+COH5wH5PvHoI2TW75z/9RoeplXm0XWaESrbv2YkDv7jkGPYDDUNQIVdQINUh0WVNaHknXOVaYjh/YwzMjbiA0MoorrrsRRwnedHnJh73NbufAX1tJOhXH5pXL6HzdYK+2c5ua/SnWtHLRgYqGEQYVAgveJh8cHMzs2bOpX78+oaGhBd7eVyZMmMC4cePIyMggzqnZfNpDkqv0T0WXz51B/cvbkpqUxKwpT7N93W+079mPe196j4hy5Uu7eMVGa/hhpuL9x2xkOBXV6mjGv+qhQUvvfgeR5SvSsc/1Pi6lbykFkeUhMvoAe/78hiv7BzBglG/mLK5cI4axL73HoV07KF+lOnFHD5ORnkbV2nXz3K5j1+6MuaEn5QO9a9qNcChaRttonjnJujW5uT4zWGDNBk3ofN0gYho1x+XKkLAvbnf/73UCgoKpXLN2obaPd2p+O+GhcpCmSaRBWGYf/T179uByuahduzYBeYwY2L9/7lfeS5MHxY40BwdTPQWeWMHtcrHnz03Y7HZim7UslvLEHz/K1AfvIiwymi7XD2H7ut8AWPvDAuZPe50rB9xMzfqNST596qKu7ackwlsP2/h1vlV5uPpGk7ue8Xh1QXHrml8JjYgkplHZGn6gKJp27ELfEfcQEubbs96sZpOf5kzn1fH/ovN1g3no7Rk5rhtgWNfvYsIUtkLcOGhTiqrBiqrB1lnzP+maw2ma42mah96eWaTXUViXTveRPNRs0IQqteoU+W7P4+kmy//xsOmUh5NOk6effpqGDRvy4YcfFlNJS87RNJPlxz0cSDELNYPO8jmf82D/znw1teh3BWZNhuJyptPm6j40bNOBTn0H0m3wrYx96X069R3Il68/x1sPj2HVom+4s2NdZk+djGn6phboS3u2wn/62Pl1vkFQqGbCa27GveJd0AN89cbz/G/kIOKOHfFtQUuQYRiMfmoKtz70FGkpyRzd+3ex7n/x9PdZuWA2zrRUAOo0uZyQ8LGepPsAACAASURBVIgch2uwG4pGkTauqWLDHn+Q3bt2FflzZgW/QZtoG92r2IgJM0rlznO/CPviZGrNgRST3054SAyMosplNWjYqHG+27388su0adOG+fPnl0Apc/dPuslHP66jTZMGvPyQd8Ma56Rh245Uj61P1SLcSZuaDKsWrWNUhyZ8Nvlrtq+vQ6M2s6lYfSYLPurEicOfsOCjkWxf2w0I4PiBF3j9P2tIS05mzlsJTBoIL46x8clzBisXKk4cLp0ZgLy1bqli0kA7xw4oYptppixy03Wg9wV2ZWSQkZ7GqX+OEhBY8uPK+NrOjesY1bEeL95za7FN4WeaJh88+QAv/HsoLqcTsCp/07ccZ/yUDzFNk20rl5F+ZDdfPHEfQxpG0yDSTnRUJAMGDKBBgwbFevd7kE3RPMpG+RN/seTTt9i6+pdi23d+/KIZB+Drt19ix/pVjHr6VSpWq1Es+7z90Re4/dEXCAhUmFpj5PFtffToUTZs2MCff/5Jv379iuX43kr3aA6navanmiS7NDv37OXI3r+pHtug0Pu8LLYBby3/0/sypMLGFVY3vJ0bFQd3KVKT3MCLwB5mT90EZI2nfv60PfcAYzJv2OkE3EtqUgw7Npiw4ThQBbDe+6iKmnotNPUv19TLfIRFUuoWfGTwwRMGpqnocoPJ2Bc9+V6EPZ8jIIBnZ/9I/PGj+XY0uBjVbNAYwzCo27xVse3T5XRy3YgxxB8/eqbpTylFsMNOtQAX44f056cfl+LIHNzMlTmiZVJSElu2bAHwyRnkb0sX8eZ/H2TwyHto0uHKYt9/Tvwm7H9fvoQtvy2n162jii3ss5x0an6PN2kVbeQa+KNGjWLgwIE0bpz/WUBxSXZptieaHEvX59SUWnXtzRtLN+Y84lMxSjoFa39QrF5ssOlnRYYz+/E2Ax0JCr2VchW/o3zVtkRXMgmP1pSvYo0+GF0ZwqM0NjtnHzawOarjSndxZG8yz42qR2SFy6leZz57/ozi9AnFuqXWrfsAdoemdTdNlxtM2nTXhe7NUVgZ6TDtCYPvP7e+wG75j4ebx5sFfutXzJ3JwV3bubL/TdRq2NQHJS19QSGhPPj2DGrUbVhszRyBwcHnjMFjNxSxYYrYMAO7Yef6/v1YvuzHMyGfXVbIT548mWHDhhEbm/tUjwXVuXNnRo4cSa9eXWgUaeOvRBPTx6elfhP2148eT69hdxLbtHguJsYdO0JEdIUzU7kdSTPxxEHr6Jxvysrq4lgStNbsSdbsSMx5EvDA4GBqNmhSLMdyOZ0knY4nunJVwJocY9V3ip/mGPy+XOFxn30vGrQ2adXF6h+98ecvmPdeKq27xfPQ29dkrpH7GOk5qVo7jCbtO/L35g3cdN/vNOlwFcf2W7fP78x87NqoWLPEYM0Sg5BwTadrreBv0sH3837u2QpvPGBnz58KR6Dm3hc8BWq2yW7pl5+w+ddl1GrY9JINe4BmHbv4bN9Vgg2aRhqE2M9+Jnfs2IHdbs8x7LO4XC6mTJnC1KlTi60sHTt2pGPHs91mowMUG095SHX7LvAv6eESfOmZEdezfd0qJr47i+adu51ZHhmgaF/eVqihF4pDnFOzI9GT780dxWHLqhU8dktvmrS/kme+WMLOjYr3/s9g1x/WpSDD0DTrrOnYW9Ou54UDUO3ZuonoSlWJqpj7QGz5OX5gL+HlyhMSnvM44af+gV++NVgxV/H35rOXqMpX0Vw5wKTL9Sa1GllnDGkpybxw9y3UqNeIEf83OdfaZV7DJWgNOzcpFn1qHdP0KKrU1Ex8102dImT0gb+28uXrzzHqqSkFniXpYnPi8AF2b9lItdp1i1Qp2bjiB47s2cktw+/g8mqROd4vExERQVJS/jNvRUREkJCQUOiyeMOjNUfTNJEOxYaVK2S4hNJ0dN9u5r33KsvnzjgzNdtPc6ZTsXrNM+N3J2Rofv7HQ9vyNsqd9+GaO3cu3377LePGjePyyy8vtnKle6wPyaFUk1MZ+Yf8J89Owu4I4Nrb/025SoUf4qFqTCxoTVqKk9fvt/Hjl1aYRlfWDBxjcmU/k6g8cqlOk6K/B/l1py1XCfrfadL/Tjj0t4cVcw1WfGPN+/nNuza+edeG3aGpXBNcGU/yz8HvSU+twV+/G9RubDX7pCYl5vplAlZTzd5tivU/Wk1WB3aeHaul30gPwx40CS7iLRY1GzThgTen57/iJWDx9PeZPXUyN417hGGFDHuHofjxk6n88sMimkc76HrPPTmul5zs3Zy63q5XEMnJyfz2229kZGRw3XXXYVOKy0J8V