From bb785deb30ebd938a2c23620e99eb75c101fb2b6 Mon Sep 17 00:00:00 2001
From: Tom Hosking <tomhosking@gmail.com>
Date: Tue, 9 Apr 2024 16:16:39 +0100
Subject: [PATCH] Add model - Hierarchical Residual Quantization (HRQVAE)
 (#140)

* Add HRQVAE model - seems to work OK

* Bug fixes

* Add example notebook and update readme

* Update expected output in test

* Fix typos, update docstrings
---
 README.md                                     |    1 +
 .../models_training/hrqvae_training.ipynb     | 4125 +++++++++++++++++
 src/pythae/models/__init__.py                 |    3 +
 src/pythae/models/auto_model/auto_config.py   |    5 +
 src/pythae/models/auto_model/auto_model.py    |    5 +
 src/pythae/models/hrq_vae/__init__.py         |   19 +
 src/pythae/models/hrq_vae/hrq_vae_config.py   |   37 +
 src/pythae/models/hrq_vae/hrq_vae_model.py    |  128 +
 src/pythae/models/hrq_vae/hrq_vae_utils.py    |  153 +
 .../models/nn/benchmarks/mnist/resnets.py     |  230 +
 tests/test_HRQVAE.py                          |  760 +++
 11 files changed, 5466 insertions(+)
 create mode 100644 examples/notebooks/models_training/hrqvae_training.ipynb
 create mode 100755 src/pythae/models/hrq_vae/__init__.py
 create mode 100755 src/pythae/models/hrq_vae/hrq_vae_config.py
 create mode 100755 src/pythae/models/hrq_vae/hrq_vae_model.py
 create mode 100755 src/pythae/models/hrq_vae/hrq_vae_utils.py
 create mode 100755 tests/test_HRQVAE.py

diff --git a/README.md b/README.md
index b601f2e8..ecaacdc2 100644
--- a/README.md
+++ b/README.md
@@ -107,6 +107,7 @@ VAE with Inverse Autoregressive Flows (VAE_IAF) |  [![Open In Colab](https://col
 | Regularized AE with L2 decoder param (RAE_L2) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/models_training/rae_l2_training.ipynb) | [link](https://arxiv.org/abs/1903.12436) | [link](https://github.com/ParthaEth/Regularized_autoencoders-RAE-/tree/master/) |
 | Regularized AE with gradient penalty (RAE_GP) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/models_training/rae_gp_training.ipynb) | [link](https://arxiv.org/abs/1903.12436) | [link](https://github.com/ParthaEth/Regularized_autoencoders-RAE-/tree/master/) |
 | Riemannian Hamiltonian VAE (RHVAE) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/models_training/rhvae_training.ipynb) | [link](https://arxiv.org/abs/2105.00026) | [link](https://github.com/clementchadebec/pyraug)|
+| Hierarchical Residual Quantization (HRQVAE) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/models_training/hrqvae_training.ipynb) | [link](https://aclanthology.org/2022.acl-long.178/) | [link](https://github.com/tomhosking/hrq-vae)|
 
 **See [reconstruction](#Reconstruction) and [generation](#Generation) results for all aforementionned models**
 
diff --git a/examples/notebooks/models_training/hrqvae_training.ipynb b/examples/notebooks/models_training/hrqvae_training.ipynb
new file mode 100644
index 00000000..dc71a51b
--- /dev/null
+++ b/examples/notebooks/models_training/hrqvae_training.ipynb
@@ -0,0 +1,4125 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Install the library\n",
+    "%pip install pythae"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import torch\n",
+    "import torchvision.datasets as datasets\n",
+    "\n",
+    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "%load_ext autoreload\n",
+    "%autoreload 2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "mnist_trainset = datasets.MNIST(root='../../data', train=True, download=True, transform=None)\n",
+    "\n",
+    "train_dataset = mnist_trainset.data[:-50000].reshape(-1, 1, 28, 28) / 255.\n",
+    "eval_dataset = mnist_trainset.data[-10000:].reshape(-1, 1, 28, 28) / 255."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythae.models import HRQVAE, HRQVAEConfig\n",
+    "from pythae.trainers import BaseTrainerConfig\n",
+    "from pythae.pipelines.training import TrainingPipeline\n",
+    "from pythae.models.nn.benchmarks.mnist.resnets import Encoder_ResNet_HRQVAE_MNIST, Decoder_ResNet_HRQVAE_MNIST"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "config = BaseTrainerConfig(\n",
+    "    output_dir='my_model',\n",
+    "    learning_rate=1e-4,\n",
+    "    per_device_train_batch_size=64,\n",
+    "    per_device_eval_batch_size=64,\n",
+    "    num_epochs=100, # Change this to train the model a bit more\n",
+    ")\n",
+    "\n",
+    "\n",
+    "model_config = HRQVAEConfig(\n",
+    "    latent_dim=128,\n",
+    "    input_dim=(1, 28, 28),\n",
+    "    num_embeddings=10,\n",
+    "    num_levels = 3,\n",
+    "    kl_weight = 0.01,\n",
+    "    init_scale = 2.0,\n",
+    "    init_decay_weight = 0.5,\n",
+    "    norm_loss_weight = 1.0, # 0.5\n",
+    "    norm_loss_scale = 1.5,\n",
+    "    temp_schedule_gamma=10,\n",
+    "    depth_drop_rate = 0.1,\n",
+    ")\n",
+    "\n",
+    "model = HRQVAE(\n",
+    "    model_config=model_config,\n",
+    "    encoder=Encoder_ResNet_HRQVAE_MNIST(model_config), \n",
+    "    decoder=Decoder_ResNet_HRQVAE_MNIST(model_config) \n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pipeline = TrainingPipeline(\n",
+    "    training_config=config,\n",
+    "    model=model\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Preprocessing train data...\n",
+      "Checking train dataset...\n",
+      "Preprocessing eval data...\n",
+      "\n",
+      "Checking eval dataset...\n",
+      "Using Base Trainer\n",
+      "\n",
+      "Model passed sanity check !\n",
+      "Ready for training.\n",
+      "\n",
+      "Created my_model folder since did not exist.\n",
+      "\n",
+      "Created my_model/HRQVAE_training_2024-04-05_09-25-49. \n",
+      "Training config, checkpoints and final model will be saved here.\n",
+      "\n",
+      "Training params:\n",
+      " - max_epochs: 100\n",
+      " - per_device_train_batch_size: 64\n",
+      " - per_device_eval_batch_size: 64\n",
+      " - checkpoint saving every: None\n",
+      "Optimizer: Adam (\n",
+      "Parameter Group 0\n",
+      "    amsgrad: False\n",
+      "    betas: (0.9, 0.999)\n",
+      "    capturable: False\n",
+      "    differentiable: False\n",
+      "    eps: 1e-08\n",
+      "    foreach: None\n",
+      "    fused: None\n",
+      "    lr: 0.0001\n",
+      "    maximize: False\n",
+      "    weight_decay: 0\n",
+      ")\n",
+      "Scheduler: None\n",
+      "\n",
+      "Successfully launched training !\n",
+      "\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d501d8fe00884236b5bbef59c0a552bb",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 1/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7478459637004a60a9153df0ce7bf4f3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 1/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 101.965\n",
+      "Eval loss: 54.0718\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "29c201c72220479c8a72971e75986c05",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 2/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "040d5486808a45d2885b579795631382",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 2/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 55.6926\n",
+      "Eval loss: 50.3147\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0d1b24f2f71a470f9bdcc02ef19eceb2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 3/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8b2403855c8d42a695f4747b926c9ec9",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 3/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 51.5036\n",
+      "Eval loss: 47.2385\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "512f9a1407f747c6be0f4b743697adf3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 4/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "aa70c481e5a84998827dad60f93ae24a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 4/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 47.8921\n",
+      "Eval loss: 43.9516\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "559271072f91499893e9df72ef07f695",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 5/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "239ae82f69254e29847b16b02b2234b3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 5/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 45.1995\n",
+      "Eval loss: 40.7514\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "248dda2ba3ba4b49b69b0c29dcdd82a3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 6/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "797ec95e73e64093b7383b753d22136e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 6/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 43.1035\n",
+      "Eval loss: 38.6163\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "003a4a0a4efa44e48489929e6e8dad81",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 7/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "4711f1d7c0314d599a169cf5f6b0e453",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 7/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 41.9886\n",
+      "Eval loss: 37.355\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "73f5c1ba217149c7a6a5aa077499ae41",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 8/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c2a65d7ad4844305988730148f7d18d6",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 8/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 40.9794\n",
+      "Eval loss: 36.1791\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "9d58ff5db2834b62a8c7c240242083cd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 9/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e671ee4f1d3b40f4975de5c5fecf4183",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 9/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 39.8063\n",
+      "Eval loss: 35.226\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "af549f8782944f0cbb4e9763003316d3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 10/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8e8644c877674d3da058c73691795057",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 10/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 38.8261\n",
+      "Eval loss: 34.2064\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "b79368ea434441fe88af1252d1179696",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 11/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "79e76823022044b5a8e6a04650de8c91",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 11/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 38.3993\n",
+      "Eval loss: 33.9564\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "305a70ad9d1f452891eee485f31b812e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 12/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "58aa4c9291f74f40bc62b04a72920c69",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 12/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 38.0116\n",
+      "Eval loss: 33.4185\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7092b3d7a016463896338f2ceeba7459",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 13/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f9abe23b552843c3b1387ec6380eddc8",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 13/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 37.8371\n",
+      "Eval loss: 33.0628\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "93e56686c07843fbb11b2070726306a0",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 14/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7c7adb468b7b48bdb72f3b48cce56c3b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 14/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 37.4615\n",
+      "Eval loss: 32.6782\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "82d2cc8b6a654384be820b663810cbcd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 15/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0071b3b62c7443dfba75832cdf71f0cc",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 15/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 37.0028\n",
+      "Eval loss: 32.3543\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "47e04c91d1cb43928aa0e1429ba3984b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 16/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5f6e26ad24fc49bba56917f6e224eb9c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 16/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 36.7326\n",