0n0adgrpWoAU4AeWFfQlgLjtdYHfHnc3Gxd/Qtrly7khrv/k++YOefb8NNiJt81BLfLhcd99pRvxdxZrFwwh4nvzqJ1t96AFb4rT3hoHGlQJ+xsbXLWrFl8+eWXzJo1C6fTSVhYGLfeeiv3339/gdsDE12aw6kmx9M1SW687r3w05zpfP32y9gDAug3cmyBjnm+yPLVuXXiaWZPDePvPxR2h6b/KJMbx5peT5pRHLTWHNm7i5NHDuU54fZldWHYgyZDHzDZsUGxYq5i7Q8GcUcVh3cDPAK0Z9vazky83o5haILDnyEj/UvqNl1HcHgAjsBkVEoT4pKWkZrsISO9D3HHODNfKRTfWC07Nqxm0afvcEW/G2l7Td/C7+giU//ytrTu1vuCCVC8EWxT1A03qBGiOHzTQGpWimbAgAG5rh8WFuZVzd4Xkw/t2bOHXr16Ub9+fa677rpi3//5fNaMo5QKAf4AnMCjWPdJPwOEAM211il5bV+YZhy3GzZu1qz5fS2OBq0IjYSIbJ0WHh/Wl00//8D4Vz8s0LgUR/ftZlzP1mf66eYkMDiE15ZsuOADWi3YoGmUwU9LFnPDDTeQkZFxTjA7HA4cDgezZ8+mTx7TyZta4zStZpp9KSbxhWymSU44TfzxI6QkJtCokLfaezywYq7iiym2M9Pitb7a5M4nPFTz4r61dx4Zy/6/tnLbpP8VugzZaa25qV4kGc50Zm6PK9DNOc60VJ64tT/NOg6hVqM7ObrX4ODfin3bFPv/SkebzYEDwHqgEdACiARWZe6hAVCHGvUX0qSdSYfe1lgtjtzvsfPaZ5P/j9lTJzNg1Hj+9dgLRd/hJcxQitqhigYRBZuJbsyYMUybNi3PNnuHw8Ho0aOLtc0erMEI+/TpQ5MmTZg6deo5zYbLly+/qJpxRgF1gAZa678zC7EZ2AXcBbxS3Ac8fRratVZA+zPLoitrajexhldt2Po2Yho2K3Dzwbz3Xs13kmG3y8W3017jrmdeP2f5kTSTjTv+5r5Bg3Fm9vPNzuVy4XK5uGHQYN798XfK14zFUBBogCPzpCDdA07T+9p7XsIiowiLjCrUtlrD6sWKz1+ycTCzqaJ6rGbEox7aXuN92basWs6hv//Cbi/4XJ45UUrRuN0VeDxu0pISCxT2G1f8wLa1P+NypjL0gTuBs93snGl2fl3wBgknT9GgZSMO7trBh08fJjDgFI7gBqSnxpOavAul/uapGQf5+ZtZ1G48FEdA4a9BZNf9ptuIqlCJhq0L/4VoKEW4A0JsigAD7ArshnWjj8L62VBZHVfP3qdgAm4TPBpMNC4TUt2Q7Nak+PAiYmFEBiguj7IRmcsYVnm5//77+eSTT/IN+wkTJhSliDkKDg5m+fLlxb7f3PiyZv8jEKS17nze8hUAWus8L7sXpmZ/4gR0v0Zz6lQqbhXC6ThITzn3AxAaqbn8Sk2rriatuupcJxXObkij8mfa6PMSEhbBzO0nL1j+ziNjWTLzw3Oaf85nszvoNWzkBV8WZYHHAysXKOa8aWPfduv9rFRDc8sEDxWr/8JHz9xPrYbNGPfKtHz3deLwAQKCgvnr9zW07NLzTG+m0pKSmMD6H78jICjYq7Fm9m3fQurWlTQefDdgjaRYu0kLpr/wGCvmzqTP8Lu4+9k3fF3sPIXZFZEBispB1qO4J+dJ91ijPO5PNUnw4hpRYR0/uI+UhFPUyaUHncOw5qKoF64u6PK8atUqAgICaNasWZ5DmQAsWrSIwYMHn6l4ndm/l2fdRfXFF1/w+++/8+yzz2LL7CJ2sdXsmwDzcli+FbjRFwesWBHW/Q6Ll61FNboC07Rmk9+7VbF1rTUf5ZG9ipULFCsXWNXm2k00DVubhEdBWCREV9FUqAoR5a2bccIiIT3Fu4szabmst3zujDyDHsDjdrF8zgyfh/17j00gOCSUm8c/muPt4tm5MmDF14o5b1ljrIN1pjT4XpOeQ00cAfD35hB2b9mIMz0t32NvWbWCR2/qQZ/b7ubu//nmdXo8njN/MN4IjYikyw23eL1+TKNmaM72ysiaceim+yaxff0qeg8f7X1hc6G1LnA/83CHIibUoHKQOqdroS8E2RQxYYqYMINTGZo9ySZH03Sx9hNfs2Q+z44cRGhkFG+v2HpOD6QAm9VPPq9hyB9++GF+/vlnFixYQN++eV/v6NOnD5s3b2bKlCl89tlnJCcnExYWxvDhw5kwYUKx9q/PyZ9//skLL7xAly5duPbaa312HF+GfTRwKofl8UCOo1gppUYDowEqV65cqFMcDZCebHWPA6oCVWOhUyxwC2zdksCKH49w6EBj/t51JXu32ti7Nb9wCAPyr9kHBweht/96wXLvvyySctz+6JHDzJvzFSt+XEp6WhpBwcF06X4NAwbdSNVqucz3lgOPx8PCj95EKcUt1/VA5xKKTqfBD4uq8s1XNTl5wvpCqFw1jUE376fbNcdwBGjYbb3XNXUGz015nVoxtXMse3ZrZr5j7SuIfNctDK01T056iGv7X0/7ToWffDtfmZ+v7KoDb7/3PjYS+GfFHGZ++jEej4cJEx8p8O43rl/Hx++/S78bBnFN77xrlAaKIBsoA/ZjPUpDKOAyrYfVylO04G97WRRdrr6GwKAgUv/4iciq1bEb4FBW8+ZhrEduIiMjqVGjBhkZGV7nyODBgxk8+NxRcQ8ePMjBgwcL/Tq8cfjwYSpXrszOnTsJCbEGSUpOTi72Jh5fNuNkAK9orR8+b/kzwMNa6zy/aHzVz/7L15/j8xcf5/rRExj24GS2rlEc+luRkmTNpRl/XBF3FJJOWbPlpCSCad4DTANyr50bNge9b825GaYozUC59QLKEhAUTPcbb2PA6PH59l5wZWSw7KtPSU1K5Ia7/3PB8yeOwE9fGSz42CDhpFVjqlFfM/heD1f2s+5kLQqtNft3bCGyfKUidfnMzdY1v/LI4KsJDgvno/X7CQ7NuwfFs3cOJjgsnBGPTi5Q76z8piVMPBXH8OZVcQQGMuPPE/meQZ3v5bG38fM3sxg+8WkG3zsxx3VC7YraYQY1Q1SBLkiWhAzTGuUxPkOT7NKkeKzrTt5mjc1QhNkhwq4ItkOUQ1ExqHCjT16sLrZmnFPkXIPPrcZfIuo2a0Xrbr2pXrcBAUHQsos1KXFuTBP2br2Phwd+QkZ67mFvehzENhuf43NdbxjqVZt910FDgbP9+ZfNmY4zNc9OS2Skp7Fk5gcsm/3ZOd0/c+IICKDXsDvPWXb6BKz+3hpIbMtKhdbWH1Td5lYXynY9Nd5MrZuSmMDOjWupWb8x5avmfLahlCKmUfP8d1ZIjdt15s4nX+Gy2AbY8+kOk56awrofFqAMg3uef7tYyxFRrjzjXplGTOPm2AtxTWL8qx/R8qpr6Hzdha2dhlLUC7eG381rLKbSFGBY/cUvyzaSp9vUZzoauLVGYbW5aw0urcnwWBeFIx2KCAdl9rVdzHxZs18GBGitrzhv+fLM4xb7BVrw3R20WTVsV0YGpsd9ZrnN7kBrB6ZnNoatN6OfNukz/NyBk7zpummz2Xni84W4Mpx51uTzklv3z/PFHYVViw1WfafYtlZhmtYfliNQ076npsctJi2u0AUav2XKuDtY/rV1zeHa2+++4Hm3y4XdUTy9b4qDx+1m95aNHDuwh6sG3Jz/Btn4csLxvATbFG1yuFnvUuZ0Olm5ciUxMTHUqePdCKtxcXFEREScGdzsYrF27VqOHz9Ov379fFKz9+UQx98CHZRSZ35DSqkYoHPmcxeV1t1689qSDeeM3RESFkGvYSOZ+tMGBt/bE9OjeOcRG1PG20jN1kxfNSaWie/OIjA4BCOXdnKtNU/fPoDn7rwRZ1pqgYMeznb/PHe/1kXqZbMVL96zh5Htl/Cvdgd4/zEbf642MGxWH/mxL7r5eL2bB9/ycPmVBQt6gCYdrqJhm46E5DDtotaa0Z3q89yoG8lITy/w6yqM5ITTeT5vs9up37JtgYO+oOZ/8AaT7x7C6RPH8133rYfH8NI9t3J03+4LnqsYZHBlJf8KeoCJEyfSvXt3PvnkE6+3GT16NCEhIXz77cUTM0uXLqV9+/a88847JCYm+uQYvmzGeR+4F5inlMq6qepp4CDg/Yy/PqC15sThA4RFRReoT3bVmFhGPfUKG5YtpmrturTveXao4uETTS6rq3l7ko3lcwx2/q544M2zk0a37tabRz+ay+PDcr7abpoeMpwFGwjsfNl79CSfhsXTDb7/3OCfQ1kB8RnwPwzbI7Tr8TQdrzVp210TmvtIAF7recu/6HnLv3J87p+D+4g7dhj3eleJTMN28ughJg3sxvV3TaDvHReO2V+SZxmrFn3D1jW/cM3Nd+TZxAZWD5TTJ44z/OFnziwzlHWjUN0wVSoTXpS2Hj16sGzZMipW9H48oKSkJNxuNzVr1vRhyYrX1VdfzX//+18+//xzVq1aRaAP/k58OhCaUqom5w6X8CPWcAn78tvWl804k+8ewm8Lv2biu7PodO3AAh8jLwd3WRNI79tuDR9w/V0mN91nEhjsXX/7olLKYMzzTj6bbCMx3gqH8HKaxm01toCPOHFwFgP//W869S25MfW11pw8cpCTRw7RqG0nnx/v1/lf8eKYYdRq2JRXv1+Pcd5Fh7ceHsMfv/7EqCdfpk33gnd1K0gzzqpF35CWkkTLq3rkeVFaa82uTes5uGs73QbfimEYBNsUraNtRHs5/6k4KyUlhaCgoAJ1wy1LLrYLtGSOgTPIl8cojGoxdQkvV55UH5wu1agHL37r5uP/GSz82MbsqTZ+/sagyw0mS7/Mv799UWkdxpsTrV9rk/Ymg+81ufyqrIust2U+fCclMQG7w0Fg8Nmrc0opKlavScXqJVPTuqLfjZgeD+17Dbgg6LXWnD5xnGP7dxMR7fvRI72dFFwpRf2Wbanfsi0AFQINWkcbBJbS6KkXu9DQIo48dwnyy2kJh/znMT774wjXDLmjwNtu+uVHtq/7DWda7jcRBQTB6KdNXpjnpnYTzT+HFF+9YcPlLP6R887lAIYT28zk/qlu/veVh1ZdvetNUxzenPhvhjapyI9ffnpm2bKvPi32OUW9cdX1QwgMvnCmEqUUE9/9ggff+px6l19Q+SkT6oYbdKggQZ9dYmIia9euzXOd3bt3s2DBAlJS8u7B5q/8MuwdAQGFbv9855GxPDywK8cP7s133QatNC8vcPPIB24G3+vBZvftUJD2AAdPTL+XV77zcNWAcy+yHj+4j+MH8i9zUTRu15mAwCAS408AcOjvHUx96G7uveZyTp/8x6fHzo0zLQ1XRsY5y2x2O1f0u7HE2sB3b9nIrFef4cBfW3N8ft57r3Jr86r8+MXHtIhUNI60SdfDbA4cOEDFihXp27cvbrc71/WmT59Ov379eOihh0qwdBcPvwz7wtJa06BVO+o2b02Vmt51A7PZoX1PzfCJJj1vGYqtmAb/yk4ZBo7AIB6ZNouWXXLudvn1Wy8xunMDFnz0ZrEfP0unawfxyabDDJnwfwBEVajMVdcP4erBwws8pHRxmPnKU9zXoxXP3TkYj9vNlPEj2L/D+3lzi8v3M6Yx8+Wn+Oa9Vzm6bzfvPDKWIY3Kc33NQIY0Ks+mX5aSdCqOtV9/Ss2wi7ON2Zdq1KhBo0aNqFevXp53s9asWZN27dr5dMiBi5nfTl7y8tjb+HPVzzw1cxE16jXyahulFBNe+7jQxxwwejzLZn+W781VOT1vszuwOxyMemoKe/7cyPI5M0hNtq45aNOkdbfeefb2cLsyCIssR6M2vrtAen6ziSMwiPFTPsSTR23Ml9p078u3016nTtMWrFw4h+VzPmf5nM/5z+ufFGg8nKIaePf9nDxyiGadujK2++W4Ms6OfpqWnMQfv/5EYFAQDz/0gF/2uMmPUoqNGzfm+96MGDGCESNGlFCpLj5+G/aJ8XHEHz/C0X27vQ77osrqb5/TTVM2uwPT9OBxu+jUdyAnjxzi0K4dpKUkExwaRtdBQ+l/57jMG6ZGZHavPMWMV55i58a1NGjVLs9jj33pPe554Z0SCZN92zcz9aF/ExAYxP++WorNXjofs3otWvPB2r2EhIWjtSYkLJzUpESuun5IiZajSq06jHryFcb1bH1O0GfxuK3PwtChQ9m8ebPPB966GMmXYNH5bdiPevJlHIFBVKhWw+ttThw+QHi58gSFhKKUwlDgMQvWdTXr5qxvp73G8jkzzgnzUyeOs+q7uXS6diBX9r8p332FRZVj9FNTvD72+T1TfKVi9Vrs276Zei3akhB3olSacLJk3UehlCpUN8vi4s2cCL6Y2PpS88477xAbG0uPHj3OWX7w4EECAwOpWLGifDHkQiYcL4DnRg5i/U+LmTZzNsNu6IfdUDg9mr0pJn8nFX2I18RTcQQGheTYi6Qokk+fIjQyqkT/CFYunMOSzz9gwusfl2rY+0JhhkvwdjC8kpjY+mLVqVMnVq1aRbVq1Th06NA5n+dhw4YxY8YMPvnkE267zbfdi0vCxTZcwkVBa51nN0qHoagZatCxgo1QMw3T46Fb28vPjDQYaFM0jLDRsYKN4CKOIx5Rrnyhgv7w3l28dO9wbqwXeeai3zuPjOXovt2Ypsm4Xm24vVUNDu3+q0jlK4jOfQfx5IzvLrmgLyxvh7n2xcTWl4r333+fUaNGsXLlygsqLg6Hg4iICOrVq1dKpSv7/DrsN/3yIxN6t+XnebPOWW4oReUggzblbfSsauPycjYqBhks/eEH4uLicrwNu3ygomslGzVDS/Yt3fDTYib0assv874gIz0NrTVpyUksmfkh43q2ZtnszzAMg8DgEKrVlj+E0hKUz3DLWXwxsfWlokmTJrz33nvExMScWaa15tChQ3z88cecPn2a9u3b574DP+e3bfYAf65awd5tmzl55CABNkWlQEWlIOsRkMsY4VFRuc/f6jAUl5ezUTVYsfmUSZrHt01kR/ftZvJdQ3IcTTProt97j45jyvfrCAgILLE2e3GhawYNZfGMD/Od63T48OElWKqL386dO2nYsCEdOnRg1apV0l6fB7/86w+0KcoFKG4adhvvz1vCpHvvpmcVG62ibVwWYuQa9N6qHGTQrbKNuj4ec9zbidAXfPBGiQ1VIC4U5lA8O+mBfIfc9dXE1peaN954gyuuuIK4uDj27NlDREQENWp439HCX13yNXuboagQoCgfaE2KEOlQZ29Dr9QI8K7b5W233cbvv//Om2++SZcueQ7FD4DdsO6ErBVqzdF5wqlJdVOs83SWpbltwWr+CndYE14DuE1rQgqtNVkj/KvMhwmY2prCzm2CR1sPpaxhmbPeJ6UUdgWBBgTYwK4UGmvS61Q3eMpoB4MsFYMM2kQbOCrXZfbs2flObC3dLvM3f/58Vq5cyffff8/QoUOJj4+Xi9peuETDXlEt2KB6iNUkUxzTmW3atImtW7eemSPSW6F2RbMo665IU1uz9aR5INltTdsW59SkugsXWN5e9EtNTiLCoQgwrC8hu8J6GGAoK3Sz3qGsktgyQzaLW2ucmXOMZq3nNiHQZn2BRgcU7wxDptZorFPP3E7NTa1JcUOCS3PCaU2F5/Rx05m3lFLUDbOGJ856T0p7YutLxYsvvsi8efO45RbrxjibzUZ0dHQpl6rsu+TC3mFAhAPalM//tnOPx8Obb77JsWPHePbZZ/Ncd+XKlezcuZMmTZoUumyGUoTYIcRuXdCtlTkwX4apSXJZXwBuExJdmpNOnW+bf1BomFfd+YKCAula+eL6VXvzpZF1JhHusKbA01pz2gXH002Op2sSMkon+CMc1hd8+RyGJo6NjWXq1KnSl74IWrRoQYsWLUq7GBediysBvGAohbd1S8MwmDRpEqmpqUycOJHIyMhc1w0PD6d169bFU8jzBBiK8oFcEA6pbk2iS5Pqsb4Q0jPnNrEpq8Y74OZhzP70gzzb7e12O8OGDfNJucsapRTlAqBcgI2GEZDi1hxMNdmXrMko4M1vhWE3rLlhY8OUDGQmypxLLuwLQinFyJEjqV+/PvZSuqU/LyF2RUgeffefmfQA3878NM+wDwgIYNKkSb4oXpkXarfugYgN0+xL0RxIMUkpZJNZXmyGolaIFfQyLLEoq8pewpWw11/P/8Ll7NmzWbhwIUOGDKFXr14lUCrvxMbGykU/LzgMRb1wK4zjnZrDaSZH0zTpRWzfj3AoaoQYXBaiJORFmeeXXS8LatmyZXz88cds3769tItygayLfqNHjyYiIgLDMIiIiGD06NFs3ryZPn36lHYRy5ToQKs9/ZoqNjpXtNMo0kbtMIMaoQZVgg0qBBqUD1SEORQOw2qOsRmKAJsizK6oEmwQZFN0q2yna2U7sVKbFxcJv6/ZA2zYsIGNGzcydOjQHHvbjBo1iubNm3PVVVeVQunyJxf9Cs5QOV8n8cZyw7ooLMTFRMIeuP3229m6dSstWrSgbdu2FzzfsmVLWrZsWQolE0KI4iFhD/Tv358WLVoQGBhY2kURQgifkLCHPPvYL1u2jG3bttG7d2/q1q1bgqUSQojiIxdo8/HRRx8xduxYFi9eXNpFEUKIQpOafaaMjAy2b99O8+bNz7k9v3///iiluPrqq0uxZijefwAABkxJREFUdEIIUTQS9plq167NkSNH2LdvH7Vq1Tqz/MYbb+TGG28sxZIJIUTRSTNOpqZNm1K/fn1OnDhR2kURQohiJzX7TN999x0227mDpy1evJhy5crRqlWrfMciF0KIskxq9pmygt7pdPLLL79w8uRJ7rvvPjp06MCmTZtKuXRCCFE0EvbnOXbsGA899BBVq1alQoUKNGjQQG6oEkJc9KQZJxutNY8++iipqal8++239O7dG8h98gwhhLhYSNhno5Tis88+K+1iCCFEsZNmHCGE8AMS9kII4Qck7IUQwg9I2AshhB+QsBdCCD8gYS+EEH5Awl4IIfyAhL0QQvgBCXshhPADEvZCCOEHJOyFEMIPSNgLIYQfkLAXQgg/IGEvhBB+QMJeCCH8gIS9EEL4AQl7IYTwAxL2QgjhByTshRDCD0jYCyGEH5CwF0IIPyBhL4QQfkDCXggh/ICEvRBC+AEJeyGE8AMS9kII4Qck7IUQwg9I2AshhB+QsBdCCD8gYS+EEH5Awl4IIfyAhL0QQvgBCXshhPADEvZCCOEHJOyFEMIPSNgLIYQfkLAXQgg/IGEvhBB+QMJeCCH8gIS9EEL4AQl7IYTwAxL2QgjhByTshRDCD0jYCyGEH5CwF0IIPyBhL4QQfkDCXggh/ICEvRBC+AEJeyGE8AMS9kII4Qck7IUQwg9I2AshhB+QsBdCCD8gYS+EEH5Awl4IIfyAhL0QQvgBCXshhPADEvZCCOEHJOyFEMIPSNgLIYQfkLAXQgg/IGEvhBB+QMJeCCH8gIS9EEL4AQl7IYTwAz4Je6VUfaXUa0qpzUqpZKXUUaXUt0qpFr44nhBCiLz5qmbfE+gGfAL0A8YAFYHVSqnWPjqmEEKIXNh9tN//b+9uQuSooiiO/48gYjSoIa6COmYjRhGVLETFiCZbP8CVYECEkEUkLiQQCDh+RFdGQVQQUVwILqKgceVH0IUyYFRER40Kzkh2RsNITAhIjovqwaFTjT3TVV3pqvODpplXXe++xeP261dVd94CXrTtxQZJB4E5YCewtaa4ERFRopZkb/toSduCpJ+AdXXEjIiIwcZ2gVbSGuBa4IdxxYyIiEJd2zhlXgAEPD/oA5K2Adt6fx6XdHiFsdYCZ/y6iKhI5lfUbZQ5dkVZo5Zsqw8kaTPw4RBBPrV9e8n5u4GngYdsvzZEPyORdMj2xrrjRDdlfkXd6phjw67sPweuHuJzJ/obJG2nSPR7xpHoIyLiTEMle9sngB+X27mkB4CXgGdt713u+RERUY3