+      "Eval loss: 32.5343\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "1807286b2e06492abbe2a810bf544b3e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 17/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "426b4f30a0d0485d97076a92105b5fa8",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 17/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 36.6532\n",
+      "Eval loss: 31.7224\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e304d76af83f4828ad126dec9703fa8e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 18/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0f069aac9b9a40849e3b916eec4fad4e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 18/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 36.1471\n",
+      "Eval loss: 31.621\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "14ffc4f983824a68b7530a04b8eb232a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 19/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "47725e447af340ff90ba97133932a891",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 19/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.9401\n",
+      "Eval loss: 31.5179\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "380cc284533147b7b8c98e09572c3ea0",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 20/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "35175d46322e439e984e6516001af97e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 20/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.7204\n",
+      "Eval loss: 31.0614\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "429d22f06e034434930b0ea9ea6b0fdd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 21/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "88627775b21a4b6dbc491704155cc5c5",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 21/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.755\n",
+      "Eval loss: 31.0521\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d59685564dc1438a9e1f0915dc63407c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 22/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "468e0fc716d54961b2d17d179f212f0e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 22/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.5125\n",
+      "Eval loss: 30.9777\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c9f67ead36324f2daf57ec117b39fb6c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 23/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c79babb1c3cd4fe1bb22f62c3cf06489",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 23/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.2867\n",
+      "Eval loss: 30.9093\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8d08cf036e4d4fe1951ec5bdeb5e183b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 24/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "87a1cfdd659241c5bcc715484d64f049",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 24/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.2111\n",
+      "Eval loss: 30.7066\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "92f4300be883420ab2fd40cdaa40a502",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 25/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3c0066c79d0644a390e46a05581f0da8",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 25/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.9795\n",
+      "Eval loss: 30.6674\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a9faf68dedb5445fb41e0f1fc23ad906",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 26/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3f20523d05ad40cd8ed1d177156ebe12",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 26/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.0619\n",
+      "Eval loss: 30.4481\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "650da9751489466eaab021719180cba9",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 27/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8b323851eca64a48943c598733b6dc6b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 27/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.8622\n",
+      "Eval loss: 30.4097\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d588d387dfdc4fa89fda44777e98406e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 28/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "39b94484b5334eb48a2b12c9c1389139",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 28/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 35.0153\n",
+      "Eval loss: 30.397\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "46f5dce209ec4597b00c241851d6eab7",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 29/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "1b5817f950bd45d7adcdb75186b8006a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 29/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.9702\n",
+      "Eval loss: 30.1431\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "73381dc97fd143d09e3355a3f2e12013",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 30/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "47f67b92a75d491582311c7d8b7cdca2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 30/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.6816\n",
+      "Eval loss: 30.2088\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "da93d7011160430b9d411d226189219f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 31/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d491604ae5a941b2a7b311d25a6ec8d5",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 31/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.6856\n",
+      "Eval loss: 30.1159\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3925490aa3a34424842a6ca419a6d76c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 32/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0c923ddef6574378a15bcdd36fed27c5",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 32/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.8329\n",
+      "Eval loss: 29.9623\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "ed0e905a40104af191db49dcfea4f1a9",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 33/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "b42faae585ea4fb2b272dfec3c548714",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 33/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.5247\n",
+      "Eval loss: 29.9011\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d68ddb361660408ea05e79dd426f7d2d",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 34/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5e430fc6f52a4a479ed7a73c42927caa",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 34/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.4178\n",
+      "Eval loss: 29.7393\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "ac05ddd6ad0242f4aecb9544437d1270",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 35/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "b92cb38120aa4411a010162c0c3452d6",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 35/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.5664\n",
+      "Eval loss: 29.6141\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2260481b355d4dd29b93e37b9dcfc87c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 36/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "6690b8eb5c9e49f5bda2dee95e6f96fd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 36/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.4733\n",
+      "Eval loss: 29.6439\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "fd9f1e23660041fdb6a152fafa494ce6",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 37/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3eb37f83586b41e5bb5fe038f9da094f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 37/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.1947\n",
+      "Eval loss: 29.6019\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "24d4a6c6595144ebb03e70db02786586",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 38/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e2fc117b8698473890755eba5e10e61c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 38/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.2864\n",
+      "Eval loss: 29.5434\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d1382fa9ab844614b3a9c3ebf0da6338",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 39/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c25780b6ccfe418782a8709c3e647ac0",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 39/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.4757\n",
+      "Eval loss: 29.6029\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "304109b81e8b43c2a1fda17dd24192d9",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 40/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e256820dc7164312948cc1a654343b88",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 40/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.27\n",
+      "Eval loss: 29.8103\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d96ef58e6497439e8f395cb3427d7e02",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 41/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8ee31386798e4c1286a089e22daec32a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 41/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.1275\n",
+      "Eval loss: 29.4015\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d1f799fcaf254f34948cd6bf66a631d8",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 42/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a84f853ff27b4d25a6dfb01923e5c479",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 42/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.124\n",
+      "Eval loss: 29.3188\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "384d0e44709f4d148567e6a782f3625f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 43/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "352eabd5bdcf426c843cfd8b25cb1f03",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 43/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.1174\n",
+      "Eval loss: 29.6814\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c7217204c33e4f24aee8923e5b38ee77",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 44/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "b9acebb236674fb2ab904ec77079bbfd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 44/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 34.2653\n",
+      "Eval loss: 29.2887\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3e42728c434948a8b4f6d47c05afe10c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 45/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "205307ccc3f14e1d8077b78309b4e251",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 45/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9657\n",
+      "Eval loss: 29.2766\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f6112c5549a24834ae33f8d290696fd1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 46/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "ab72b45451ed44778e7191f668976d8d",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 46/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9284\n",
+      "Eval loss: 29.4838\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "551664d2376141d380efa9fc865b8972",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 47/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "1e223e40a9eb4c369dae3a40a987fed6",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 47/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9108\n",
+      "Eval loss: 29.2926\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5aca815525b5432498e6d18d9ea819bc",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 48/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "62c295fcbb9b4f8d8d47981b839dfc3b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 48/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9725\n",
+      "Eval loss: 29.2128\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d0d485bfef3741128e5c58c65035e4c3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 49/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8692f84b8aba43cc86c6125c45ac2bfe",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 49/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9662\n",