aLtBKuhd4HXjV9qN1xRnglTHHi27J/Iq6VT7HhtqzX3an0m3AB8As8DBwesnhU7a/rjxoREQMVNfdOHcA5wE3Ap/1HZsHpmqKGxERJWpZ2UdExNml1VUvU5AtqiLpMkn7JS1I+kvSO5Iub3pc0Q6S7pP0tqR5SSclHZb0jKTVlcVo88pe0g6Kh7TeAL4CLgZ2AdcDt9r+ssHhxYSQtAr4BjgF7AEMPAWsAq6z/XeDw4sWkDQD/Aa8CxwBbgCmKe6CvNn26cFnDxmj5cl+LfBHX0G2iygKsh2wnYJs8b8k7QT2AVfZ/qXXdiXwM7DL9r4mxxeTT9Kltn/va9tKsVC90/bBUWO0ehvH9lH3fZvZXgBSkC2W4y5gZjHRA9j+leLmg7sbG1W0Rn+i7/mi915Jrmp1si+TgmyxAtcA35W0zwIbxjyW6I5NvfdKclXnkj1DFGSL6LMGOFbS/idwyZjHEh0gaR3wBPCR7UNV9DlRyV7SZkke4vXJgPN3A/cDO5b+JI+IOFtIupDiQu0/wINV9TvOEsdVSEG2aMIxylfwg1b8ESsi6XzgALAe2GT7SFV9T1SyT0G2aMgsxb59vw3A92MeS7SUpHOB/cBGYIvtb6vsf6K2cVai4YJs0Q7vATdJWr/YIGkKuKV3LGIkks4B3qQoNXOP7ZnKY7T8PvsUZIuRSbqA4qGqk/z3UNWTwGqKh6qONzi8aAFJLwPbgb3A+32Hj1SxndP2ZD8NPDbg8LztqfGNJiZZrzTCc8AWiru5PgYesT3X5LiiHSTNMeDfCQKP254eOUabk31ERBRav2cfERFJ9hERnZBkHxHRAUn2EREdkGQfEdEBSfYRER2QZB8R0QFJ9hERHfAvCiAH3lGbPlsAAAAASUVORK5CYII=\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 75100 -0.2593\n",
-      "train: 75200 -0.6236\n",
-      "train: 75300 -0.08129\n",
-      "train: 75400 -0.3822\n",
-      "train: 75500 -0.6475\n",
-      "train: 75600 -0.4307\n",
-      "train: 75700 -0.4553\n",
-      "train: 75800 -0.4995\n",
-      "train: 75900 -0.4326\n",
-      "train: 76000 -0.5237\n",
-      "train: 76100 -0.6655\n",
-      "train: 76200 -0.3014\n",
-      "train: 76300 -0.4181\n",
-      "train: 76400 -0.5699\n",
-      "train: 76500 -0.5513\n",
-      "train: 76600 -0.5511\n",
-      "train: 76700 -0.4707\n",
-      "train: 76800 -0.5458\n",
-      "train: 76900 -0.4743\n",
-      "train: 77000 -0.2894\n",
-      "train: 77100 -0.6029\n",
-      "train: 77200 -0.1736\n",
-      "train: 77300 -0.5338\n",
-      "train: 77400 -0.6966\n",
-      "train: 77500 -0.5877\n",
-      "train: 77600 -0.3778\n",
-      "train: 77700 -0.5219\n",
-      "train: 77800 -0.4417\n",
-      "train: 77900 -0.711\n",
-      "train: 78000 -0.4956\n",
-      "train: 78100 -0.6015\n",
-      "train: 78200 -0.2942\n",
-      "train: 78300 -0.5023\n",
-      "train: 78400 -0.6093\n",
-      "train: 78500 -0.5752\n",
-      "train: 78600 -0.4982\n",
-      "train: 78700 0.5446\n",
-      "train: 78800 -0.4493\n",
-      "train: 78900 -0.514\n",
-      "train: 79000 -0.5537\n",
-      "train: 79100 -0.0422\n",
-      "train: 79200 -0.1685\n",
-      "train: 79300 -0.5534\n",
-      "train: 79400 -0.5862\n",
-      "train: 79500 -0.7527\n",
-      "train: 79600 -0.6949\n",
-      "train: 79700 -0.537\n",
-      "train: 79800 -0.2861\n",
-      "train: 79900 -0.592\n",
-      "train: 80000 -0.1725\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 80100 -0.6382\n",
-      "train: 80200 -0.5556\n",
-      "train: 80300 -0.4024\n",
-      "train: 80400 -0.6858\n",
-      "train: 80500 -0.4718\n",
-      "train: 80600 -0.4404\n",
-      "train: 80700 -0.1249\n",
-      "train: 80800 -0.6377\n",
-      "train: 80900 -0.6056\n",
-      "train: 81000 0.04496\n",
-      "train: 81100 -0.5175\n",
-      "train: 81200 -0.459\n",
-      "train: 81300 -0.6442\n",
-      "train: 81400 -0.4858\n",
-      "train: 81500 -0.6696\n",
-      "train: 81600 -0.5937\n",
-      "train: 81700 -0.677\n",
-      "train: 81800 -0.7368\n",
-      "train: 81900 0.02835\n",
-      "train: 82000 -0.4812\n",
-      "train: 82100 -0.5787\n",
-      "train: 82200 -0.5941\n",
-      "train: 82300 -0.7607\n",
-      "train: 82400 -0.01379\n",
-      "train: 82500 -0.544\n",
-      "train: 82600 -0.4083\n",
-      "train: 82700 -0.5421\n",
-      "train: 82800 -0.5329\n",
-      "train: 82900 -0.6963\n",
-      "train: 83000 -0.5836\n",
-      "train: 83100 -0.7281\n",
-      "train: 83200 -0.07811\n",
-      "train: 83300 -0.6082\n",
-      "train: 83400 -0.7577\n",
-      "train: 83500 -0.3074\n",
-      "train: 83600 -0.5638\n",
-      "train: 83700 -0.5026\n",
-      "train: 83800 -0.6515\n",
-      "train: 83900 -0.6092\n",
-      "train: 84000 -0.6231\n",
-      "train: 84100 -0.7262\n",
-      "train: 84200 -0.3121\n",
-      "train: 84300 0.2351\n",
-      "train: 84400 -0.7038\n",
-      "train: 84500 -0.6218\n",
-      "train: 84600 -0.6128\n",
-      "train: 84700 -0.5159\n",
-      "train: 84800 -0.6961\n",
-      "train: 84900 -0.4771\n",
-      "train: 85000 -0.3866\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 85100 -0.7066\n",
-      "train: 85200 -0.4893\n",
-      "train: 85300 -0.3894\n",
-      "train: 85400 -0.6736\n",
-      "train: 85500 -0.5539\n",
-      "train: 85600 -0.2071\n",
-      "train: 85700 -0.5272\n",
-      "train: 85800 -0.3043\n",
-      "train: 85900 -0.3074\n",
-      "train: 86000 -0.4278\n",
-      "train: 86100 -0.7092\n",
-      "train: 86200 -0.5411\n",
-      "train: 86300 -0.6925\n",
-      "train: 86400 -0.719\n",
-      "train: 86500 -0.5624\n",
-      "train: 86600 -0.5489\n",
-      "train: 86700 -0.6837\n",
-      "train: 86800 -0.5912\n",
-      "train: 86900 -0.686\n",
-      "train: 87000 -0.608\n",
-      "train: 87100 -0.6017\n",
-      "train: 87200 -0.5255\n",
-      "train: 87300 -0.493\n",
-      "train: 87400 -0.4132\n",
-      "train: 87500 -0.374\n",
-      "train: 87600 -0.5148\n",
-      "train: 87700 -0.5235\n",
-      "train: 87800 0.2358\n",
-      "train: 87900 -0.6785\n",
-      "train: 88000 -0.4794\n",
-      "train: 88100 0.801\n",
-      "train: 88200 -0.2338\n",
-      "train: 88300 -0.6678\n",
-      "train: 88400 -0.5584\n",
-      "train: 88500 -0.7697\n",
-      "train: 88600 -0.605\n",
-      "train: 88700 -0.1149\n",
-      "train: 88800 -0.6044\n",
-      "train: 88900 -0.6848\n",
-      "train: 89000 -0.1964\n",
-      "train: 89100 -0.466\n",
-      "train: 89200 -0.4586\n",
-      "train: 89300 -0.4717\n",
-      "train: 89400 -0.5004\n",
-      "train: 89500 -0.7392\n",
-      "train: 89600 -0.7288\n",
-      "train: 89700 -0.7451\n",
-      "train: 89800 -0.701\n",
-      "train: 89900 -0.5205\n",
-      "train: 90000 -0.6016\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 90100 -0.6851\n",
-      "train: 90200 -0.7329\n",
-      "train: 90300 -0.6505\n",
-      "train: 90400 -0.4302\n",
-      "train: 90500 -0.7013\n",
-      "train: 90600 -0.6514\n",
-      "train: 90700 -0.5535\n",
-      "train: 90800 -0.4494\n",
-      "train: 90900 -0.3957\n",
-      "train: 91000 -0.2012\n",
-      "train: 91100 -0.3615\n",
-      "train: 91200 -0.7466\n",
-      "train: 91300 -0.6954\n",
-      "train: 91400 0.2363\n",
-      "train: 91500 -0.4793\n",
-      "train: 91600 -0.3451\n",
-      "train: 91700 -0.3886\n",
-      "train: 91800 -0.6714\n",
-      "train: 91900 -0.4519\n",