+      "Eval loss: 29.2485\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2dc89d21f6c24d2585695211f3b5adfb",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 50/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "aad3eceed51241ad88f7b09dc1fbc896",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 50/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.6761\n",
+      "Eval loss: 29.1596\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e86b7a1f47224177a0df7dba48a67f56",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 51/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "01700c482c7c49b688121d5729a26d93",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 51/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.8387\n",
+      "Eval loss: 29.2874\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f1b0aa8ae864491b88d57c5594243c48",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 52/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2f31461553b1450e9bbbb47ffdd3251f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 52/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9582\n",
+      "Eval loss: 29.0576\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "fe75bcb2a4fb4310a901b6086f8e8f1c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 53/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e5a1a7e6ba754dae8acbcf168f3f17ff",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 53/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.9945\n",
+      "Eval loss: 29.1711\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d8f88ac697c44f35bccd2cb26e745179",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 54/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7ed0536cb9c946b49de5b46baeb81f50",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 54/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.7572\n",
+      "Eval loss: 29.1158\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "41d050b4296b499ea19416d4589a4353",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 55/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "fa431fd05da44213bcc1b697d630d923",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 55/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.6628\n",
+      "Eval loss: 29.0065\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "169158d150e14d4fbcda23e4b9cbbccd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 56/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0697a8224d4a4089aee977bdf6ba97f2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 56/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.5971\n",
+      "Eval loss: 29.0604\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "59873742cf1d4ebba99a19600e3f0383",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 57/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "eb86f7f7887d42f2a6c7de0e161ea38e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 57/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.7197\n",
+      "Eval loss: 28.9971\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c145f1a8211840648b0138d23ad3fcee",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 58/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8033192266c846d79c5da197619d1d5f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 58/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.5317\n",
+      "Eval loss: 28.8572\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "9e356ca93ed24cb984b2e748b5c73e1b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 59/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "75978a63b45745aa90952b0f61b7c858",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 59/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.45\n",
+      "Eval loss: 28.8384\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "df4d4264ae4c489d9813eee00027de93",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 60/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0c57cff1a17f4898a3c0b3e3e0e52a9b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 60/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.5517\n",
+      "Eval loss: 28.6644\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3efe860826524478b41fe1acdfeccf3b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 61/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c47010982fa6424ab4010992f2146b6c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 61/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.3453\n",
+      "Eval loss: 28.7286\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "77865d5642a84a5ca8de449a3317c30f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 62/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e0199b2f07ea41ffae3973a3ef4230d0",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 62/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.3515\n",
+      "Eval loss: 28.7265\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7c1337a5b3f74902825a6d5a806189b4",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 63/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "1925050c8c7f46718a51bccc84355e77",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 63/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.6179\n",
+      "Eval loss: 28.8094\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "4926407b64e04e83abe857ac94283445",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 64/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "4f599fe256b34ff99b87376e6b856920",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 64/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.618\n",
+      "Eval loss: 28.7035\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "95810b0b2e264364a1646f5bcb3e2807",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 65/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e467276f5ef84ba4bc3f9e6f7e97d967",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 65/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.3259\n",
+      "Eval loss: 28.8633\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a76f775892524cedaa197021b4ee290b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 66/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a31a7c69e9a44799a9c570a89e71ef6e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 66/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.4899\n",
+      "Eval loss: 28.6412\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "829f46a27c484b6c80ce071c49948e19",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 67/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "ac7d09f005284b8fb0653d443bd39094",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 67/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.3958\n",
+      "Eval loss: 28.812\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "1c27bed86dca4c6893c5a9f3f33ce736",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 68/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "56d3a6e1a1d3410d88672e74559d0266",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 68/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.4267\n",
+      "Eval loss: 28.8075\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8b9d7783bd154a2abdeeac8180853605",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 69/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7af11a4ccba0406288d95d5f330b6f90",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 69/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.3737\n",
+      "Eval loss: 28.5985\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7b7c280b20ed4656950d55c4820db0c1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 70/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "04f8a1750e124496948aa73ab7f3b273",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 70/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.4183\n",
+      "Eval loss: 28.5121\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "6de5fdc09af547cdba07144d60bdf8a3",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 71/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d4a62abfdda04ad6b186bfff9a0a7995",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 71/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.2056\n",
+      "Eval loss: 28.6505\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "eee32eee264240a59b5738aa48de3b8c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 72/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "682e8e9a8be540bb9020ca462d9e99ce",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 72/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.1596\n",
+      "Eval loss: 28.6393\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c332b5cf133d4ae4aebaa70609d8ce37",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 73/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "cc9a037b110943e29acc7bfb2d7fac7a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 73/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.3922\n",
+      "Eval loss: 28.3878\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5399b94a4b7c4b7f8c1174412084d704",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 74/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "6e6fc46dee314ab085e6767084f70857",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 74/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.4299\n",
+      "Eval loss: 28.5715\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "640ebe21362c4199b643c467851d8dbf",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 75/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a14486955a214b22940c1ae9e2a629d8",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 75/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.2898\n",
+      "Eval loss: 28.7902\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "86fc2ebad25f4ee39624952d048df45c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 76/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2c58e75dc6f6478caaa49574d9be193a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 76/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.4332\n",
+      "Eval loss: 28.3917\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "efb86c1d12d54bd18ed5f621a43992c9",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 77/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "deaa3d5c00b64addb7c223f3b9e51f4f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 77/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.9395\n",
+      "Eval loss: 28.7671\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5fd0c3776b424e2484b662e7d79f683c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 78/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7b7dc14a836842679f6e9fd8ff98db58",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 78/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.1894\n",
+      "Eval loss: 28.6109\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "ff0688ec7b704a80ae29d4efca9c13c2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 79/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c4b83b0bd2a8480c968ca5616ddd4957",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 79/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.1998\n",
+      "Eval loss: 28.5882\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "3206cc2662e34d9aa98b1431cd6f22f0",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 80/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8a779d8ae28a4dcfad3a8c97a802ec33",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 80/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.9539\n",
+      "Eval loss: 28.6687\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "95a43389c27245f8b6cc041ef0982518",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 81/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "05a75bc82e2c4202baaa95ef2d3ca88c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 81/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.8146\n",
+      "Eval loss: 28.4852\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f7f28714242f419298c01964fa3a6f9e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 82/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5536681dcbe14a44b3c486560e680433",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 82/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.093\n",