-      "train: 92000 -0.136\n",
-      "train: 92100 -0.4423\n",
-      "train: 92200 -0.7106\n",
-      "train: 92300 -0.4152\n",
-      "train: 92400 -0.6478\n",
-      "train: 92500 -0.56\n",
-      "train: 92600 -0.6129\n",
-      "train: 92700 -0.7449\n",
-      "train: 92800 -0.6618\n",
-      "train: 92900 -0.4279\n",
-      "train: 93000 -0.4671\n",
-      "train: 93100 -0.4209\n",
-      "train: 93200 -0.3372\n",
-      "train: 93300 -0.4563\n",
-      "train: 93400 -0.3912\n",
-      "train: 93500 0.004813\n",
-      "train: 93600 -0.5544\n",
-      "train: 93700 -0.7031\n",
-      "train: 93800 -0.7427\n",
-      "train: 93900 -0.6343\n",
-      "train: 94000 -0.5875\n",
-      "train: 94100 -0.7114\n",
-      "train: 94200 -0.6236\n",
-      "train: 94300 -0.6985\n",
-      "train: 94400 -0.2566\n",
-      "train: 94500 -0.4993\n",
-      "train: 94600 0.1537\n",
-      "train: 94700 -0.6535\n",
-      "train: 94800 -0.3541\n",
-      "train: 94900 -0.6471\n",
-      "train: 95000 -0.398\n"
-     ]
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {
-      "needs_background": "light"
-     },
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "train: 95100 -0.6686\n",
-      "train: 95200 -0.5709\n",
-      "train: 95300 -0.7112\n",
-      "train: 95400 -0.5405\n",
-      "train: 95500 -0.6825\n",
-      "train: 95600 -0.538\n",
-      "train: 95700 -0.6084\n",
-      "train: 95800 -0.5746\n",
-      "train: 95900 -0.6882\n",
-      "train: 96000 -0.7418\n",
-      "train: 96100 -0.4833\n",
-      "train: 96200 -0.8083\n",
-      "train: 96300 -0.4333\n",
-      "train: 96400 -0.2728\n",
-      "train: 96500 -0.5085\n",
-      "train: 96600 -0.5066\n",
-      "train: 96700 -0.4973\n",
-      "train: 96800 -0.5204\n",
-      "train: 96900 -0.156\n",
-      "train: 97000 -0.4574\n",
-      "train: 97100 -0.3583\n",
-      "train: 97200 -0.7343\n",
-      "train: 97300 -0.5368\n",
-      "train: 97400 -0.6153\n",
-      "train: 97500 -0.3679\n",
-      "train: 97600 -0.5705\n",
-      "train: 97700 -0.5925\n",
-      "train: 97800 -0.538\n",
-      "train: 97900 -0.6398\n",
-      "train: 98000 -0.272\n",
-      "train: 98100 -0.6799\n",
-      "train: 98200 -0.6772\n",
-      "train: 98300 -0.5507\n",
-      "train: 98400 -0.6174\n",
-      "train: 98500 -0.5484\n",
-      "train: 98600 -0.6124\n",
-      "train: 98700 -0.6531\n",
-      "train: 98800 -0.6382\n",
-      "train: 98900 -0.2357\n",
-      "train: 99000 -0.5752\n",
-      "train: 99100 -0.4605\n",
-      "train: 99200 -0.6954\n",
-      "train: 99300 -0.615\n",
-      "train: 99400 -0.707\n",
-      "train: 99500 -0.6642\n",
-      "train: 99600 -0.7637\n",
-      "train: 99700 -0.1616\n",
-      "train: 99800 -0.505\n",
-      "train: 99900 -0.1382\n",
-      "\n"
-     ]
     }
    ],
    "source": [
     "epochs = 100000\n",
-    "for epoch in tqdm(range(epochs)):\n",
+    "PLOT_AFTER = 10000\n",
+    "PRINT_AFTER = 1000\n",
+    "for n_iter in tqdm(range(epochs)):\n",
     "    model.train()\n",
+    "    \n",
     "    data_train = dataset_train.generate_curves()\n",
+    "    \n",
     "    (context_x, context_y), target_x = data_train.query\n",
     "    target_y = data_train.target_y\n",
+    "    \n",
     "    context_x = context_x.cuda()\n",
     "    context_y = context_y.cuda()\n",
     "    target_x = target_x.cuda()\n",
     "    target_y = target_y.cuda()\n",
     "\n",
     "    optim.zero_grad()\n",
-    "    y_pred, kl, loss, y_std = model(context_x, context_y, target_x, target_y)\n",
-    "    if epoch % 100 == 0:\n",
-    "        print(f\"train: {epoch} {loss.item():4.4g}\")\n",
+    "    y_pred, kl, loss, mse_loss, y_std = model(context_x, context_y, target_x, target_y)    \n",
     "    loss.backward()\n",
     "    optim.step()\n",
-    "    if epoch % 5000 == 0:\n",
-    "        model.eval()\n",
-    "        with torch.no_grad():\n",
-    "            data_test = dataset_test.generate_curves()\n",
-    "            (context_x, context_y), target_x = data_test.query\n",
-    "            target_y = data_test.target_y\n",
-    "            context_x = context_x.cuda()\n",
-    "            context_y = context_y.cuda()\n",
-    "            target_x = target_x.cuda()\n",
-    "            target_y = target_y.cuda()\n",
-    "            y_pred, kl, loss, y_std = model(context_x, context_y, target_x)\n",
-    "            plt.title(f\"epoch {epoch}\")\n",
-    "            plot_functions(target_x.detach().cpu().numpy(),\n",
-    "                           target_y.detach().cpu().numpy(),\n",
-    "                           context_x.detach().cpu().numpy(),\n",
-    "                           context_y.detach().cpu().numpy(),\n",
-    "                           y_pred.detach().cpu().numpy(),\n",
-    "                           y_std.detach().cpu().numpy())"
+    "    \n",
+    "    writer.add_scalar('train/loss', loss, n_iter)\n",
+    "    writer.add_scalar('train/mse_loss', mse_loss, n_iter)\n",
+    "    writer.add_scalar('train/y_std', y_std.mean(), n_iter)\n",
+    "    writer.add_scalar('train/kl', kl.mean(), n_iter)\n",
+    "    \n",
+    "    if n_iter % PRINT_AFTER == 0:\n",
+    "        y_pred, kl, val_loss, y_std = test(model, dataset_test, writer, plot=False, global_step=n_iter)\n",
+    "        print(f\"train: {n_iter} loss={loss.item(): 4.4g} val_loss={val_loss.item(): 4.4g}\")\n",
+    "\n",
+    "    if n_iter % PLOT_AFTER == 0:\n",
+    "        test(model, dataset_test, writer, plot=True, global_step=n_iter)"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {},
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2020-02-01T03:19:26.014782Z",
+     "start_time": "2020-02-01T03:11:00.300Z"
+    },
+    "scrolled": true
+   },
    "outputs": [],
    "source": []
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Test"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "ExecuteTime": {
+     "start_time": "2020-02-01T03:19:32.800Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# Final validation\n",
+    "loss = torch.stack([test(model, dataset_test)[2] for _ in tqdm(range(100))]).mean().cpu()\n",
+    "print(loss)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "ExecuteTime": {
+     "start_time": "2020-02-01T03:19:32.800Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "writer.add_hparams(hparams, dict(val_loss=loss))"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1624,18 +1106,6 @@
    "language": "python",
    "name": "jup3.7.2"
   },
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.7.2"
-  },
   "toc": {
    "base_numbering": 1,
    "nav_menu": {},