+      "Eval loss: 28.4506\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "b5a78c2a386a44b8a702a6e486cd028a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 83/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e7d2329c96154277a4104eb7c40f37fd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 83/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.1327\n",
+      "Eval loss: 28.3635\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "bc25c0dc7aeb483ba00c3835f70a11ad",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 84/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5711268bced24520a2fadfc031869e0a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 84/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.1589\n",
+      "Eval loss: 28.4366\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5f2417a7805d47348ce3868048303204",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 85/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "dffe793c587c414b90e74b5fae0171f2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 85/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.1564\n",
+      "Eval loss: 28.4373\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "65b305778db94887b115df05fb8a66e2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 86/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "807d8da0ac4a4d038177a2c0511562f7",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 86/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.2252\n",
+      "Eval loss: 28.3446\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a14fd1a5607d40eba8c35343cd51018a",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 87/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a3ee84fe52314ea69527164fe83e1ba1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 87/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.0373\n",
+      "Eval loss: 28.4645\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8e8a82987c6e4ec398d1df89b873618f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 88/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "76718ed957584b78b223efe61d168075",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 88/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.9789\n",
+      "Eval loss: 28.3844\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e505397cc0d4482d94dd6551bb0c67e1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 89/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2f8252641397454db7e5e297cc08821e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 89/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.0804\n",
+      "Eval loss: 28.396\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7c16c34ecf564a04bd94e60442b4ed03",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 90/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "1f4e85d78fd04fc19a18a25bbf3a5cf1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 90/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.0642\n",
+      "Eval loss: 28.6572\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "7118d54002dc48e198e38ef41838f2ba",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 91/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "6ec3966cd54142c99ac67ca93ef5a142",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 91/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.036\n",
+      "Eval loss: 28.3789\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f4a6240163944caa824a90ec1ce50805",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 92/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "286d160165e44f5ea501ce6a176d6a5f",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 92/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.0854\n",
+      "Eval loss: 28.2493\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "948dbc673e9a4ea290f56f641de39937",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 93/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "ed7776aed0e740ecb933ec050b520856",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 93/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.9204\n",
+      "Eval loss: 28.4381\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f04bb312a181499b82f7a81851f0d5e6",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 94/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "bdad1f06852c4332877e0e8f896a99d6",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 94/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.913\n",
+      "Eval loss: 28.2786\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "50c339da0edb4aa2a2cf9e22fa31e997",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 95/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "6c32b0f2c70e41b3a8a20d80e853d4fd",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 95/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.8294\n",
+      "Eval loss: 28.234\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "43a4bafbe8be4179891b59a11a66ae7c",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 96/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "38fb1620e29c4d498c1f6fa4000428c1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 96/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.8744\n",
+      "Eval loss: 28.3384\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2961d5ea86084d5a9fb76dde06f95cf2",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 97/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "c5b20c9e60af41d9a3adde22a60658de",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 97/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.8647\n",
+      "Eval loss: 28.6488\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f50b7f265a7f4358b4e0e845387c505e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 98/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8980ec435cbe49249418489d210bf705",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 98/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 33.0428\n",
+      "Eval loss: 28.2822\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "5462badb4cf3439b8b57a1a809b292d7",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 99/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "4c941bd296e64026aedd7ca23663b908",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 99/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.8344\n",
+      "Eval loss: 28.535\n",
+      "--------------------------------------------------------------------------\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a5ad8b37d7cb4be196ba7f6a4b5ecd26",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Training of epoch 100/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a69cfed8a877428aa6a36870b2b39467",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Eval of epoch 100/100:   0%|          | 0/157 [00:00<?, ?batch/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "--------------------------------------------------------------------------\n",
+      "Train loss: 32.6901\n",
+      "Eval loss: 28.3918\n",
+      "--------------------------------------------------------------------------\n",
+      "Training ended!\n",
+      "Saved final model in my_model/HRQVAE_training_2024-04-05_09-25-49/final_model\n"
+     ]
+    }
+   ],
+   "source": [
+    "pipeline(\n",
+    "    train_data=train_dataset,\n",
+    "    eval_data=eval_dataset\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "from pythae.models import AutoModel"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "last_training = sorted(os.listdir('my_model'))[-1]\n",
+    "trained_model = AutoModel.load_from_folder(os.path.join('my_model', last_training, 'final_model')).to(device)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Visualizing reconstructions"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "reconstructions = trained_model.reconstruct(eval_dataset[:25].to(device)).detach().cpu()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 1000x1000 with 25 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "\n",
+    "# show reconstructions\n",
+    "fig, axes = plt.subplots(nrows=5, ncols=5, figsize=(10, 10))\n",
+    "\n",
+    "for i in range(5):\n",
+    "    for j in range(5):\n",
+    "        axes[i][j].imshow(reconstructions[i*5 + j].cpu().squeeze(0), cmap='gray')\n",
+    "        axes[i][j].axis('off')\n",
+    "plt.tight_layout(pad=0.)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 1000x1000 with 25 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "# show the true data\n",
+    "fig, axes = plt.subplots(nrows=5, ncols=5, figsize=(10, 10))\n",
+    "\n",
+    "for i in range(5):\n",
+    "    for j in range(5):\n",
+    "        axes[i][j].imshow(eval_dataset[i*5 +j].cpu().squeeze(0), cmap='gray')\n",
+    "        axes[i][j].axis('off')\n",
+    "plt.tight_layout(pad=0.)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Visualizing interpolations"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "interpolations = trained_model.interpolate(eval_dataset[:5].to(device), eval_dataset[5:10].to(device), granularity=10).detach().cpu()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA/sAAAIHCAYAAADAX0zsAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/H5lhTAAAACXBIWXMAAA9hAAAPYQGoP6dpAADI+ElEQVR4nO39Z7Mt13WmiQ6JEknQAIQnHOG9dyQBeqMiW6qS6aqOjopoE/3PqqM7ShHVFdV1o0KqK3oCIAgSjvDeHHgPgqA30v1ylXrnc84ea6+NbXLleZ5PM2OuszLXfKfLs8c7xx/90z/90z+ViIiIiIiIiCyGPz7oBxARERERERGR3cWXfREREREREZGF4cu+iIiIiIiIyMLwZV9ERERERERkYfiyLyIiIiIiIrIwfNkXERERERERWRi+7IuIiIiIiIgsDF/2RURERERERBaGL/siIiIiIiIiC+NPtvvBP/qjP9rL5xAR2TU4X/3TP/3TAT2JVKnH3FCPg2XVfko9dods565NqQev//Ef/3F3H+woRT3mxR//8b/8vXedOadbP5y7ds529WD7/+EPf1j93Tt/LBERERERERGZI77si4iIiIiIiCwMX/ZFREREREREFsa2PfsiIpuCvrF5oR4i/wLHg2ci7Q3bnXecn/YH9ZgXu3X2gXrtDnt59oF/2RcRERERERFZGL7si4iIiIiIiCwMw/hFREREROTA0dYyL9Rjf9hLO4R/2RcRERERERFZGL7si4iIiIiIiCwMX/ZFREREREREFsYiPfvr+EvSI8F/ZzqJ3aHTo2tj9dgbsl3XaVP12BvUYxlskh5zfrbdYpP0OBpQj3kxZz2OxtSUO9VjP3Q8GvVYh52+4+yUnXynf9kXERERERERWRi+7IuIiIiIiIgsjH0P489wB4Y+/PEfj//38Cd/8idHLFdVfeQjH5nKJ5xwwlCX1x/72MeGuj/84Q/D9a9//eup/O677w51b7zxxlR+8803h7pf/epXU/n3v/99bSqdHrxODd73vvcNdcccc8xU/uhHPzrUpQYf+tCH2nv85je/mcrZxlVVP/vZz6byT3/606EuP0uN9yM8bbdCqbpwoHX0eP/73z+V2eYf/vCHp/IHPvCB9h7/+I//OJVTm6qxzTl2clxxfCxVj9SAc9mf/umfTuXUpmrUgPNc/ruqUQ/289/97ndT+Re/+MVQl9rl547EXuizH3qQ1KPTim2+3Tp+L39TN3byetX4WIoeOSa6tZ9jZ7t1R7reCrb5b3/72y3r9kOPvWCVNl19p1Vevxc98ns4l6UGS9FjFdttq66O37FqT7fV/dmmqcGq/dVS9Fhn7d+qblX7b1cPkhrkOnOk66NBj53+u3X06HTNNt4PPXbSFv5lX0RERERERGRh+LIvIiIiIiIisjB82RcRERERERFZGAfq2ad3ldfpvb/44ouHuhtvvHEqf+ELXxjqTj/99Kl84oknDnX0XL7++utT+amnnhrqvv3tb0/l73znO0PdCy+8MJXTS151MJ7xnZJ60B/M6+OPP34qf+ITnxjqrrzyyqn8+c9/fqg744wzjliuOtwT++qrr07lJ598cqi7/fbbp/IPfvCDoS71eOedd4Y66kEPzW6wWxqnHmwbXh977LFT+aSTThrqcrx86lOfGupOO+20qXzhhRcOdRyDr7322lR+5plnhrp77713Kt91111D3YsvvjiV33777aGu82DOeays8m/nOSLHHXfcUHfmmWdO5RwrVeO4uuqqq4Y66pF9O9u4quqxxx6byvfff/9Q98orr0xlnj9CPXJ8zFkPtj89jnlWBc+tyLUltakazx+55JJLtqyrGtuO54g8//zzU/npp58e6nLdWYoeq/zaeTYF+3WeI8IzeHIdyrmrahxzhG2V7fzSSy8NdW+99dYRy1WbpUfn3yapAee21IptnOOOZyLxs/kMHK8///nPpzLbPMcS1/NN1WOdM5GoXWrFc3bys5znOF9tdb+qcZ/EM3jyDBjWLUUPst0zeNiO3Z6a817S7fd4flVe5zg6Epuix6q67uyW1IrzzFafO9J1tk93LhnP4MkzX375y19uef+q3dHD1HsiIiIiIiIi4su+iIiIiIiIyNLY8zD+LhSDYSsZzlpVdf7550/lG264Yaj7y7/8yyN+rmoMnWHIE8MfMqyGKeMyVCbT8FWNofuZZqzq8NRWGR510GlK1tGDocgZun/dddcNdX/zN38zlWm5yHZlmBlJ2wVDNjOshqFknR4ZYlM1hp3NTY8ujJ8hkh//+Men8uWXXz7U/fVf//VUZtj4KaecMpWpMcOjMlyJuqaNgH0+9WAIGkOgUo/dSlOyW6nEujBU9uW0UnBO+vrXvz6Vr7nmmqHunHPOmcqpTdXhYX/Zzs8+++xQd9ttt235rFnXpeWrWi/N0nbZLT3yuks3WTWGGKe1q2q0tlxxxRVDXYbuM8SfYzD7K20uaTXKsVJVdccdd0zlpejRhbpWjW3AUP3U4Nxzzx3qLr300ql89tlnD3UMI89nYKh+tvmjjz461N19991TmfMV15NuvtqpRWw/9FjHdnTBBRdMZa7D55133lTOuavqcD1yTHIPldYv2lweeOCBqczxQD3WSUO2XfZiH7DK5pJWFu5Fs9/Tmpr7ANorOV9lKDJtR2kDO3To0JZ1XOs5XrIPbpIeHEu5vrMdzzrrrKnMuT2vaVvl+0hes1+//PLLUzltqlWj5Zg2Cn7PpurB6w9+8INTmXrkHMV9Wa5D3F91axbX2pyHqEda9tg264yPvbS0+pd9ERERERERkYXhy76IiIiIiIjIwvBlX0RERERERGRh7HvqvWRVKpL0T9CHkX6WTGNU1aeZokcivbX0O6X/jF609OWsSp2Uv2tuqS/yeTq/ctXov+P5CunRZsq8TKuzSo/03tArmD5PegxTD3oDu342Nz2SLjVP1Tgm2FapHT3BDz300FSmb499N/s92/zkk0+eyvSt5Wep+Sqv3JzI/rHKA5t6pKevamwrpknM8yc4l9Gv2qX8SV86PZ95zfFBD2ZCbeY0XlaNj/TAUo88f4LnweS/Y9tkatCqw8dLkj5bepJzXK1aP9I7OGc9Vnn2sw/yTItrr712KjOVbrYVPa/UJ685X+VZDLlHqBrXs1WpQunlnCur9Mj2YQrWPNPipptuGupy/qJ3lm2Tbcc9XK79TB+Wa/2q1MZ7kUp3p6xzhgLPGMk1lHpkSlbOV6eeeupUZp/v5ge2a+rKPp9rBrXi7+CadZC8lzNGct7Jcyqqqi677LKpzPOSsh1z3a/q9eA+LdMef//73x/qco7iv+Pv4JlVB8k6Z/Dwd+ScQD3yjBHWpY48K4Ztk3sIPmu+xzD1d6bbo0d/nfGxl+u5f9kXERERERERWRi+7IuIiIiIiIgsDF/2RURERERERBbGnnv2uzzm9F7RC5S+fObFzRy6nWefHkt6WW+++eap/LWvfW2oS08P/Wbp51jlo9zL3Inr0t2f3jd6T9Kzwjys6X2hji+++OJUznyUVYd7/m688cap/PnPf36oS08P9UitlqoHvT7p1aLv9/HHH5/K3djh+QqZy7Sq6sorr5zKqU1Vf6ZG58PfrVyvewH1yH7Fus47ys+mP/Kee+4Z6nK+olZsx/Sjpe+8avQzc57LOn7nfuZ6XZdOj1WfTf8dz5RIX356I6vGMxSeeOKJoY5zYua8pj8zxxLPfMmxw9/UjY8567HqDJ7M8Z3nS1Qd7uFPcn1/5plnhjqeOZK6cm3JOYr3z3FH7+hS9ODvSn93nmdQVXX99ddPZeamznMRHnjggaEu9whVo+ZcI3K8nHPOOUNdrkv07qY/tmpe8xXpPOLUI73eV1xxxVD3mc98ZirnWSBV4zrE/RXHR85JOVb4vewPzz777FSmB5lz4qbqwX6W/fO6664b6nIvdMYZZwx1qQf3ZdQjxwfX7JwT+Y6T+2++43CfuKl6sJ/l2RQ5P1WN44XrcLbHO++8M9TxOtcMfk/eI/cIVaMePFem02M/8S/7IiIiIiIiIgvDl30RERERERGRhbHvqfcyhIHpPX79618P15nu4+GHHx7qMuSG/y7TUjA1T4ZdVo2hUwxhztB0hirlNcM0GN47t9CZpLNVsO0y5IXhlBmKzHQWacegVgx5SQ0YVpXPyu/J0L6l6MHn7GwWDEXO9uG/S+0YxtSFQHUho9Qjr6kHx/2cw8y652EYZn6W8wXtEsn9998/lakjw88z7I/hnDl2OAbzeajVpurBsD/O39lWDFnN8EqGVt55551T+bnnnhvqOCdlyp+0vFRVffzjH5/KHIOpD/XgvLspenSpKKvG9HZMhZi/mWvL7bffPpW7cVQ1jgmmxMo69p2cozZZj4SpKBkmnP2TlqAcSy+88MJQd+utt07lRx55ZKhjP890xpk+rqrq3HPPncpdurBV89WcbRVJl9qtqurMM8+cymyrXJcZtp3jgzYwtlWOAVoFMt0f59LcQ3FtmbMeHav0SMsc0x1m6D6tK3fcccdUTvtD1eH7gk6PfB7uA3L+6tJdVs17vsowfobtM01e2hoyNWjVOD6YVvWuu+6aymkprjo8/WTOV7RupJ2Jz5bjhXNg9/6xn3r4l30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGHsu2c//Qydt6Rq9BPT95Bee/679GqedtppQx29N+mTyX9XNXrV6BXM8wRW/Y65+WSS1IPeRJJ60MeX/j+2R5eC6tJLL93ymh7DTCvD+6cHt/OQVc1bj+5MC3rfM/0HvUidnz59UumjrTrcS5ueZJ53kV61HA9Vox6d5/VI13Oi04PexUyFSO99esbpEU996JXkeEl9cszx+Zj+stNjU8dHl/qQn2V75Bkf9Pildt3YqRpT9WRqoKpROz5bjh32o00aHwl/I33x6eHnWMqx89RTTw11mUaUKajo80wPbO4RqkbPJf266d2kHps6PvicbKtsD/r7c7zcfffdQ11e53k8VYfPX3kPrh+Z/vDpp58e6nKt2+T5qoPniOQZBpz3U488M6Fq9IhTD2qe3m/ur3L+4v27feKc9eiehfMTf/Nll102lXmWUa4ft9xyy1D3ox/9aCpTD46zPOek84hzbcn5a5P3uwnXVupx9dVXT2XufVIPjo88g4fnXfCeOe/k/fg81CPnPe5L9OyLiIiIiIiIyJ7gy76IiIiIiIjIwtj3MP6E4SYkw1MYqpLhLww9zhQNGYZcVXXRRRcN1xlKlqGuVVUPPfTQVH7iiSeGukyhsalhM4TP3aX0YF2GRGXKkqoxHClTllSNWlWNemVoZ9UY6vfYY48NdUvRI591VQoPXic5PtjGGdrHUMJzzjlnuL7++uuP+GxVY/hzhtpWjXqsCreeM+ukQkwYHpbhe0yZl+GtnOc4XjLlDMMwM80PbUcZ5rYq7G9TWBXunqGODJ9M2I5p/aK1i6HhOT6oVVoA0oJUNYYLrmNHmBvds3EM5FzDMPqc62mdSA2oB0PDM2UZ570MtaQdIEP3lxoWyzY/7rjjtqzL8GNaxDq7H9Mt5n6Le6+0wHCtX4oeXWrKbt6hHSL3Pkyvl6mmc56vOjy1cYYfc2+cIf5s427vt0l6JFwTOLfnXoj7pAwNz9S5VePYoWWPmuf38j0mxyfnvRzbS9GDlpN8N6sa+2unx3333TfU5dqb1teqw9s1+wTXlpyvaDHowvjnYsvzL/siIiIiIiIiC8OXfREREREREZGF4cu+iIiIiIiIyMI4UM8+6bwN9DtlyrD0TVZVffnLX57K9MfSo5EpTei9yRRA9EJ1fulNZZUnOevpd8rUJOmbrBrTHZ555plDXfqSqsZ2PXTo0FCXvnCm0kpP2aZ4lFax6gyFrO+8eRdffPFQl2cqdOmpqkYPJs+tSO9gpmWsGj2XS9WDXtI8t4LevEwVwzQ++Vl6NenjSx8bfcc5PtLHWTV6xJeqB+er9CGzzXN8UMeu7hOf+MRwffnll09lpnLKFGVMTZnryaaemVA1ekfpEeeanXMU/ZDZ77nWZrsylS59xzfeeONUZnqm9D0zBVOuJ0vRg+1Pn2vuhThfZb/vUrSxjr78PGOEHtw8I2mpemTf5bkITHubnnmefZC68ntSV+7LeG5Fjg/uxRJ6zXNMch/CcT8n+GypB98FOF+kPpzbuzNw8p7p8z7SPTLVNNf61JXnZOS5JqvGx5zX+2xXzk9ca3OvyjbP+YO+/Pwsx053bgbP4Emffpeyb67zlX/ZFxEREREREVkYvuyLiIiIiIiILIxZhfEzNCLDT1iXIR0M98gw5RNOOGGoYzhOhsNkGFXVGHLDcOcMW2ZIyZzDZjrYxp0eDBfL8G+mb8swcurBsOUMH6MeqQFTX2SYLEOeNtVy0YWgVY16MDwp24qhfFdcccVUZjgtQ6kyhR71yDA32jHyeTKkv+rw8bIprApTzn7WpQjL0L2qMXSM44NhZkmGjlWNIYPUMZ+160dHup4rbH+OgZwH2I7ZzgwvzrmNobYMr81+T2tR9gGmFUoNqMdcwwBXQT04t+f46dIaMXwyQ2Y5z/A6w/o5Xp977rmp3PV56rGp6wf7HMdA9nPub3L+4Pdk+D3XBNpl0uZCrdJqxDUixwD12CSyD3I8cG5Jmwv1yDFx9dVXD3W5T+X8xL1xrj3UI8OfOZelPhzn1G7OZF9iW7Hv5pzEtsr5q0ttzH0AbcXXXHPNVKaFMscn9ci1f9X4yD44t7U9n416cAzk/MW5Pf8trV65D+Bcxs/m2KLNJfcX3Huto8dBMc+nEhEREREREZEd48u+iIiIiIiIyMLwZV9ERERERERkYczKs9+lUupSHfziF78Y6p599tmp/Morrwx19GzkPeiX/dKXvjSV6QNPjzg9S5vqgV1F513MNkifd9WYFoMpXegtTujL/+xnPzuVmeot70E/TZdCcM6ses7sk/yNeU090pvHtmKbp/+I/sxPfvKTU/nFF18c6p5//vkt79+dBTFnVo3rbMvuN3O+ynH1s5/9bKjj2Rh5TQ9u+s0efvjhoS7TJjK12aaOD85HnKPT55g+46rxHBGm6qGHP+l86JzL8uyYLFdV3XvvvVOZHr8uldXctMnn6dJDVY1zNlOtpSeWZ3pkSip6YOl7zrZkOtL0xNKrmZ+lHt2ZCnPWg+3PtTfnGuqR7UovberBOs5J2XbdGQL0S+c8t0qP/M1z1oPzLtNx5vhgWtXs95yfMt0h5yDqkfMXx0eOX+6FOz3o4U/mdv5IPg998NSjO8Mgz0m48MILh7rUg3smnlWS44za5TpArZJuj1A1zqdzW+u78cH0wTmfMaXhJZdcMpW5v8ozFTgeeE5CnqnAvXDOX927Ks8O4nXSpdPebfzLvoiIiIiIiMjC8GVfREREREREZGH4si8iIiIiIiKyMGbl2Sfpg6APIz1m991331CX+XT57+i5zFyK119//VCXHlh+z09+8pOpTC/cfvowdpNVnuT8XfzNL7zwwlR+4IEHhrrUgx5xev7Sj3bllVcOdVddddVUpr8qPbD082+qHoR+qzwngZ7L9GjTw3Xo0KEtv5O+sczvm76oqlGfPCejquruu++eyqv0mLMHNuGzdR7xp59+eqhLrx59+dl29GvTY5Z6pDewavShZz7lqqof/vCHU5leOHqk55yXt/Pn8pyEnCN4hkGXYzzPm6DPmOMj14/M8c465rtOj/Jbb7011HF8bEqed84lHPfPPPPMVL7//vuHuvRrss/n2kvvKv3CuX7keKgadabm6c/k2sLxsSnwublGPPjgg1OZecRznmGb5x6K5wLwHllPD2zqTA9ufpbfyTlyU+Yrnu2U46Fq3MfybIpsH+rRnXvEs0pyv8Xxmt/LPUN6z9c5Y2Ru5/N0Z1o8/vjjw/VDDz00lTnPZBvQs5+f7fbQVX3b5T04lvkes93vnBudHtxD5TXbPOd9jp3cF7Ad2Zdz3uEYyHbls/J7tlu3n+PDv+yLiIiIiIiILAxf9kVEREREREQWxqzC+LsQBoa/ZOhjhttUjaEaTGHC0IxMwcMUCeeee+5UzjRjVVXf+c53pnKGTFcdHt4759QwHV16CYbjv/zyy1OZv7/Tg22e4TgMnfof/of/YSp/5jOfGeq++93vTuWnnnpqqNtUPdYJAWNqmLROsC51fO2114a6zubyv/wv/8tQl6kpv/CFLwx16+gxt/Q8W7FKjwztYijwo48+OpVzrPB7GXbJkLSck/73//1/H+pyjvr85z8/1H3zm9+cypusR/Z5Pid/R9olGBKY/7YLg6XlguMjQ8X/t//tf9uy7lOf+tRQ9w//8A9TeR095jxfUQ+GOmYa3LTBVY0h/2zjtFLQGsD1I61eDNXPtf6KK64Y6jJMugsT5/Wc9WDIKlNbZb/jvijnGeqRcxItQeS6666bykw/maHhtLnweTaVLjUl1+W0tlC7XIfZ53P+WmXRuvbaa6fyZZddNtSlzrR1zDl9W0dnTaWtgmvvHXfcMZVpEcuUbWyPtEpQD94zQ7zTNlw17n8ZCp5za2eLrJqXPp0efKd46aWXhuvbb799KrNfZwpDWq67tMecE3Md6NqVbZzf073/VB3c+4d/2RcRERERERFZGL7si4iIiIiIiCwMX/ZFREREREREFsaBGqNWpYjIevq30/uSad+qqt59992pTK8NvS/pAbzrrruGuj/7sz+byukJqaq6/PLLp/K3vvWtoY4+kE1hnZQdbMdsZ6ZhS28a24b3zLMYMn1bVdWXv/zlqXzCCScMdZmKgx7DpejR6dOlvWKaxNSDvkHeI3222eerqj73uc9N5UwtV1V19tlnT2WmL+v6wJz8ZWTV+Oj8gHlWBc9JyPmq85tVjX5/pgq94YYbpjL1SL/yKj/sUvRI7xy995l+kqQfknpwHUotb7rppqEuU1UynVznEZ9zm3es+h057jkGch5iarH0PK7SI3XmOSI5Brh+demyNkmPfFbqQe9orgscD5nujmnxsq14hgLnllyXvvKVrwx16dnnOrSOJ3lToUc597HULtP0MYVh6sF5jnrk2Pr6178+1OVZAEx3mN/LftTpMeexw2djH8wU0vRh5x4z+zHhPoBtledfsD/kZ7mnzjOBeI9NSdVaNWrAfsWUtJk+mHvIPCeBZ/DwHTBh2+VelXX5rDxnJ/d3c9XDv+yLiIiIiIiILAxf9kVEREREREQWxqzC+Blal/VMN5IwxU+GbTA0hPfMz2bIctWYNoNhsRkOxVDCOYcudazSI6/5mzNUhXpk6CXDoUh+lqFkGXaYKfr4PKv02NQwZf6u1IOfzXZmWFGGjTN0jGRYE0NvM7SPoZ5bPeemke26anwkDN3KeYZjIPXg2CE5J1GPzq6Sz9PNs0e63hT43BkGuY51JDVgCCDvkRrQLpPfw3+Xeiyl/btUrVXjWsyUVNnOnOc6HdmX0y7BsNycB/nvch6c85qwDl37s557nwxT5t4r2456MGw80x/ys928l9rNJQz2vdKtCVVjH2SYcOpBm0u2OduYlsYzzjjjiPerGlPPcW1Ju0aXSmzudFY7WlJyL3rPPfcMdd34SD3Yxtwn5T6W/SHXE1qVO1vFJtGl3mOa7tSH60eOic6ixb0XLRhpkeFYynamrSK16sL/j3S9X2zuLlxEREREREREjogv+yIiIiIiIiILw5d9ERERERERkYWx75799CPSW8EUXemLOe6447b8HvowOp8t/YB5j3POOWeoO+WUU6Yy/SSZ+mKTWUeP9MUwvUX6lKhH522mxy9TaJx55plDXabbYwom+mWTTUpt1elBb1he05u33fFBeI/UOccD65jyJ6831YNM+DvYd/Oa2qXfqzu3gn2TeqQnmSmY8v7ph60a/YDrpBDcJLozFfibOo92epnps6VfNjXgnJhrDf2YXYq0TWVVv8l2pc+0O98goR8y14uqcY1gXfYHesRz7CxlfNCz351p0Z0pwT1TwrmMnuQcH1yjEq4f3T3n4oHdDvlsnEu656Z22e+7tll1j9SHa0uOiXX2t3Nuf7KOHtm32c/zTIl1ziTivJfzDnVNPZiGLvvHqrNSkrlp1aXe4+/INbs722gdPfgewfUlSQ8/zwzIZ+dzzyVVqH/ZFxEREREREVkYvuyLiIiIiIiILIxth/HvNBS6C61kSCTTIGS6u3PPPXeoyzAahiNl2hCGUDC071Of+tRU/tf/+l9vef/nnntuqHv55ZenMlM08J57ETqzW6HpqQfD9hmSd/zxx0/l8847b6jL8BeG66UerOM9Pv3pT0/lP/uzPxvq0srxxBNPDHUZdsYQn7mE0WyH1IP9mm2X4+Xss8/e8jszpU7V2Fb8zgwTr6r6zGc+M5U/97nPDXWp3TPPPDPUZRg5Q+A2KQyzSzdJfXJuOfXUU4e6HFsMHcsQPY5B2pdSgxtvvHGoyzByhv3lmNjkVFadzaWzIXGeyX7P9kh9OB5yDqwa56srrrhiqMv1rUsDN+f+v4rOMkdy/NACw36fZIgk1+9M7VY1rufnn3/+UJf3ZLhml75qqTaL7f6ubj/DOZHj5corr5zKTJeb426dPdMm2fI6OptF1x5dOmnOgdxjp1WVc2KmNltnjeA9N2W/tU74+zrp7VKPLl1y1Th/cQ7MNXudsbup44PPyT6Yv6sLtyfdGkVdjznmmKlMrWjFS7o02J0e+6mNf9kXERERERERWRi+7IuIiIiIiIgsDF/2RURERERERBbGtj37u+UtSA8LUxXRn5oeL/oh0xv24osvDnXp56aH7KKLLhqu//Iv/3IqX3755UNd+kJuu+22oe6uu+6ayvTg7ocndqd6dGcorPJvn3zyyVP5wgsv3LLu9ddfH+oee+yxLb/zggsuGK7/zb/5N1M5/X5Vox4/+tGPhrqHHnpoKvMMhXVS3uw3nb+HHi76VXO80LOfYye9eFVVTz755FSmb496/MVf/MVUvv7664e61OMnP/nJUPfUU09NZZ6hsKl60KPfpabM8VA1ppGk3+zQoUNTmXpceumlw3WeK3LdddcNddnOOR6qxjNGVvmV56QH6fyQ9IHn3MZ25ZkKSabV4ZzIdehrX/vaVOa5MplOjHrkmQocD3MeHx2rvIpJl1KSumbKJepxzTXXDNd5jkWm4asazyp5/PHHh7o8U2EdL++ceS8pBLs0otk/O0941Th/8bPPP//8VH7llVeGupyjVnmQN2V8rKLz8na+4/wsteI8l/sE9uucrzqPetcfjvTsm0L33DvVg/+O6XJzn9alnuNeI+dPzpf7cWbYQbDOOR5Jtgf7Lt9HzjjjjKncpUjmfiLntrnq4V/2RURERERERBaGL/siIiIiIiIiC8OXfREREREREZGFsW3P/k7pvBSrvKLpb6H36Nprr53K9F3kPemtSO9s1egBfOGFF4a6b3zjG1P5//1//9+h7umnn57K6+R8nDOrvKPpGef5Cunnpi8p/X/UIz0yVf1ZDN/73vem8t///d8PdZ0em+pZYvvT+5O+LfohL7vssqnM3OB5VgZ15PhI3ytzt99+++1T+bvf/e5Q98wzz0xlevY3SY981lXexPRq0Rt23nnnTWX6+f/qr/5qKjNveHcWA/O+3nnnnVP5jjvuGOqeffbZqZwe6KrN8sB2fkhepwbU4+Mf//hU5tpy4oknTmXqwbNKLr744qlMPfIciwceeGCoyzMUlqpHB/cFxx577FTmnJTX1IPnVuR44Vk6jzzyyFTO9aKq6t13353Kq/Ylc9YjWfWcne+4W1uyjms9zxhJ7ahHevZ/+tOfDnU5125Kex+JnfqMu7mMnuCEe2Gu57kXyz5fNZ5VwrUuNV91FsSc2ake3We5tuQ9eM5Sri1VYzvnuSG85j3Sw8+6TWIdPfKznVZdHc984bkuOfdzfHD+2up7OT65nuTz7efctrm9RERERERERESOiC/7IiIiIiIiIgtjz8P4GaaQIQ0M72WYxHPPPTeVTznllKEuw5QzlLJqDJVhOqZML1JVde+9907l/+f/+X+GultvvXUqZzq/qjHEZlNT81SNYUQMJ2UKu5deemkqZ1hw1ahHlqvGUP1VemQo8n/7b/9tqEs9mMoqv2eTU8Hks9OOQD0ylVSGBbOO6aky/SRTulCPbOfvfOc7Q13qkakoq8aQwE1K7dal3WIqFs5fGU6Xv7+q6rXXXpvKn/70p4e6tMAwzIyh4WmPuPvuu4e61IOpQtOCwd+xSfNX6rNOv+LYyX6elrCqqs997nNTmWGYnL9ynDFU/5ZbbpnK1CrTYXYpl+bOTm0u7IM512VIf9WoD8PG045RNer86KOPDnU//vGPp3KXem+T14/tpm+rGuerLl0w0yRmm9NyQX0S2vIyBSzT9eaYmEvqqvfKKjvIdsN7qeMxxxwzlRnGz/TW+b20TuQawTkp+8qmtv8qOj3YB7vQ+c4Cw/Wk0yPXKIaGc5+wRLp+to4e2XZcv7n/zX6fa3vVGNbP++U4m6utYp5PJSIiIiIiIiI7xpd9ERERERERkYXhy76IiIiIiIjIwth3z35Cjzh9W+mjYxqE9H8xHVL6lujtePjhh4fr9I0dOnRoqMszBPism+SrTDo96BGn7zj9w/Qrp3ZM+3XaaadNZXo1eRbCU089NZWpR3qY6MHd1FQ9nU+MvjmmZkm97r///qEuxwvbOMcL75/joWr0iPOchuwf6UHms2+SHqTzJHMM5G9m381/y76bKcHoKcv2rxrTVTFVaJ7bwLk0+8qmzl1Vox6rvLx5/eabbw51jz322FRmm6c+9I8z/WSOgVdffXWoy/FCrXI92WQ9ku68C8KzKPJMC+qR/lSmEnvllVe2/CzvkeePUKuljI916Oa2HAPUI9chpnWl7zjXJa4tqcdSU+/tFP7mbA/qkR7h9O/z31WN6wf3F1263O7ZyEGlFttPco6gRzuv6dmnHvkew8/mfned948uFeBS9ejS8uU1zx/hGMh1mnrk9/A9pjvzZC/02En6S/+yLyIiIiIiIrIwfNkXERERERERWRh/9E/bjCPYSdjAunThFyRDmbpUB6vS6Gw3vcNSw186uvZnXerBNCFdaHrXrl1Y6Do6LoWd6sGwv9TnveiRY6sLqd4Pbfj7D6I/5DN0ejDdS4aWUY/uHvyNGYrMeS/12Y8w5bnpQVIPhuulPnxurjWc65IMI2fYX17vR2qxOeiRdGnguvHBuYxhyxnGz3tk+DmtNBm2TK2ORj3yuls/mAKMqfhOOumkLT+bFhjaYzKEmSHlezFe5qYHyefjnNOF8Wf7V1Wde+65U/kjH/nIUJdWPKbyTdsL7YR7kVp3k/TowvjZ5zMteFXVJZdcMpWZtjJD95m2Mm2CtMB089c67djtZ+ZsderWllV6nH/++VOZeuS6lLazqtGixLmMFoydWpQ6PTobwT/jX/ZFREREREREFoYv+yIiIiIiIiILw5d9ERERERERkYUxK8/+btE969y8P9tl7h6mZD/6ykH//k3WY6fjo/vNB/37N1mP9JStk9ao+8374QPvoI9xUz1+6/zb7jcf9JkWS9Fj1VyWn6W3OX2N3ZkW+3EezCbpQbJdV42dPG+Baa/SW0zfd6eVeozks/N38LyF9B3zrJLUoztjZJVHfzf06cbu3FlHj/SQd2n62ObpA1+VBm439OBzrzpbaE506wfnpNSAZ8fk93RpmJnCvBsvO9WGz817Hgn/si8iIiIiIiKyMHzZFxEREREREVkYiwzjl3nRpQvb7r9b99/K1uxUj00OdZwz2a5s024MdKGOBz12Dvr+74Wd6sHQui7UUT22zzp6JAzDzFDHddrjaEi9tw6dHlwj8ncx7VWGIq9jEVOPkZ3qwTR9TBG2FV1K5PfCpqZ6I3uhxzr9s2sr/rtuL3g06NGtH5yv0i6xjh6dBWW39DD1noiIiIiIiMhRiC/7IiIiIiIiIgvDMH4RERGRA2KTw8iXiHrMiy78XW32H57O32Wekd1nJ7YK/7IvIiIiIiIisjB82RcRERERERFZGL7si4iIiIiIiCyMP1n9ERERERHZC/Qdi2wfx8vBsh/pMGV38S/7IiIiIiIiIgvDl30RERERERGRhWEYv4iIiIiIiAyYen3z8S/7IiIiIiIiIgvDl30RERERERGRheHLvoiIiIiIiMjCOOo8++t4T0wnsfeox/6zW/4r9dgd1GNedHqwjdf5rOwM9dhc1ENk83Gs7g85X3ZtvhM9/Mu+iIiIiIiIyMLwZV9ERERERERkYcwqjJ8hX3/8x3+8Zd373ve+qfwnfzL+jA9/+MNT+U//9E+HuuOOO264/tCHPjSVf/vb3w51v/jFL6by22+/vWUd/x3Z1BCY3dIj25h6HHvsscP1Bz7wgan8m9/8Zqj79a9/PZV/9rOfDXW//OUvpzL1YPtvqh5ku3qwzY855pgt63LsVPV6ZDv//Oc/H+p+9atfHfFzVYe3/z/+4z/WEtipHh/84AenMsfORz7ykeE663/3u98NdalPtj+v+e/Y/qnPJo+V1GCd+Sr7fH6uahw7/Ld/+MMfhrrs9xw7eU091hkfm6RPF9KdY4dtvo4eObbYrqlHV0cdOz2WurakVqlN1djnOXZyLquqev/73z+Vc/2uGtuZbf773/9+Kq9aHzo9tru2bFJqsW5fRq241qRe2cZVfd/NdlzVpt360Wm1FMtHt+502pH8zev0T352p3ps9zs3ma5d16lbR/Nkp3p0/WbLf7P2vxARERERERGRWePLvoiIiIiIiMjC8GVfREREREREZGEcqGef/jv6ENLrTS/YOeecM5Wvvvrqoe7yyy+fyjfddNNQd/LJJ2/5DG+88cZQ99Of/nQq33nnnUPdf/tv/20qP/TQQ0Nd+vmr1vOfHSRsf16nPzK9eFVVp5122lTO9q+quuSSS6byZz/72aHu9NNPH67TU/baa68Nde++++5Uvu+++4a6b3zjG1P5wQcf3PLfVY1+zc6vfNCs8nell5XevOznF1544VB3/vnnT2XqccYZZwzXH/3oR6fyq6++OtTluQkcA7feeuuWddQj/bLr+Aj3m1VerBwT9LJ+7GMfm8pnn332UJdz2Y033jjUnXXWWcP1KaecMpVfeeWVoS7nnccee2you/vuu6fyKj06P/mc5q9VPsbUgGtNnk3BNSHHQM5dVb12HB/Zl5988smh7oknnpjKjzzyyFDH8y9SD55/kV7nOY2Vqn68UI8cO7nuV1WdeuqpUznbu6rqzDPPHK5zrnv99deHuuzLL7300lD38ssvT+Wnn356qMvzYKr6sxi6tWVu7PRMizxnh2Pn4x//+HB93nnnTeVs46qxfdjncy7juOIYyDbnuQCpzyadfbGOJzjXfp65k+t3VdVJJ500ldnmqTPbI+eZd955p332/Gx37tKqs0q2W3fQdFpx7HTn9fA35n6P39Od3cO5Nb+3O+eH89xW37Gd67nCPXQ373W6Usdcv1adm5Hf051Vwn1ZtjH/3XbwL/siIiIiIiIiC8OXfREREREREZGFsedh/F0ocoYhVx2eZipDvBmO/1d/9VdTmaGWJ5xwwlRm+D+fJ0PCmMYnwweZsi/DKBhW9uyzzw7XGe7MsKaDDn/pQo7YdhlCzNDw1OOGG24Y6j7xiU9MZabaY8hRtiVD0LLt8lmqDu9LCUNoM40iU5TtJDxmN+lSHtE6kX0y+2pV1Ze+9KWp/MUvfnGou+KKK6YywzDZjqlHjquqMdSS4eb5WT53hjBXVb311ltTubNcHDRdyFfVOH+w71577bVTmXrk3EYbRYb/V43zRVpnqsa2u+iii4a6DD9nqCf1ePPNN6cyU47mGDzouYt04XOcyzL8mzaw1IfzHMPGc86kDSzDZFP/qqrHH398Kn/3u98d6rrxwdD0DL2c09x1pOscL5wTcr6gtSutLZdeeulQl5akqnF9oZ2OId7JCy+8MJV/9KMfDXWHDh0arjOMOf9d1Wj968KU55DKqlv7u7ksxwBtFdQnw8avvPLKoS7D6tkfss/TVtGNAe690gLAMOW8/9z04Fy23VS63AekjaKq6vjjjz/id1aNcyTXiOzL3ZpQNc5Dzz333FCX44XzZYYwU485z23dPMd25H4r33k4BnIu4/sHx2vC/pHXO7UvsT/OaV9GOnsf+zwtY501NudB7oVTZ77HUrscrxxLubbce++9Q12uLabeExERERERERFf9kVERERERESWhi/7IiIiIiIiIgtjzz379Bak14Q+SvpZPvOZz0zlf/fv/t1Qlz5L+jDS9/DjH/94qEsvWNXhvqHk3HPPncr0Qn3yk5+cys8///xQR29Y+sLpddlvL1J3hgJ9QPSzpD/yX//rfz3UffnLX57KJ5544lCXXkmmMKRHOz1F9NOnn5ne2euvv34q09NHT1l+b+c3O2i6sVM1jpdPf/rTQ93Xvva1qXzNNdcMdfkb0ztcdbjPNT136RmqGscEPfupR+eVrRpTJVLz9PEdtEd8VSrE9G3RS/y5z31uKqc2VaMXjB4ueoJTD85l6Y+l9/+6666byvwdHOeZpo9zWc5fcxorVX1KKp4Vkl586pHeYqbrYrrD1IMpqVKPbv2gx++OO+4YrnMNox45f81ND5LrdLf233zzzUNdpgelV5JzS44XphbL8y94LkCe+0MfOtO8pj7UI687X+tBe8KrxvmrS015wQUXDHVXXXXVVKZHn/fIuY1apZeV60f2B/q+eZ3jg2Mg17Mutdjc6NK68hyXTHfI80c4XrJdOQZTD64fOX91Z5NUjalE84yAqnE951q3KSmqq/ozFHLs5DtE1eHtmtpxjUit+O+yjmeGse/k+wnTX95///1TmWtb7sXmrkfOZd0+gGOHZ39lu3KNSA2oa77zcK/Bs+Dy/YT77WeeeWYqv/jii0Ndnv3GFNXbwb/si4iIiIiIiCwMX/ZFREREREREFoYv+yIiIiIiIiILY088+12+0PRP0MtAf0/6IuifSD/co48+OtT93d/93VSmZ58+iPRx0Tv51a9+dSr/z//z/zzUZe54eqL5PJnrlb61g/bsp/eInnDmCM283szjnZ4y5uu85ZZbpjJzGPOz6eFn/8h85P/+3//7oS79NPStUY/0MPHMgGyf/fBVdvlbu7FTNXqD0nNaNY4XnmHwk5/8ZCpzfLCt0lfH/pG+/H/7b//tUJe+1yuuuGKo4zkB6VOiD73LmXoQ+nR16YFkfvacI6hjtjn1eOCBB4br1JL3z37/l3/5l0NdemJ5TsVTTz01XD/55JNT+bXXXhvquvzCB32mAsk5ietHeuYvvvjioS7n6DxPomr0OFaNHnHO5anHV77ylaGOXsGtvrNqHC8cg53vuvNZ7oVW/M51vJM5R9Czn/sArt/UJ/Os83nyLAZ6LtOryTamRzmfnW3eraedz/Igxk72Hfq3c62/9tprh7o8Syk9+VWHt1WewcP+mGOA4yHv3+V8rzo8j3WSv5F5zDkPJgetR3eGAs+UyH1RnqdQdbg+XS778847bypzvsy9Oes4X6UvvDujin2uO+NiJx7l90q31qU+fG+5/PLLp/KXvvSloe6yyy4brnMuYXukn5/zZb6r0PfNs2PyHSfnx6rRB86ze/KMJv7+uZ0Pk8/HdszzvXLdrxrnsqpxXcgzd6pGPfiumO8q1INnKiQ8tyL3d5yvOCesi3/ZFxEREREREVkYvuyLiIiIiIiILIw9CePPECiGQ2U4DsMSGHKU9QyFePPNN6fyN77xjaHum9/85lRm2ArTr+TzsS5DvJlOIcM4GNbE6wyLPWioR4bWMeyQeuQ1Q5GzrZg66lvf+tZUfvjhh4c6hhzl8zB8L/Xhs2UIEtPN8JrhMQdJF/rapXarGtuAvzFh//ve9743lWmrYBhmhmsxzKuzwGQoFZ+b9owc51345EGHiVMPhovlHJEhX1VjGB5T3PzgBz+Yyt///veHukOHDg3XOX9SjwwRZJtnn+e8y3ZNzRk+edAaJAwt5PyV/YxpXTM1InXN1FG33nrrUMcw/gwFZptnSC21ynsynDhDK6tGWwFtYDlfUpv91mpVasocLwx9TStFWuSqxvZ47LHHhjraXjL0lHpkGD/XgOzzTDHKMOWcI/nZ3KcwbP2gx05nq+B6mjY9hrrm/ob9kTaLDOnuwlC5v8t0bpwDMzVo1binYBrk/J65zWXdeOHakjasf/Wv/tVQl3YhpiTmnjZtckxzm/Mn92WpD/cT3ENkqkqO187S2s1lB0Hqw/1uzl+0HWWa3S984QtDHfdp+Zs57yfcl+W8Q1skU4U+8cQTU5na5dxGC+XcUiF2e+McL7TlpYXub/7mb4Y62i1TZ67LeU9qleH42d5Vh+uT8xk/m7YnzmXvNc2uf9kXERERERERWRi+7IuIiIiIiIgsDF/2RURERERERBbGnnj2E3o90mtA30P6u6pGn0p6GapGvw/9LOnnpy+Jvq30kTHVQno96CNM6NtjKqv0s8/Bi5Tk89CjwrR06X9jSq5MN5Htz88yLQXTraQe9NlmGpNME1M1+nnYH+iLSb3m5qvM+9OXQ32y7ei3Si9W+uSqRs94anqke3R6pDctPdD8dxy7PLch9Zqbr7I7f4T65DXbMTXgnJRacezQc5m+ylNPPXWou+GGG6by2WefPdSl34znmDC9X/YPemm79pgb+Zvpmc9+Rn9q9lf6tfnZ1IPjI9NhZiqxqj51bKbGrBr1Yt/JPjcHX2XSpatiujS2XZLpiOgBZlrRbAOenZPnaLA/5Px57733DnU8FyA9mFwjU9e56UHSn8oziS699NKpzHkmxz39qPTs51zCc0xyHeZ+IlPy0oOcZ85UjT5k9occL93a0qVY3S9yfHTnTTC9Xvr0uV5w/urOj0oNeKZG7s2fe+65oY7rR96TqcVSD66Rc1tPUg+eM5R6fPnLXx7qbrzxxqnMccW9WOrBs3zyHecDH/jAUJdrAscOz7jIe3K+Sj26tWUO4yOfgWeuXHDBBVP5a1/72lCXaYjp0ef+N/s23+Nyb8Z9Ub5jsM9T19zvcrzmmKBWXdrM7eBf9kVEREREREQWhi/7IiIiIiIiIgtjz8P4u9BXho0wHD5DuRjml+ExTGeRIR1M98Iw8gxX+j/+j/9jqMvwHKYWy1BkpppjmFOGauwkZcJu0qXeY2gK2yrDUR566KGhLsOGGYKWaXxWWQVSj//1f/1fh7q0VVCPTIl1++23D3XZj6rG38XQvjnBMFDqk23H35jjg2lKUg+GjTN92YknnjiV//2///dDXafHPffcM5VvueWWoY6hhGnnmVsYf0I9eJ0hvAyty2umVTr33HOnMkPHeI+cB//H//F/HOoyRRbD3DI90ne/+92hjmHkGdLMPjf30OQk+z37UZciLecg9muuNaeccspUZvjgNddcM5WZZir1yFSxVVUPPvjgcJ1hf/yeg15POrowfoa35phgyGr+O67nHEuZ4jLbv2q04tEyeOedd05lhokzrH+7ehz03LUqFWKG8Wdqt6rRdpJ9vKr//Zz3ujbIf5vtXzXOSbR9cY7M9YPzVa4nc0vr2qVCpB6ZVjXX76qxzzFdF/dCnWUuxw734jleqDFtk6krbZqbZDvKvRAtQZ/85Cen8k033TTU5dxGG8Xf//3fD9fPPPPMVGaIf96foeGpT5cyvGrUmZqnBp2ldQ7jI9uDc9JnP/vZqfxXf/VXQ12OJY4P6pFpd2n/7azCXYpPvudmO3P9znbu3k12ood/2RcRERERERFZGL7si4iIiIiIiCwMX/ZFREREREREFsa+e/YT+hXoS0nvO1Pfpa/y6quvHurOOOOMqUx/F++RaUz+p//pfxrqMhUf0yf8l//yX6YyPX5Mz7Rd39hBw2djWopMScV0aplWh2m/UivW0RuWWtKTnH5Zem/+03/6T1OZevAe6evbJD3oR8zfxZRDqR3TfqWXlSmomA7z2muvncp/8Rd/MdSlN40pmP72b/92Kn//+98f6pgmsPNPHzRdihPOX+lVZJ9LH12eg1A1njFCjyP9d6ndV7/61aEuvc55ZkJV1X/8j/9xKv/whz8c6nhuQ56rMefxQbo0rzwrJLWilzlTjWU6w6qqiy++eLjO9YPezXwejoH/+l//61Rm6irqkc86t/GRrEoHlO1M733CMZDnJuRYqerH0uWXXz7Upbf11ltvHepyzWB6P47l1KPzXM4N6pPnevAMhUyN2KUYJZkGrmqck3j+RZ4zk17ZqvFcF+416IHt1o856bHqDIVci7lPytS2PNMiPfTpAa/qzzDgnJhtzjMtct/cecJ5zfbfJD3y7KkrrrhiqPvUpz41lTkH5bvC3XffPdQx7W22M98bcg/HupyDqGN3ttDczq3ooD45f1x33XVDXZ6Xw7GTZyFw78N9Uo4f7lNz3uF5Zt25CN0Y2E89/Mu+iIiIiIiIyMLwZV9ERERERERkYex5GD/pQhgYvpeh2vfdd99Qd8wxx0zlK6+8cqi7/vrrpzLDPRiqk2kZGMqW4VH/1//1fw11/5//z/9nKr/44otDHUOnNoUuTWLV2B4M207LA9NiZCh4lqsOD9XJMMCPfvSjQ13aOv7Df/gPQ91//+//fSoz3ckm6dGF7jA8K8NLaWtI2wvT+GR4K8PTGF6blhimucpwV+rxne98ZyozDJNhZ3MLH0vy2Th38LkztIt2obRZMHVShp2xjnqkXYZpEjOdW9ooqsb0oEzxw/RZB51yZ7fI38X0RBmixzbOUPBMi1h1+BjMkH/OM9/+9rencq4XVWPoPsMFqcecQ/fXIdd3hjrmb2aY8jnnnDOVs/9XHT6X5JrB0NdMQ5bzU9WYZolh+6vCZOcK19bOZtGl5WMq3Qz/z3D/qsP3QtmWrMuQWaaOzTmqC5mtmvcclW2+TipE7kVzf8U2z7kt975Vh9v0MoycbZ7rNO18qcEmp2NNVumR7xhM/Z37K2qVa/1555031NECkXMd99Q5dvhulBq8l/Ewp7HD9ue6nPMQ97QXXnjhVE7dqsb2SUtxVdWll146XGefYD/Pcda9K70XK9Fe6uFf9kVEREREREQWhi/7IiIiIiIiIgvDl30RERERERGRhbHvnv2E3gZ6JNLLeP/99w916ctIv0bV6GlKTyX/XdXoe6V38ic/+clUvuWWW4a69Jut8ljOyRdD8tn43F0asPQ4Vo1p8S677LKh7mMf+9hUpp+feqRPh+lf0nd82223DXXpWWdqnk31lPG56c1KH92hQ4eGutNPP30q05efWtEDS19+epgyNU/VmE4staka/ZnUo0vdNGe61G5V4+/kuRHph2RKy9QjvZlVh+uR45VpfXKOyrmrajxvYynjg3CeTa81U6e+8cYbU5l65HpCfyznq1yz6APPMxTox8wzHdbx6M95LSHdmTzZ/lWjv579M1OHdv7xqnGNuvPOO4e6PPeHZ5ykR3kdj/4m6UFyPeEYSD3YHul7Zdoxpt7LfRJTGqYvn+km0x+7jid5znqsSk2Z8JyCbB/6vnNMcL3IvRevOSdmO3NOyrG7zv52bnqsc4ZC7kU5BnI9ZxunDz/PPKo6/Byq/Lc8YyQ132k6vbnT6UHPfs713HvleV559gXJfXHV4elAUx+uQzkncXxuwjlH/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYB+rZX5XXPX1C9EikR4N5vNO3RE8fSe9k5sesGn2W9JTN1ZfxXuBv6jz89LOkR5t+zNSAfhpqnucmUNc8J4D+pnV8lZuqXefhp8cu244evxwf9EVR19Qjx1zVqAfz8q7jy99UPfjc29WDZ5N0HjP285yv6DvOHMrMK0/fa8em6LFqXGcfZL/O80DYV7OtqBXHUupBD2zeg57PvMeq37Epeqwi27kbH+y72eZdvml+dh0f+lL0WMcXnr+Z83d67c8///yhLjXIM4+qDm/X/Cz9sbmecQx268ec238dqFW2Ffebed4Ez63Ic0ToO+eeIc+HYa7yvAeZS67wdWEbr+PZz7kl97dV4xlF7PPdep5+/qqqCy64YCp3+yuySeckJJ0e3ZkJVWM7516nqup73/veVOZ7Q85R/E6et5DnW3E955lVySbo4V/2RURERERERBaGL/siIiIiIiIiC+NAw/hXkeEPDBfLFAqXXHLJUJdhNAyx+elPfzpcpz0gQzKrxhAPptDI1FoMEV0nlG7OdOEnnR7nnXfelv+O4YIMV8vQS3420/YxpWKGzC5Fj3XCfxielOmqmG4kxwTTG9LKkiG1HEsZBnjyyScPdZnqjSG7S2Gd8ZH9lemRcgxkOqqqw9OBZngrwwUzJPCpp57a8h7UcVNZNa6znm2VKXa4JmS4HsNgGWKeFqVMSVY1pvCj5SJ1ZT/i75pLGOCR6DToUilxfGQdwyVzrWVbMBVipq7kmp3h6BmmXtXrQeasR9KFzFaNYbP8TbkXuv3224e6/CxDbz/xiU8M1+ecc85UZkreBx98cCpTj63uV3X475hT6tB1wpS7kHvufTJtIfdMOXY4Hq655prhOueka6+9dqjLdK2crzYhtdh26PTobBXcJ/3whz+cyo8++uhQl+s714TPfvazw/XFF188lW+88cah7p577pnKnR5kk9aP1IB7WP6O3Ldwf5P71tSmarS9MPX3X/zFXwzXl19++VS+4YYbhrr83k3Uw7/si4iIiIiIiCwMX/ZFREREREREFoYv+yIiIiIiIiILY1aefXobMk0F/Xdf//rXp3L6LKpGH/4PfvCDoY5pGTIdDFOaHHfccVP5qquuGurSW8sUPyQ9GuukJNtvVqUi6fT48pe/PJUvuuiioS79f5mypOrwNH0nnnjiVD722GOHukwbQ/9feqqYHqtLKbgpfqaqwz1N6c87++yzh7pPf/rTU/nMM88c6nIM3HbbbUMdPcnpwaQeeS4Az2nIVDUcH10qq03Sgz7wnEvOPffcoS69kvQrp1cyfXpH+mymhmHamNSA4zM1ZxrTTSLnKLYN0xrl/M0xkHX0Y6aXmOl36MH8zGc+c8TvrBrnwXvvvXeo61LNzXkMdJ5ktv+HPvSh4TrbLuf5qvGclbvvvnuoy7WW/uw8p6Kq6s///M+nMvXIMclzM7IvbXKqvc6/zdR3eW4FUxTnmGAKsDzfIL+j6nBP8qWXXjqVOV/lmS+df3rO7U+oR/YrjgfuN3P8sJ9nWrZHHnlkqMszcXiW0cc//vHh+rrrrpvK1INjcrvMWR/2q+znq/TI38VzoDL1W56nUDX2AZ5hwTMUcvxQDz7PVsy5/Qn3sDkP8/dyjs5+zj1lnrvD9Kypee7DjkQ+A/XgfmMr5qqHf9kXERERERERWRi+7IuIiIiIiIgsjAMN418VNp4hFTfffPNQl2kq+O8yVcx//a//dahjarFM1XP99dcPdVdeeeVUvvrqq4e6TP2QIT1VVb/97W9rE+lC0KrGsMibbrppqMs0LgxBy1Dxv/u7vxvq3nnnneE6Q/syFL1qTLHIcKiHHnpoKjMtBsN65pyKr0sNw1DLtDUwbUuGezME7Tvf+c5UvvXWW4c62kzSLpEhy1VjqDj1eOCBB6YyQ6HnbGUhXfo2hgFmyCTTgWb6Q7ZHzldMQdWF4lKPnC9pbUo95tz/V5FjgmF+DNvOuYRh/DmWmMbn8ccfn8qcy5m6J8cZQ8ozVRDDOZmuaVPJNYLtn2tr1Thf5NxVNVoZuJ7mms1QdKYVzRBKhjTnPRmiOdfQy1VwLGe/5u+nHmmr4NyWqd8OHTo01OV60qW0rBrnL47X/Les6/SYs1bUI38/5wDa4rI9uGan9Yp72CRTHVYdbifLMcr0fgnDrTfFake4h8r+Sesh97sZNk5raM7trOvGIO3AOQbZ5uwDySZpkGOCvzHn4UyNWnX4PjEtv2zz/CzX7NQ1Ux1WHZ5eL8cH7XV5z02yev0z/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYs/LsM3VPemDpCU5vxdNPPz3UpSf5/vvvH+rSh1M1+juYpiR96fT+ZNox+ggzDcTc6Tzi9CSn54znG6QeTEXywx/+cCrTH/v73/9+uM62pKcs70HvT3pp6f9jqrE5+2u61GJdaq/Oe8R0bplajKkoeS5A6sPxmh5YpnrLOmpF5qxHjgnOTyeffPJwnT556pH+TKY7TA8/z7CgJzbTI3Wpm+gHzc9ynJM56dGlFuPvT49+1ThHMeVOei4z9WHV4Wd8JPR65xhkXbcOdGl85uwHpB7ZP5m6i+dW5PkGXFsyVSe9kjmfM/UhdaUnM0kPbOq/yXAsp0ec51Rk+1eNeximsso9Fc/ASZ3pe/7kJz85XOeegeeR5LxHzXnuz6awjkecfTXnBK7LmXqPWuW8k2cnVR2+DuW6nPsA3mOT9diuR5zzE8fLm2++OZU5Bl577bWpzLk890Jf+tKXhjqmjM7+wrMxUg++t8xpTVgH6pHniPA8A64n+V6RbVM17pt41kGu0X/913891DFldaYJ574g05HyXIBN0MO/7IuIiIiIiIgsDF/2RURERERERBbGvofxdyE2XSolhthnWNGrr7461GWYE0NseJ3Pw7DQDPlhCG+GmPC5GcqV92AY5EGHf3SpxRgKnGF/THmU38NQ5AwFZ/gmw8MyDJFh0hmC1lkO2K82KdVYPiv7FdN+ZRggQ5iznzOFSdYxBRbHQNbz/hlSy/D/ZFPbv2rsS7TrMCwzQyaZ8ig1oB0j25GaM8wsQw2ZSqtLj7SpYctd2Dj7LtOjfvGLX5zKDAnMsHG2cYbecg5kmHKOO85tqTlDC5eiR66LnB+YnvVzn/vcVGZYaoYUdynBGAb7la98ZbjOVFepcdVoq2AodK5Rc7ZRkC7VG8OSs/2rxrB+ho3nnMR+nWPiq1/96lB33XXXDdfZPxh6m/dkf8h9wZzbv2r7e1qm3vvsZz87XGf4Ny0PDz/88FR+4YUXtvx3f/mXfznU5Z6tagxNT3tl1WiroJVp7hpsBfXI8cE9E9N7516IemR6UM5XOa6+9rWvDXV8j7jvvvum8n//7/99qHvrrbemMu2um6RHNz6yPWgF5Xqe6aU7WwXJdYjrN9eBb33rW1OZacLzs5tka/ln/Mu+iIiIiIiIyMLwZV9ERERERERkYfiyLyIiIiIiIrIwDtSzT484/Sx5Td9x/lt6odLLzHvQW5zpN5imJD3jb7zxxlCXvvTObzZ3Ug+2Tfqbqsa27FL+nH/++UNdepjSD1t1uJc2fTqZyqxq9Ewz3WL6Mek32yQ90tNE/zY946kHP5vnTVx55ZVDXfqHzznnnKGO15kOhec0pO+YaWNSD7b/JvnNulRv7Mt5FgJThOX4+PznPz/U5TzD+erSSy8drlMfjsFHHnlkKj/55JNDXaaf3GQ98jez/bvxQf9y+st5Hkx6ifnvLrjgguE6xxk9yenHZF2OnU1qf5LrB/Wg1zvXF55pkeOFKeLefffdqcy5jJ7kTImU7V81puFNP2zV4Z7YTaE7X4DrN/dXWc+UbXkeCM9eyHmOeyaeiZTz0He/+92h7oknnpjK9M5u0pqd8Lm7fsW5PscP2zXnIX5n/juuF+nRrxrTUqc/uWqc9zYxtdiRYFvlOsh9O/tuzkk8jyTTuXFc5djh/pbp3P72b/92Kv/4xz8e6vIsgE0dD1Vj3+HePN+jeNYX9zfZrlyXc0/LM5HybB/2Y/ry/8N/+A9TmWnCs79s4njwL/siIiIiIiIiC8OXfREREREREZGF4cu+iIiIiIiIyMLYd89+Qh8KffnpWXnooYeGuvQP0xfz7/7dv5vK9B7R25x+J3py8/6PP/74UHfPPfdMZXpNOn/NnL0ebH96mtKHnf7gqtG7mh7kqqq/+Zu/mcr07DCPNTVIMm9ytn/VmIc2fVlV825z5knOZ2XfzfavGnOLpv+xavTL0pP813/910e8X9WoY9Xom3r99deHugceeGAq/+AHPxjqMhcwc4rPWQ+SY5m+UuamzmvmQk6PH89FuPDCC6cyfYP0n+Vcc9dddw113//+96fynXfeOdRl39mknL18tnx29sfHHntsuE4/Nz2XmdOXXvP0L9OPSX1yXaD/7x/+4R+mcuZlrhrHxCadocBnyzmKv5Ee1DxTgfN86kEffn6W63f6+auqvvnNb07l//yf//NQ96Mf/Wgqcy7NtW/O7U/4rL/85S+ncs7PVYefaZH9Ls9mqao6++yzpzLXKJ7FkHBf8B//43+cyswjnnPkJnvE81m7PSz3LJxbsi/nmlA17nfpZc6zMLju3HLLLcP1f/pP/2kqs39seh7xf6bTI8c95yeui9ddd91U5pqdeyrmjs++fNtttw11/+W//JfhOuerV199dajLvfImjQeSz842zrPQfvjDHw513G/lGV5nnXXWUJce/vTvV1U9+uijU5lnJlCPPOeF7xGbPCaq/Mu+iIiIiIiIyOLwZV9ERERERERkYex7GH8X0pEhaFVVL7/88lRmiEeGG3/xi18c6jIMkOGCDOvJkJsMi66quvXWW6fyN77xjaEuwzcZxs/w9zmH4GRoCkPp+LueeeaZqczwpAxvpR6nnnrqVGYYJvtAhu7QOpGpezL8qWoMoWb70zowJz34LNk/OR7YPx988MGpzNC+/J5Pf/rTQ12mlGFYMkOXMqUeQ/UzLJYh5V3YOMfgnOjCxjkemG4wUynxN2bYdqaXrBrDa/nvOAbuuOOOqcxUVtkfnn/++aEux8QmhY2THMsM46fVK8OPOSfceOONUzktFvzs22+/PdQx9DXTV91+++1DXabbY7j5poZo8lmzrRhCzHGfn83w5qqqa665ZiozlW62HTX/9re/PVznmMhUe1XLSc+aUI+cv5k6iuGtGULLuSxTHFKPnMsYtk8rS85X3ZzUpRDcJDo9OJczTDnb57LLLhvqMqUebXn5Pdwnc9+aVifOSZuafrKD4zp/M/su1/e0JTHlaqaX5jtGzjPf+973hrrcM1WN89kmW1m2C/XIvptW3KrD053n+GEYf75j0FaR9l/uU5nCu0tRvOn4l30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGH80T9t0xjC9Cu7cnN8J33H6YE94YQThrqLL754Kt98881D3SWXXDKVM2VJ1eE+qSeffHIqMxVHpjNL30fV6P2hH3STvB6pwSo9MlXMSSedNNRlqphMWVJVdd55501lphKjBzN9O/fee+9Q98orr0xlemlTgzl79FeRGrD96UU65phjpnJ6lqpGTxn9f/lZap4+46pRD6b3yzHA8wXS/zdnj/4qsn3Y/jzvIL33mQqmatSDfrPUke1IT1lqkOOhahwD3TkJmzQeSI6JLgVV1ZjWM1OJVY2plLi2pHeSKdry3JKq0adOH3pqwDVhk9aIjtSA44F65Dk79IHndY6HqlGDN998c6ijHm+99dZU7tblTR4DHTlHcb5iqrccH/SB576JeuTay3RhPLch91ubvC5vl24PtWp85PrBOSnTGVPHHBPcT/G6O7tlibwXPTLFJNNW5lzGe+QcxH0q3z9yjVjieCDr6JHvf1Vjv+fZXzlHsV+nBjy3hOckbOqY2E7f8S/7IiIiIiIiIgvDl30RERERERGRhXGgYfzr3KML2WT4TX4PwzK6FGUMN+7SI+X3bmroxyqoRxfynxow/CY/uyq9S5easQtFzuulhEOtGnNdyH/qwfCoro27duX42G5Y7FL16K47PRhe281BnQWCdUejBgnbvNMj5yh+Z7Yr5/auzTc5peFO6daELmSTdVwzks4O0elxNLQ/2S09uGYk6+iRHI16cN7p6rq5jWtG0s1XR+Oc1LFbeuRnu3eDddboo4FVc1LSadV9b6fHUlJ8EsP4RURERERERI5CfNkXERERERERWRi+7IuIiIiIiIgsjFl59nfKXj3bUvwc+03ny3kvbaoeO2MdPbo25veox8hO56Gd6iE9e7EuqMfO2Q09bP/dQz3mRecD3y5L0WMOe43t6rGUNu+Yux5HgwYJ9djOuXH+ZV9ERERERERkYfiyLyIiIiIiIrIwts6vskHsVgjHnK0Km8Ruhdiox+6wWyFoR1uo1LrsRvvYxnvDOu26W7Yj2Zqd6iF7g3ocPEd72shkr/b0R7s9YqfMQQ/Hx7+wk9/vX/ZFREREREREFoYv+yIiIiIiIiILw5d9ERERERERkYWx7dR773vf+7as22QfTHpIduonOYjfsY5vbg7tvF126gc86N84h9Qke02nzRJ/79w5GvqciIiIiByZ7ez9/Mu+iIiIiIiIyMLwZV9ERERERERkYfiyLyIiIiIiIrIw/mSn//Af//Efd/M5dpU//uN/+T8Mehnoc83rP/3TPx3qfv/732/5PXP7/XP263Z6kNTjT/5k7J6pBzno33/Q91+HnZ6LkDpWjWNA//jO6fK6d1p5hsLe0Omx3X+37r8VERER2Qv8y76IiIiIiIjIwvBlX0RERERERGRhbDuMvwt/P+hwRT5bpgl8//vfP9R9+MMfHq4/9rGPTeWPfOQjQ90f/vCHqfz6668PdW+//fZU/vWvfz3UHXR7HARdSHGG49Mq8YEPfGC4/tCHPjSVP/jBDw51v/vd76byz372s6HuF7/4xRE/V7U/ehxECO92w/G78UGrBK9z/FC7bOdf/epXQ91vfvObqZzjaMls167S6cE6Wifys6zLduYY6CxJS2WnenTpWLt78LNpc+EYmJsNbD/YDfvQOjYwkv+W33O0jIlkp3rs1D7UrZFHY/uT3dBjVb/W+rV9tvuOs9Q02JuKffzI+Jd9ERERERERkYXhy76IiIiIiIjIwvBlX0RERERERGRh7Dj13n57H7brCa+qOvbYY6fyaaedNtRde+21w/WVV145lenvf+yxx6bynXfeOdT9/Oc/n8r0x3Ye5U32jHQapK+S3u48F+GUU04Z6i699NLh+pxzzpnKbKsnnnhiKj/00ENDXecRX4pnfB1vcXq7eS5C6nHCCScMdWeeeeZwnXrRl//8889P5WeeeWaoe/PNN6fy3NNW7pTOX08/fc5RPIsizwrJMytYV1X10Y9+dCpnn6+qeu2116YyzxjJMy04HjZ1TurGQ9XY5l0d5/1jjjlmKlNHfk/OdezX77777lTO9aKq6re//e1U3tT2J6vOm+C6kOR8leWqUY915nbeP8/W4Tk7S1kjklV6ZL9nH+zmstQj+3HV4elx83v5PLlvOohzdvabdcZHpwe/J9cTrgnUo0uXm2NgieOB7JYeJMfVKj268ZFaLWXPtA7dHqr7LLVKHanHOv18088Y8S/7IiIiIiIiIgvDl30RERERERGRhbHtMP6DCFvYaZhyhu5/7WtfG+r+/M//fLg++eSTp3KGiVdVPf3001OZ4TcZNsLwEoafdOE4cw7P6cJkGWqZGjC94QUXXDCV/+zP/myo+/znPz9c57995JFHhroMDWebd2Gg6+hx0OE5nVWiCyHurCwnnnjiUJd6sP2vuuqqLe/5wAMPDHXZdi+//PJQlxowVKoLV5tb++c1+xGvM5yS4fhph6C16BOf+MRUTltRVdXZZ589XGcYGq0sd91111Sm5SLDllfNOQetQdLN+wwL53VaIHI8VI1tfuqppw51eU1by/HHHz9cZwrQxx9/fKh78MEHpzK1ytStXWjn3OhCK9nnuQ6kBYVzUurRWYlogaHmaV956qmnhrrUJy16VVXvvPPOVKYec6YLzWefT/sW6zkGzjrrrKnMOSjHAMPvef3iiy9O5eeee26oy/3Vk08+OdSl7WWTQsq590g92P4cA1l/+umnD3WpB8dHznPZj6sOT1GcGrz00ktb1r3wwgtD3S9/+cupPOc9K6EeOV9wLucYyPq0l1aNeuQ7RNU4Jl955ZWh7qc//elwne2cNryqcezwe3IfMOf1gnTz1XHHHTfUsV1POumkqXzJJZcMdWecccYRP1c1ttWhQ4eGulyHq8Yxke8bq+py3purHv5lX0RERERERGRh+LIvIiIiIiIisjB82RcRERERERFZGLP27Cdd6iR6Bc8///yp/IUvfGGou/DCC4frV199dSqnx7JqTC1GD2x6NFZ5mLqzB/J6nTY+CD3Sb9N59umFyvR6n/3sZ4c6eqHSw5SevqrRj8mUP53vu2sr6rHTdt0LPdZJLZbpkKp6v9n1118/lakHfYWdTynTuZF12qM7p2C/+znv13n2eU5CepLpG0vf6+WXXz7Upf/smmuuaZ8vfX3d+R8cn10bk53OSftBPhv92xwD6avMcyqqqs4777ypfMUVVwx16d2kj5N9Pj2YHB/5fPSWd6m0yNw0SHIMMI0nry+++OKpfPXVVw91qQ/1yLaj5vQkv/HGG0f8d1VVb7311lR+9tlnh7pOjzm3P+ekPCfh3HPPHepyPFSN63KuCVXjHirLVaOHnmkr6RlP3/Ftt912+A/4/8MzX3K/NbdzdUh3llH68tmvuRe97LLLpjJTROf6wTkpz5jgmtR59r///e8Pdd/5znemMr3M6Xueu2c/9WB7fPzjH5/KXRruqnGO6taIXPerxr0p78/1I98xvve97w113/rWt6YyU7em5nM/YyT14Jycetxwww1D3XXXXTdcp145VqrGtYb7gOy71IPvdTk+qMc3vvGNqfyDH/xgqMtxNtczRvzLvoiIiIiIiMjC8GVfREREREREZGHMOoy/C3/PcJBM21NV9ZnPfGYqZ+hg1eFhNBmq8fd///dDXYYwMzwtQ2fWSWU195C0jk6PDONn+GCG52SKjKrDU5FkqF+GMVWNIYEMM0s9VoXx5/Xc2j+fZ1V4b2erSGsLwzczbDzT9lSNVomqqjvuuGMq33777UNdpjGhHuvYKuakxzrh7qQLBc7wWlolMsUMw8qoR1qN7r///qEuQwI5zx10u+6UVVaWpLMWMdw5w2sZWpj3yDmn6vB1INucaX3ys3MN7VsXtmO2Ode2Lh0ow7/T9pJhl1XjmGC6MNq5Moyfoa85Blaly50zXdh4tivHCsNbMwUo0yTmHMU+nxZGtn+XVpW2jtScodC5njCd35zp1uHOalc12u1oTU37Cu1CXRg357a8Z9o4qkaLGNNWZpjyJqUK5TjPfk2rHfemqQfnqyStwFXjXE/NmSY8UyzeeOONQ12mO+Ta0qWmnLMe7I+ZVpX7VNpc8r2C39OlgU6436U+qTnTUie8R+7L5rrWb84KJyIiIiIiIiLbwpd9ERERERERkYXhy76IiIiIiIjIwti2Z/+g6fxnTIvx5S9/eSofd9xxQ93f/d3fDdfp2X/mmWeGuvTFdL78dTzJm0znJ0+PWXrCq8aUJvRGZrqXqqpbbrllKlOPLr1Fly5szh7xji4NXFXvDUqvWvrCqkbvJP1/1OOuu+6aytQjz1ugr3JO6fR2i1X9Kr3F9LKmh56+zvTE0iP+0EMPDdfpEaevMlOLsW+sk1psU/Shd5Rtnn07/eJVY9vx9z755JNTmX5M+peznZ944omhLscH75Hjc5M8ySR/169//euhjp75TGtET3KOHfpsU0fqQQ9s+m6ZXi+fj+tQekDZj+Y8HvhsuU/p0hJWVT3yyCNTOf3BVVUPPPDAVOb5H6kHx0N6cKvG9Fk5P1WNPnR6Z9Mj3aXZrTp4fbq1Luconr+Sc3nV2B6PP/74UJfjg+fj5DzDtmLa3TzPis+dY4fnyuSz8R5z8yh3euRcy7mE+5v8XTxLJ/s9z53Ka671TBmXqY85J2Y60jzXp2qc2+Y+X3XnoeT44PzEcwpyjsrzJarGuYXzXtaxjZkCNvXgGMj3mEwZWDXuJ7iezyVVpX/ZFxEREREREVkYvuyLiIiIiIiILAxf9kVEREREREQWxsZ49rs8yV//+teHurPPPnsqv/DCC0Mdc7enJ/ndd9/d8v70Xew0j/hSoB6Zs/RTn/rUUHfqqadO5UcffXSou+2224brzB3e6cE2Xcc3Nmc91vG6Zx9k/8x8uuedd95Ql35I5mq/++67h+v0dXZ6rHrWpMvFfNB05ySs6nPp1aLPNb179PGlB5NezfSPV435XekjzO/l+MzfsSp3/ZzGRzeX0rNP33H6KpmfPb173dkL/Hdd27Af5zhjnuZOq7l5khM+W2pAzz49ygl9ppn/mHU573Bc0SOeMBdz3oNnOKTPk+OTHPR8lf2Bc1D2XfrpeaZEzjv0K6eW9Ij/5je/mcocg8yNnfu0LFeNnljq+Morr0xlznPkoD3j3ZyUfZfjgWdaPP3001OZZ1Gk75gecd4z4XlWn/jEJ6byxRdfPNTlnoE55/PZOM+SOelB/3R3Pg7PL8pzCvg96RmnRzzvz7mEXvPUIz3hVeP4yM9Vjfsyjg/2h4PWI+dLtmP25TzTpWqcA6rGtmSb5/kLrNvqO6oOX9/z3ZHzVZ7/xj31fffdN5Xnqod/2RcRERERERFZGL7si4iIiIiIiCyMWYfxZ3hphtRUVX31q189YrlqTN1zxx13DHVMZdWFiOX9Gf6xTqq3pZDhptTjpptumsqf+9znhroMCWSYeBc+SFaFV/4zcw6DXUWX3pCkHgz7yxC9a665ZqjL8KjHHntsqGOYWYbU8nlynHU2ly7McO7stO8whDhDwhiSl2OAIbMM9cwQMI6H7AMMl0sNGCY9ZzrrBFO0MRw+fzP7YIYiMx1PjgH+O15naDjTh+XzUMfsH6tsFXOC4zz1WJUqNNuAIbO5DjPl0aqw4eSjH/3oVE5tqsY2Z/hm2ixW6dGt/fsN9cixzTmAc1LONV06Nf7+XKM5BrkOZegrrROpM+eybnyQg9Zju7YK9uOuXdkeaV/hb0ztOAfleKga7ZYcH7mn61KFbrIeqQHtEGy7tF1wnKU9g2tC3v/DH/7wUJdpj6vGFG7UI9uRY3kpeuT4WJVWNccA2yND96lH/n62MdNS5zW1y30z1/NN0MO/7IuIiIiIiIgsDF/2RURERERERBaGL/siIiIiIiIiC2PWnv30LZ155plD3b/9t/92KjNFQqYT+/a3vz3U0ZOcMD1SekborUjPCL1YS/XwpweVevz5n//5VKYed95551TmmQlMk5H3oJc2PWX05aT3h36aTdVjlQc224qpcr74xS9OZeqRaXToW6MXKscEz2lIXyVTYuV1l6Ju7qQG9O11eqQ3sqrq2muvncr07aVXjb5ztnn6yNIPWzWOic7PTn/upnr4OQew7RLO7XluAn18p5122pb/rhuT1DzHVtfmHDscLwftuezIZ+Vv5PjI+YJj6eSTT57K7PPppaUeTNmW1xwD2V+oea713dkPc4P9IZ+V5xF1KQO7M0a6NLtcW66++urhOj2w1COfh3XZd7pxPTc6Pz33JTyfKPc7XdpItkd+9qyzzhrq8iylqlEPjs/ci+WZJlWbtWYn7PM5JtiO3It253jkHNWdZcRUlF/4wheG61xruvR+3Kdl39mk/S3Xtjz7YFVq49SLn+3S3Ob4uPTSS4e6r3zlK8N1nqHA+TPPVsoUyFX9mRpz0WNzZlERERERERER2Ra+7IuIiIiIiIgsjFmH8Wc4xvXXXz/UZVgsw6MybJyp3UiGwjIk7dRTT53KDPnJtBxPPfXUUMfwjwzl6kLp5k6GLl155ZVDXaZ3YwjYs88+O5XfeuutoY7hexmOfvzxxw91qQ9DfJ5//vmp/OSTTw511CPDpeasx6oUUKkHw8WuuOKKqdyFszLkiemRzjnnnKnM0MK8/zvvvDPUpR5PPPHEUNeFns5ZD0I9MhQ4261qTIWYoXtV45igPYZtdcMNN0xl9uscdwwzO3To0FReFTaezE2PLiSuCxvPMPGqqnPPPXcqpzZVY8hkptupOnytOfvss6fyG2+8MdQxlVCSawJDNKnHpoSRr0p5mmsow+gvuOCCI5arxvDmLgVV1TjumF4vQ0aZkiytAl2a3ar5hGUeiS7lKefvLpXueeedN5VzrFSN8xfTJHIPlXaZF198cctnZd+hrptCF4q8yuaS8PenBkwXlmsG57Lzzz9/uM79FdeInOu4T8vfNdcw5e2Q8wdtFJ1li1rlGKCVKPs8bRRMg5zj54UXXhjqHnjgganMsZP9atW8O2d9cs/S2SGOdJ2ktYg2ybS2/Kt/9a+GutxPVY17uNwzVVV9//vfn8r5TlM1jm2u33PRw7/si4iIiIiIiCwMX/ZFREREREREFoYv+yIiIiIiIiILY1bGKHrl0nvxb/7Nvxnq0mNHb8Vtt902lekfTx9+1eh/Sp9z1ei9oZc5/Zi33nrrUPeTn/xkuH7zzTenMn1Cc/LE0jNDPdJflKndqsaUYHmeQVXVww8/PJXpu2EqjNSDfjNqlzz++ONTOfWvqnr00UeH6/TIsn9skh553sR111031KVW/I3p1aN39eabbx6us82Z6i3HID376Tej5jzjIn3Qm6QHz/FIHzI9++nro3c254RM/VJ1+BigH22r76H/L5+NXt7nnntuuE5PPz3SB61H53Ml2T/pc815h+ckJJz3mQ4025Wpm9Iz3p2bwbMX6KXNevoBN1UP+sDTO3nZZZcNddmv2Y7psawaveccZznPsI3To8wzLXL9rlrPn3mQrPKKZl/mmS+5DvC8pPx3ue5XHe79T72YojjXHnr/c9xxXLHPd2nh5qTHKm91jg/O++n9ps84zyPheTDUJ9csjoFsR/rXu9Rm/F1zPmMkWaVH9mW264033jiVOT7yXASOK/bzhOfBdGnpct7lvpC/K6/nNB7IOnpwH5RniFGPPH+EqUF5lk+OAZ7Xk+tAl6541VlbevZFREREREREZFfwZV9ERERERERkYRxoGP+qsNgM9bv88su3/LcMWc0wcobfXHLJJcN1hktlioaqMYyf4YIZ7sFQaHL33Xcf8dmqDj4MM1mlR4bCMtQyYcqpDJtneDPD1VJzpt7L0B22W2rFUFuStgKmmNkkPTLNEcPF8rMZDlY1hsXy3zGMPEPSGH6UYbK0p2TYOvVgGz/99NNTmSGAB61HarBKjww9Zbtm6COtCvm9nIM4BjLsjGF/Ge7N1GZ5zTBMhqRl+PncwsYTtj/JsD/OM5zPk+zXbEfO9d3cn3pcdNFFQ10+O38Hx1JqwLo5wd/RpQrNlIVVYzsyFDznHc4lvM6Q1gyL5j3YHzJ0n32802NV6Ol+k21OywPDfbOdaWVJfTKtFb+X7U/tcp/EMZf65PpdVfXaa69NZc5PnD9Tjy6N6EGQetBWQn1yruE6kPst9t3s17Qdsc3TSsG+m89HzXNPR9sE1+xOgzmND+rB8ZHtyv6ZKfQYGt6lgaMeaVFhv87n4fdkmleOjzmv2STXDOrB69SD81VaWz71qU8NdflZhu1Tj85ynXpwnKWOfG62f/bB/RwP/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYs/Ls00uavhimXUs/5PPPPz/UpQ8kvcOrrunZST83PUyZ0oQp+5599tnh+plnnjnid1aNfqeD9jMR6pFeMXpf0m/EVF75G+mBpecv/S5MgZTfQ89MpjRhGrr0IFeN/jOmjDtoPbpUVvQCpQZMi5fj44knnhjq0sNPvx39ePlZpgjLcUY90vufaWqq+jMd6EM/aD0SatN5LulfznbszhhhmineM/146dtjHftDnpXBZ8v2rxp9ndQ873HQevD+nL9z/qKvMcc9z1HJOnr6uA5lPeeZ9Pxx3suzY+hzZh/IuZXjc04e5S49VNXY7zi355zAFG3peaQe9LJmH2Bqsbwn1/NM3USvfzcn8TyUOfljO68o4TjPPRXXiNSRfZdnjORnmcoq1zOe4ZDjlZrTS5v+/m6+OghyTPBZupRc9G+/+OKLU5n7gNyLch3mvJPfy5SSuZ+4+OKLt3y2J598cqjr0rOybr/XjC79JPXgfjc/y7GUbcfU37l+cG3tzlDgOLvgggumcjfv5/vFkZ41xwTvcdBreD4r9ei871zr8zdyX5SwHfk9Odfz3STfR7mHyuscq1WHt3HObft55ot/2RcRERERERFZGL7si4iIiIiIiCwMX/ZFREREREREFsasPPv0s2SOb342fY30SOT30HdBX076+ugFS38L849/8pOfnMrpF6863M/e5YM+aM9Mtiufk22VPkd6XVIPeiWzHenLocfuwQcfnMr0Qyann376cH3zzTdPZZ7LwByp6bXufIwHDfWghyn9efT+pG+JZ1qkR5meS7bHQw89NJXpP8v+QT0+97nPTWXmRGUO+rvvvnsqr8qdfpCsGh/ZlvRc5pigx+6pp5464ndU9X5yei5zrmOe5s9//vNTOb2AVYefsfHII49MZY7zObFqfOSz0weev5ke+TxTgWtS5vqtGuconv+RetCT/OlPf3oq01ebHuSq8SwAPmt6EA96LSHUJ9cBziU5z3DNzjHAMce2yzWcmuc5Fjx7If2Y1Jhtns/e7Rn2g279Yn/gGpH+VK7Z6StlW+Xvpx5su7wn75/+fnrNv/CFL0xl9ge2eZ7FcNBnWqyjB58t5w/uabNf8cyb3EPxHBnqkfV8ntzHcv3I+YtnWrDNO0/yfp9p0Z2LwGdZR48f/ehHU7k7t6M716dqXO85X+ZnL7300qEu9fne97431OX6XTXuBan5fs9XpPPss1/lfMXzP+66666pzN+U/ZV65HkXVeN6z77DPW6Setxyyy1D3dNPPz1cZ7/ib9xLPea7sxYRERERERGRHeHLvoiIiIiIiMjC2Pcw/i5snCGsGa7F9DcZOtOFQjDMjCE3ec1wi/yeDPOrGlP1MFSKIbzdb87rg07bw2djCGu2M0Mks60YZpdhZqtCVvOa4cXZPjfccMNQd/XVV09lpv9h6FL2Cf7m1OqgU+8xhJrjI5+PqcXefvvtqcyQ2Qz/ZvgkU2JluBS1yudjer38HqbHYhhg9gn+5jnpsWp85LPzs9keDM3PsEyGmXUWjNS4auwfnPdyvDLtGK9PPPHEqcwxuE6Y6l6Q7cpnYRh/XrNf5RzNkMAMm18VhpqpVDl2MiyWVq+cL0877bSh7txzzx2u00LGZ83fsZ9pfP6ZbnywL+dnu9BfzjPZ5xnmyN+c6znDQrMduUbks7H9mS43xyTHYOrRpf3aK/J3cHzwOtuSFpQcL/yN2cbc65AuxDztRGn7qhr3VAxh5j1zfeO+pLO55HUX7v1e6NYv9t18Vva5HFvcC+d+i+OK5DhjCPNVV101la+//vqhLvcJ+bkjkc/AtHRd6taD1qNLQ8wxkOODe6/UlXUkP8s5KfdUtN7ls7PPU9d77rlnKtOuk/dfp433Qg+OB+qRbcnxwbllq7pV60f+LuqR7xyZhrxqnOc4PzENclqVuRfcrh47sR/7l30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGEcaOo9evzon0gfGb0W6ZmhJzg9kPSI01eZHhb6QPJ76btIvy49VPTF5O/Y71Qwq+g8lyR9KfQi5b9lm6f3hV6Tzp9Jn1RqQE9yXtMzQz9i6nXQ5yR0dP71qvF38pyE1IPe8kxl1KWVqhr9mdSjS/2Xz8466pOes7mNj6TzIFeNz845Iduj85fRk009ctxxLstxR12365Ws6n3g3b/bb+jRZ7vms3dpI7v+yfmJuua61J0dQ11zLuW5Gfxd2z23Yr894bzmnNz9DvbP1I5zWc5B9DjSr9q1QerK83kuueSSqcz9BL2bOc669jjo8cF5n9edDzzr+DtSg25+qhrbg+Mzdc5zQqrG9MXUg2nh8rPUNX9Ht7bwN+6FZ5xn7lCPhO2Y7Uytsq5LzVnV7wtyTuK5Otdee+1UZipjrueZpq47Y+Sg9eDv5/yV8P0j95Rch1Mf1lGP3Bdw35xasq0yffFFF1205XdWjWc0cbzm7+rSvu2HHt2envfkmp1txfUj14hVeuT44Lqc895nPvOZoe68886byldcccVQx/N68hmYejx/R3f+xk7a27/si4iIiIiIiCwMX/ZFREREREREFsaBhvEzNKVL+8UQ4kyjk2nwyBtvvDFcMzwpn4GhZBkew1RvGe5x1113DXW8zlCNg0iPtF2oB0NMMnSJemS6IqbKye/JtFZVh4fK5DMwfPLss8+eyp/97GeHukw3kqktqqruu+++LZ9nznowrIrhSXnNlJLZVhnyVTXq+Mwzzwx1TNvSpSI588wzp/LNN9881OVYevbZZ4e6Rx55ZLiekx5dWG6XmofXDG3M+Yr/Lue5DLmrOjz0swszO+WUU6YyU4VmHefEp556arjOfnXQNpdODz4b56/u2bN/UlfasBKGSGb/5NjJ8GKOwbw/xy5DXzOk9yDSuW2XVaGeXWrKbI/u31HjLp1Z2pWqxvGS47FqTPXGfsN9SZfernu2/aB7HobR59zCeSb1YOhv/q5Vc2J+trMJMmQ3n5XjivaQ1I51B806qaY7PXLt5XyR8zX3CJ0eHEtpY+U98ntoq2Dq1rTRUrsMce/C+HeSWmw7dLYSzhfZJ/nekKHZtE7kfLEqbDzpUnZnmHjV2Macy9iu+R7DVLqdhTLnwf3Qg3MAbQ25DnTpm9l3812F7zTdekKLWI4tjoG8/znnnDPUsV+lHYA2wbzHbuvhX/ZFREREREREFoYv+yIiIiIiIiILw5d9ERERERERkYWx75799ETQG0f/V3p9mVIlPagXX3zxUJdeSabq6Xyc9GOmD4NeqLvvvnsq/93f/d1Qd+jQoeG6S7130J7LTo/OW/zCCy8Mdekpoh7p/6NnnylmEnp40ifDdkvv+Te+8Y2h7sknnxyu05vGPjcnPdhXqEd6k5heMNPKMBVI+v/op2c/zzGR3siq0StGj2E+zw9+8IOh7oknnhiuc4wetB6dJ5p6sO92HsTsuzwbJP1/9HvRN5a+z+5MC/oo89/xDAvq8eqrr05l9rmD9vCvMz6yTzL9TbYVPY+Z4pN1THuV3jm2efr0L7jggi3v8fDDDw919FWmHuxzBz1fJavOUMj+yvRpl19++VTmXJZtxRS49GdmP6evMtPrXX311UPd6aefPpU5HnjGRZ7pMGc9uvMuqkbfL9Op5TXbOH3Y9M7SE5tw/cjxcf755w91eQYQzwdim+f60emxjja7pWO2OfeX3N/k3M+5PecWesQTngvQrVEcSznX0S99xhlnTGV6kLv0pFzPc47YjzSi7PNd6kGeL5BnErGf5zX7dbY558Au3Tjvkc/HfVmuZ6nNke6x3XZeR4+devg7Pdiv2D/znY+/Ma+5F83+uGq+Tj14pkN+lu+VqR314D3vv//+qdyl3+zOrzL1noiIiIiIiIj4si8iIiIiIiKyNA409R5DERiGmeHw3//+94e6DLu76qqrhroM/2BIeZfOgOF6GeLMUOQMFWcqMYYhZhjHQYfBknXC+DN0n6HAGZLHsPEMW2ZdF6qS4fZVo43g3nvvHer+v//f/++Wz8ZUVl2qtzmxSo/sr0wXlqF+TIWYYUaf+cxn2ntm2BXDBzMl02OPPTbUfec735nKP/rRj4Y6pvvL9DRdKPxBsI7tKMPFujArhupnCOvXv/71Le9fNYaWMa1PhvrRLvPd7353Kt9xxx1DHbVLnQ9aj3VsFd1n09pVNYZBZjhz1RjizVA+3jPD8NgfEob/Z+g+56sM86sa58GDtoGtowfbLsNkM2y/arSBMfw+U1LxHrS5pB68f4Zasi7nMlqbfvKTnwzXGcLZpTY7CLr5iqHZuYf6/Oc/v2Ud09llWH+X5qpqDK/l/U899dQjlqvGkN4XX3xxqON8lWHkc7NVdPdnP8914Prrrx/qMlSc80zuAxhSz3DjfB6GSWfY+mWXXTbUpa2D607ajPg8tHXk/RnSvRdadd/J+2efrxptP5m+rmq0ZHAMpGWLe0/qk/MZw/i7fVruozmuaHnOMcn5crv7371KvZfzA+dk7pPS2tOFyvM9Li0w1IPvGNkeDLFPG8FNN9001GVqdlr/uJ7k/or7ghzbu93m/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYs/Ls04+XfqNvfetbQ116mOhtSK8H0znQ7/TUU09N5X/4h38Y6tJXeeeddw51nS+pSzcyZ6gHf0d6s775zW8OdakHU8ykH5P+Imqeaad4TsOjjz46lW+77bahLj3KPDMhPZ+850F7+kj3PPSH5m+mHulJpvcovU9M/8OxlJpnusmq0Tt56623DnWZvqpLzVPVe50Pms4Dy371/PPPT2X23fR7MQ1cesaPPfbYoY56pLeY6S9zLvvhD3841KUvnGcm0LeW/rc5jw/OHUxPlGe+3HPPPUNdelDTq1o1elmpB/3LOffTH5v9nmPnrrvu2vLZeP7GnM8YyeehHlwX03tNH2N6hHmGQs5RXD+Ygimhfzy9xvSB5/r+7W9/e6ijBzY9ut2ZM/tBd4YCn43jI8c9vbzpic31gvfo0lOxnme+pK7sKzmXfe973xvqHnrooeE692IHnbq1g89CPbJfMX1YnkvFtJW5DvEeXZtT8zxDgGnocizxvKScy6rGdYlnKGw39d5+sCpNYV7n+l01+vl5Hkzub7pzr6p6X3bOgxyDOdfyDAueL5brPc9b2O5e+L1otV3vOdfW7vyofKeoGtdzrtk5z+X+qerwdSjnD641eY8rr7xyqMvzFnguwI9//OPhOvXiu0rev5vbd+Ln9y/7IiIiIiIiIgvDl30RERERERGRheHLvoiIiIiIiMjCOFDPPqFHIf0u9Fb87d/+7VSmPzX9NfSP0z+cPlt6V9MLRT9PeisO2re3V3Qe5fTUVVX93//3/z2V6RlK/x/942+++eZwnf4iemDT10ePX+qzyre3KfqsOtMiPXePPPLIUPd//p//51S+5ZZbhrqzzz57KtObx5yxqTO9xPlZegNTn+7MhKrxd85ZG44H9rNsA+bmzvbg+Eg/HnMf0/OX4yPPbKgaz1fguEqvGjXmPTblTItVnuT08t5xxx1DXfZJek7Tm3f88ccPdfQo5z04PvL8EXo185prEue21GfOenA8sJ/lGQo8cyXhXJb5r0888cShjj7PXAc4BrLNeW7Fgw8+OJXpgeW+IPsZ57KDJscExwPbI73vPNso12l6UNNLzPmKPuguH3n6VXOsVFXdfvvtU7nzIPN7OZcd9BkKqQfPqmF++vvvv38qs13zTCTuRXN8UEd6/3O8sO/m93JtyT02z7aihz/7GZ/1oOev/M0804N9Nz/Lc3ZyXeAcdOqpp05lrhfMD59nKHBvnH2He6g8T+zv//7vhzrOrTm2OCdsVw96xHdLx2xj7vf57pZjm+vySSedNJU5di699NKpzD3DDTfcMFynBjnmVpHvpzxf7rvf/e5w/fjjj09lzgl7uffyL/siIiIiIiIiC8OXfREREREREZGF8Uf/tM1YgZ0c9b+b8P4ZHsNQmfwswzYYutSF4e1VKoolwJCn7eqxKjS906NLO7UpoeC7SbYr9cjQMuqR4VGr9OjS4nVpdLq6peqTejAELcPDGPbH66SzDnSpMrs6ajy3dG47hWtEXrONM7yVoa68TjiWsi27VIAcR11dN+Y2aexwTsprhqxmyCStRVnHf8ew5YQhxGllYQqqDAWnjYIhtHNKH9bB8cA5Kftypo6qGq2QGSK7qo7hznlPhqym7YUWg6xLq8yRvifDe+c8l3V7pqo+xWRa75iGLVNNM0ycKeNybmM/z3S5aW+tGu18+bmq9WwucxovHA9MQ5xtxZDuCy+8cCpffvnlW9alblWH65Hfy7bKlJ9PPvnkUJfWCaZ1pT0kx8vcUlN2eyauwznXM71eanD99dcPdZkmb9X4yHmQ4zXTtWYoftVoc2HYPm1QaR+h7Win89d2dPQv+yIiIiIiIiILw5d9ERERERERkYXhy76IiIiIiIjIwtgYz76I7A7dGQrb/Xfr/lv5Fw5iLlWrkU6Dneqzzvg4Gs8YSdZp/7zu6lZ9drvniCw1lW5H145dOr2ujvW8R3feRVe3VH26vpvtSK9/nkdC3zmv87PdeT3duRU8C4N6zPnchHXY7hhgG6fvm2eK8DwSnpWR5NkHTBOYPnym0+t8+Zs0VtaZk7Jfs03zXASm7GMq1zwLgG2VqULTv181njnC82A6X/5u6aFnX0REREREROQoxJd9ERERERERkYVhGL+IiIgsijnbjub8bLtJF5q+09+8F221yXqsszffDT32I5XuUvVg+Pl2bUekC80/2vVYZfVKOj1Yt9O04Htha1mnr/wz/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYf7L6IyIiIttnkzx+RwNHox5Hw2+cO5ua9muTWKdd87NznhPm/Gyr6J61833P+TfP+dnIOmcY0Ou+3TMtDvr37+QMPf+yLyIiIiIiIrIwfNkXERERERERWRiG8YuIyK5y0GFuMqIe80I9RLbG8TEvjhY9ut85pzbYybP4l30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGFs27P/vve9b7jOlAVz8jLMHaZM+OM//pf/b2EaiO7f8Vo9dofUo2tH9dgf1GNeqMe82G4KHvXYH9RjXmS7sh27NFs5z1UtU4+D+B3qsTVz+x2dHnNOS7dbzPl3mHpPRERERERERHzZFxEREREREVkavuyLiIiIiIiILIxte/bpX5izn2HOsN12y3ukHrvDbnmP1GN3UI950Z0rsg7qsTvsVjuqx+6gHvNit/Jmq8fuoB6bwzpn8qjH/rKT9vYv+yIiIiIiIiILw5d9ERERERERkYWx7TB+2RsMf5kX6jEv1ENElsZOUifJ3qEe80I95oV6zAtT74mIiIiIiIiIL/siIiIiIiIiS8OXfREREREREZGFcdR79rfrfaB3uPt3e+EzXuf+m4x6zIud6rHfHC16JHNOf6Me6nHQzFkPprD84z/27y4HydE4PubMnMaqqMfcMPWeiIiIiIiIiPiyLyIiIiIiIrI0dhzGn2FOexXi0YVSbTfMap3wOP6O/Lfdb2QdQ/T2IwRmTnp0n1ulR/c7urps81W/f4l6rDNW8vp973vftu/Bft3V5fV+jIdV88Gc9cgxsVt6/OEPfxjq8pq/Xz3GutRglR4J9UioRzc+9oK569HVdXp0z8o27+rWWT92g73SY51/t1M9/uRP/mXbuEqPvGabb7duDiHEu6EH6fY3nR7cQ62jR/bzuemx3fma7JYe63y20+NP//RPpzL16PZJv//974e6/F3dv9srNkmP1OC96JH9vtOjGx97xU71+Gf8y76IiIiIiIjIwvBlX0RERERERGRh+LIvIiIiIiIisjC27dnfC59n50uq6r16nUcjfRidP7Zq9F7wsx/84AeP+J1Vo5/j17/+9VC3H57Y3brHTr2s6ROrGtuVdflsnY5VvR7pvfnd73431P32t7+dyr/5zW+2/M694iD06Lx6eZ3txu95//vf396j84aldhwDeZ3aHOk7d2N87NU5Dev4pDo9so5tnu34oQ99qL1H5/XO61/+8pdDXerBsbMf89O69VuxF3p84AMfGOpyvHz4wx8e6jqPH9s16zo9DmK9WLd+K3ZLj2zzY445ZqjL8UI96KvMa847qQ/1yDVjP/yYu6XHOufjrLN+ZJtzTkp9PvrRjw51HAPZrr/61a+2rPvFL34x1KV2+7F+k52OyVV72mQd733OUWzzj3zkI1P5Yx/72FDHMfDzn/98Kr/77rtDXerD8ZHfsx/+5L0aH9s9F6Gqf//Id4Njjz12qEsNTjzxxKGO+6R33nlnKr/11ltDXWpAPXKe2w///hz06MZHzkls8+OPP34qn3TSSUMd3xVef/31qfzaa68NdTlHdXrM4YyRI+Ff9kVEREREREQWhi/7IiIiIiIiIgtjx6n3dkqXsoLXGbLHUJmTTz55KjPMK0NlVqUi6ULD854Z/lQ1hnvwOxmq091/TiEf6+hx3HHHDXWnnXbaVO7CizP86UikHgxBy77zxhtvDHWpx6owzC6NzFL0OOOMM6ZyZ4fgv+vCnRm69LOf/Wwqv/zyy0NdhkC9l7DY/dbjvYTFdvPVmWeeOZUZxp9hmSeccMJQ19ll3nzzzaHu1VdfncovvfTSULfO+NhuXcdOUsOs+z3r6MHQ17POOuuIn6sawwAZEsjxkW3JMfDiiy9OZeqR2q2TVmlueux0PaceOT44J51++ulTmeODYynX3kOHDg11zz777FSmHm+//fZUnkOq0N343lWWue3qkXutqqpzzjlnKnN8cH3PMOXHH398qMvrV155Zag7aD324h6d1a5q1CND86vG+Sq1qaq69NJLp/Kpp5461HF85Lp87733DnWPPvroVOb4SB33ygaW38O+Ojc9PvGJT0zls88+e6i76qqrpvLHP/7xoY6Wseeee24q33HHHUPdgw8+OJUZUp4WjP0I4++s0u+FHHfc61CP1IDz1XnnnTeVzz///KFuHT2eeuqpqfyDH/xgqLvvvvumMuerfD+ktWwvMPWeiIiIiIiIiPiyLyIiIiIiIrI0fNkXERERERERWRg7Tr23XTqPX5depGpMk5CeparRx0fPZfoh6fvufIX02WbKBvr/ujRX9J6kx2kv/E3rsFt6pG+vatSHvkqmeEmYFq5LQ5a/Of0zVaPO1IO/az88NdvlveiRfslzzz13qEvfUo6VqrE/cux0afro08pzE374wx8OdZnGhymXOv/XnM5MqOo9ydQj+z3Hx0UXXTSV019WNfpcmTqJ7ZHPk2cmVFU9+eSTU5l65JjgmSLrpC7drj57peM6eqT3m3pccsklU/myyy4b6jrPPueWhJ79hx9+eCpzHcrvYfqfrs3npkeyar5KPeg7Tg2uvfbaoS7PH2HqJKZsy76dHv2qcWytkyq080futF33Qw+OD/rpc7+TbVw1+sBvuummoS7nL+rBds21P8/1qRrXd/qVc82gR7xjp+26Fx5kXnfp26rG/klvcerxpS99aajLscPzFTi3pPebvudc+zmu8ntW7Z/2IpXufux3mfIz9eBZCBdffPFU/spXvjLUXX755VOZe2Gel5PzIJ8nufPOO4frTo+9eMfYrXMzOj343sT9f76PUY8LLrhgKn/5y18e6nJ88B2P5D2oVa4LPMOtSxW6H3psB/+yLyIiIiIiIrIwfNkXERERERERWRi+7IuIiIiIiIgsjG179neL9JF1Ob2rRq9xeiyrqk455ZSpTI9d+hnoa6WfIj1n9NKmz/aee+4Z6p5//vmpTJ8zn2duPuRkHT3SR5aeparRs09fTn7PKj3S48RzGlJzemc7PXiPvWC3NO704O9KPdLTV1V14YUXTmX68tNHSe8R/ZGpB/2AeU4A/cqpB/vRfuSFPWg96AO/4oorpnL246qxzTOfcdXhfTd9lpnrt2r0m2X7V4053/k75jw/kXX0SF8f56v0hTNPct4j831XHe6BTd8t9cgc1zzzJfNYr8opnddz02qd9SPX2pyfqkY9qFXqSj3oEc/P8pyG9LbSz//qq69O5VV67Mf8tVNSA7Y/x0eeR8Hc1FdfffVUTj9s1bievPXWW0Mdx0e2Hc9pyH+b541Ujd7yVTm+56QH+0qnR84PVeP8zTN4rrzyyqnM+SrnoDyvqupwPXJvyrN8cr/16KOPDnWvv/76VN7puSFzYB090rPPuT3nKK7neY+f/vSnQx3fDXLtpw899xM8XyF1XpVzfc76dHrwzJfUg2eMpD4f+chHhrqcH3h+GPe7uUbwvJ7cC/O8jTm38T/jX/ZFREREREREFoYv+yIiIiIiIiILY9/D+BOGlTEtQoYVffKTnxzqMoz78ccfH+rymilEGPKSIR4MncqwEaaYyfCTdcLK5hzuQT0YOpShMzfccMNQl+HGTz311FB3//33T2WG0TD0M8POGGaWqZsYYtOliOvC+DdJD4YnZfswXVWGGTGk+957753KtFVQj7QHsM0zzIwpZjL0ctX4mHOYcrJKjwzny7D9qtECkSHDVWMaSYb9UY8Mr82wz6o+RVmOj3XCkjdJD9pVsg3SksU66pHzVaaXPNI9r7rqqqlMK02GFjKl4jp6zFmDhGGY1CP7Ky1zWZeWk6qqhx56aCpTD4Z6ZrgzQ9NTDz7bXqSZOgjyWRn+ztRi2QYMi806WlAee+yxqcz5iumy0s7UhY1zXKUGcw7bXwfuPRkKnH2Za22uA9xffec735nK3O+yn6d9hnMSbZNJp8dejI+9SlHdfS/7YM5n1CrDvalHpr1lmHi3ZnN8Zlg/0+vlnna30hR27NUc2M27nL+6fXzauWgJogU7YTrQHHe0deSegbrud/rondzDv+yLiIiIiIiILAxf9kVEREREREQWhi/7IiIiIiIiIgvjQD37hJ6VTPlCf1Hn2X/uueem8q9+9auhjr6c9OLQH5ue3PSLV43+K3pJNsnj10H/XZ5pQL9yevcefvjhoS49TUyZR59n3rM706HTY1M9faugHumxY12efUDP0iOPPDKVmXqPHths55tvvnmoyzFJPZJN9sB20MeXHjvWpafsrrvuGuoefPDBqcwzFKhrev95jkn6AbvUMEsZH+xHnC+yPei/y3nojjvuGOpy/qInmeeY5LkVN95441BHz2GSHsyl6sHfn32Zn82zXPIMi6rRs88zXzjvpC88z7fgPbsUWPvhgd0PVnndc7/DfVKmAKUnOfdbXD94dkt6YplSMc9f4P07PTaJbHPuE5kWL+u5DmSawjwzoWpMI8l2zPmpakx/yHRyOSaoa9YtRQ/64Nl2uUZwHchUhF0aT7YV2zzXc57TkGOQeuSzb5IefNbs8/TB8zfnnibbv6o/YyS16/ZlVeN6wr3wK6+8suWzsS/tNavSLR4J/7IvIiIiIiIisjB82RcRERERERFZGPsexp9hHAybZ6h+huQxhUimV2BqsQx/YbgewzgyrIchmhmqzmfLEI9VYX9zDrPpUvUwFWIXptzp8fbbb09lhq51KZB4/9SHKUwYptuxqXqwrTIEjCFHGYr8wgsvDHVvvvnmVGboGvt52iwYopnjg7YOju1NJfVYlTqp0yNDYalHhrPS5sJ+naHQ1CPvuU77b+p8RT3WCePPOSrD86rGEMFVqazyHhw7GYrLML85t/E6dGkzqU/O3+yfqQFTIWYIM+crjoEcg9Qqv4fhm0vRo0sBxTD+bs3MOYnpDrMd2a+ZyirDxjlf5l6AeizR2sLf1KXvYntkm+d+inWE6/LZZ589lblG5drTjQ+O600aO/msq2wV+e5A+1C3h0p9OAdxvko9uNbkPfkes0lt3tHp0dkqOAZyX8T+me+DuT5UHZ6iOFMh0g6Q38Oxu996mHpPRERERERERHzZFxEREREREVkavuyLiIiIiIiILIx99+ynD5meMfq90iPOFG2ZToHejoQ+TqayOuWUU6Yy/RuZtoT+jS6t0k7SIhwU+Ts6z2vVmBqEn01/Kn3H2R70idHTlB4/6pH35/fweba6/9xJPdjneYZBpglh2socE/Tf5T3476j5OeecM5Xpd8rnoQeXz55Qjzn7M/NZ+Rs5l6Qnmed/pB+N3sBsK3oscw6sqrrwwgunMsdHak7/W2rO9t8kD2Y+6yrPfurDfp2eP/qOc11iP875qarqoosumspMA5fef96j02POrPLlJ1wjc85m36UHM0mvN+f99LxWjXpwLKXPlh7Yrl/NeXx06Q5Xefbzsxwf+Zs5BnLN5pi79NJLh+tLLrlky8++/PLLU5lrVKfHnNqfdOOD7c85ujs/KuekdVJEX3PNNcN16sOzGPIcE3qSN5V19OAcnXME+25+D++R8xXfaa688srh+vzzz5/KmWK0quqll17a8tk2lW7scjyQ1KNLd96dTUI9LrvssuE6UyPyne+1117b8h6bgH/ZFxEREREREVkYvuyLiIiIiIiILIx9D+PP8BeGEGdIfdXhYXhJhqAx5VGGezDUluGDGdbBtBgZLsWwv60+V7VZYZn5rKvaKusZTplQxwx5oeYMU85QcYay5bPSutGldNwk8jcyVVGX/pHaZdhZphOpGvXgv+NnM7yTmuezZlqSqjHMinowJK0LiTto8tloO+L8lO3Dfp5aMvQ4oR4ZVlY19gGGFqZ9Jm1OVaPmnQVpbnT9gf2qs/Z0IeXUo5uvzjvvvOG6sxYl1CPHB59tVTjjnOhShTL8O8cS19NsO/b5nC845tJmVDWuH5w/8/5MpZWab9L6TdYZL6kXw+hz3qd1Jfsn1yTqkVavLmyd63mnxyaF9XdwvOTvoBUy91TcM231uarD9eCY2Iou1dsmt3/3rPxdOe8wLV5CC0yuGRw71CP3FLx/zp9d6tZN1mMdsk9yfGT70MqS7ZPW7KrD15puz5B1hvGLiIiIiIiIyIHjy76IiIiIiIjIwvBlX0RERERERGRh7LtnP6EPhb6tLv1LpqDKtG9Vow+f6cLoWUr/E/1nCf2x6Yuhl7dLW9L51jr2ypeT38OUYPTFJPwd6WXlv0sfeHpcqw73uabfhmn5EtZl+1CrdfTYbrvulR7ZH+ij7M4poKcsvUn0p+b46PxmVeN4oZ88xy+/J8crx26XRobj4aD9Z50evE5PGT3a6V2lbyzbju3IOSl9ZBw7eX9+T/p16d3tzrhg+x+0Ht18xevs92+99dZQl23HtupS9lG7bh3I8crzR7LN6Q3svLxz8wrms9Hny+tMr9eND/ryM50e9eB5Czmfcy7L8cq5LNuc7d+1+UGPB5LPw7WN62DqwT1UrsPs1+lzpX/8rLPOGq67M2hSD67Z66SmnPOZL50eXAfz3BuegZNzCftn7ql4pgjnq5x3OD66VHN5T/6OTfKMd3Mp9ch9LMfHz3/+86nMvWj28+4MpKqxnbv9FfdQqcfc1oR16PoKf1f2T+6Fc63vzlbi+yDbPP9tlz65O6dtruPBv+yLiIiIiIiILAxf9kVEREREREQWhi/7IiIiIiIiIgtj3z376V+gR+bVV18drl988cWpzHyIF1xwwVTuPI/0tdLXmdCnRM9Gkj4dejTo50ivXJcztvN27JXvI7+Xnr6XX355uH7uueemMn0x6Q3r8oVSK/r7U69Vec2T1Ir37/KK79RfcxB6vPTSS8P1oUOHpjL9kNlWHDupB/sqfbY5JvjZvAfbMT24q/RIL9RB+506j/qq8fHEE09MZc4lCT1+J5100lTmfNXleqWP74QTTpjKnVZvvvnmULfTOekg6Dywb7zxxnD9+OOPT2XqkV49rgnpV+acQx9hzlFs89SVc1neP/2fVf0YOOjxQTo90hNeVfXMM89MZbZH5gfnHJR1na+1apwHOT5SD/qV85r9YSl6sJ/lesL2yDNg2OfpQ06oR447zm3pX+acmJ/lvDu3Nu/IZ+Nz8syXPFckx0rV2K5c6zN3O8/u6dqO+9vUgPdIHbuc75sEn5u/K9uS7yYJ14h8N+FZMdQj1wzmh08NOF/mv9vpuVNzg8/NeSfbjmdavPDCC1OZ5yucf/75U/mVV14Z6rr5iucr5JzUnWkx1zMU/Mu+iIiIiIiIyMLwZV9ERERERERkYex5GH8XUsJwvddff324/vGPfzyVGZ7EsLOt7skUPwyrYejMVjAkMEOwVqW42Wk6t46dhup0YcpdqqSqqjvvvHMqM9QxQ5mYFiNhKGGXNpGhMnndpbmiHuuEOW03hHm3Qgm7f8d+xL58zz33bPnZDKOnVSLDjFjH0MLLLrtsKnPMZVhTl1ZpP9K37UdoJ9uYoWQPPvjgVGafSz0Y5pUhYKueO9Mssc1zTDIsNu+56h5zDgPsbGCcW55++uktvyctD5zbc95jn2e7Zuo3hinnv+U6s461aFNgn+dckrYX/v5cl9mvc95j6iSmfrv88sunMteIHINMadmln9wkm0vSpa6qGts1LWFV49zGcO9Mc0WN2a7ZPtQ17RkcH531b27pWbfLqufOtfi1114b6nJ/wzbOdYltzD0c65NMm8jQ9LR+dalByZy1WbUvyXZl+uJce3JeqRrnktzPHul7co3g96Rdhrp1qVvJXMPKV0E9cn3hvJP9k3U5l2TK9qrD97+pB8dZavnQQw8Ndeukbj2oMeFf9kVEREREREQWhi/7IiIiIiIiIgvDl30RERERERGRhXGgqffo8aMH9vbbb5/K9Lrw3ybpp6G3nB6N9GXQa57XvH96ROiFo5d0uxy0v4neEnpg0yNOX0z+ZnpgO69N51ujRzr/Lc9eSD3477q+QuaUeo96sH+mR5x9Lvs9/V55D34nU1t16Q6znZn2LP2gfLZufOzF2QfvhS51EvtZesSpXf7mLqUnv5O+sTPPPHNb/5bnO+T8tepMizl7kjs9+DsyXRJ915naim2c0H+XPuOq8UyLLqUkz5zJcce+skkey3XS0OX6To98zh+df/v5558f6vLMhKqqT3/601OZnv3Uh2t2jp1VKaCSuY2Pdci1lutprv30EqcPnPsittUXv/jFqcy1hWdcdN+TsM3nPF+tQ64RXQq9bn/DccUzLb7whS9MZa7tXaq3bONNPVNkFfxd2QfZ5gnn/dyLMoUiz5G5/vrrpzLPg0kNuIfLe66ad5cyPrq9cerTvfMxhSJT8eV6zvkqzwyjHrlGrXrfOCg9/Mu+iIiIiIiIyMLwZV9ERERERERkYex7GH/CUDqGAmfYX4bGVPWhEBnKxJBhhn984hOfOOL9eM3wwQxlY9gI6UJPt8tepRbL71mV6i1D9ljXhRVliAv1YHq91I6Wi2xzpgrKFH7sV12KlbmFNXWpxRgymaF+TGGYejC0L8PDGHLE0L4MW2Y/z9DP5557bstn7ULgquanwVasslWkzYRhmKkHbS4ZEsa2yBRxVWN6UvaPDEl76aWXhrp8np2mBp0bq2wVec35IscEQ/ky/Q7nsm7N4vNkHdPKZt+hjkvRg78r5xq2R44PpgtLPbheMBSca3iS8xXXr5zb1rF9zZlVdoSs59qSenAdznBjWv0yNWjVuG8755xzhrpcs3iPHLtL0WMVnd0x26OzxVFzrgM5Bri2pAac57LvbJLNaB3WSYPcpbLN9Z37IuqRc113j44ubP9I9ZtKZwFPuE/NOYrtz+tcF7i/S125h1snla6p90RERERERERkV/BlX0RERERERGRh+LIvIiIiIiIisjD23bPfpbOgFym9DvTAbtd33flpqvrUYvms9ESnZ4Temk1N9dZ5LKtGPegp6+hStdBzSZ9M0qUWy2ddJ3XS3Oh8Sd3v6HzxbP/0F9FrxO/JerZreqGoR+dpOxr06OYynqGQXknOT7xOTyzvnx5xzlc7TR82Z7/fqpR12eZdHds4/apMQUWPcvqQ2T9efvnlLf/dEvXozkapGtuc80ynVY6l9O9XHe7PvOCCC47476qqXnzxxanc7SfInNt8HdY5u6bzoCacy0ieicQ9Q46Pbj8xF8/rbtP9jnXW+twzcf/Etsu0lhwDuX5069dSU++RTp+c6znv5/rRnUVRNa49TH+Zazj7Q3dG1tFId6ZEtjnHB/XJfs9zTLo5KudBzp1z2e/6l30RERERERGRheHLvoiIiIiIiMjC2Pcw/gyNWSe8gaEyu5U+Lf9tl2aJoRld+Nqmhp2tCsNMujQ+pAsz6lJYMMSf6WC2usc6qUjmrM06YbHd+CBdCFimgasaU10x7VWG0DLsr3uWTdWD7DQMswvJWzUn5njhnJR6dCFnR2Oqnq6uS0XYpRGt6sM5M2SQ4YJdnz/a9ehCiDk+PvShDw3XXZvnNcdHZztaqh4d2R5dutwuTLxqtK+w3VKrTo/OgnOk710inR4J1+iPfexjw3XqwTD+/N4u9d7R0N7r0Nl2uZ/q9KD1Lt8xOguheox0+yvawLh+pAa03lHLre451/nKv+yLiIiIiIiILAxf9kVEREREREQWhi/7IiIiIiIiIgtj3z37yTrehZ1+ln4Jeu0//OEPT2Wm8cnvobcjvR/8zp2mVToI9uN5Oj3oMTvhhBOmMtsxP5u6VY0psuir7dLSbTI71S7blb5veppOPfXUI/67qlEPprBMDTg+6Dmc25jYb7JdOT4475x22mlTuUtfxn+XOh+NeqzyYSepB72S9CTn+KBHPH2vPH+kS3+5zvkbm8pO9WBf5TqQ+rz99ttDXXowuUZ0qTEdHyNd+mTOO3nN1GLpGWebH+16kC71XXfeBdfz9B1nKsqq8QwFzkl5z7mmFttPOj26scN9Uo4PnqGQ60eXPnmd94+l0vXPLi0e1/McL3wfzLHF71lnf3VQ+Jd9ERERERERkYXhy76IiIiIiIjIwvBlX0RERERERGRh7Ltnf6f+qs4H032W3gp69TpP2fPPPz+V089UNXqf+J1Ho6dsu9DPQi9rembeeeedoa7LK5866vEb6cYOz0zozkJ49913h7rOA5vfw/yx1C6f72jTpmr8/RwPeYZF1TifdR4/5oTN8UFvefc8S9GjGwOd55Ie5BNPPHG4zvahHvk99M7mNXOMd3mbl8p29eH8dNJJJw3X6bNk3805iuMjr5lj/Ghku55kepA5X2Vb0gOb6zT1yHVnlR5LnK86Oj0++tGPDnWcr3JuYbtuVw/qeLTTecSPP/74oY4e8e78i4R6cL91tNO982XfPfnkk4c66pH/lu8NOXa4b+Y7RzKXOcm/7IuIiIiIiIgsDF/2RURERERERBbGxsSCrJMapvtchnSQV199dbjO0AyGaHapxY4G1tGjC7HpwpOoR96T6UXye9Sj1yM1YDhYlxqGNpc33nhjKjOsKcOUqfnRQKdHlw6U7cgws9SDqcXSZtGFKa/SYy5hZ7vJOuMjxwTXi4997GPDdbbrT3/606EurV8M48/vXceidjTQrdlsR+qR2nF8pB7U1bDYreF8kVYjrhfUI7Xs9OCandccu6uul0A3X1GPXBNWhY1nKDLnq7TlrZM++mjUI+nSeR977LFDHcdLhopzfKT9sbOiHo16dHAPlRrQ5kI9cny8+eabQ13qQStL/juOnbnocfTtwkVEREREREQWji/7IiIiIiIiIgvDl30RERERERGRhbGxRrX0PXSpSFb5tzPt0QMPPDDUpYfmlVde2fLfMVUSPVWdv2a78DfOzYfT6dF5xPnZ9B3fe++9Q136Nek3y7QlXWq33WLuenSkHvSusu9mv7/nnnuGuvT4MTVl52FSj5Gco5jqjXo88cQTU5nniLz44otTmR6/ZJUeu5HKapP06OYrpnpjeqSco+j/e+2116Zy57ncDz3mTve7uhSGOQdVVd1xxx1TmVqlB5NrRGrQ+ZX5PEeDHt1vZKrQl19+ebj+7ne/O5Vff/31oS7PfGEauE6PTZpbdotuvujm+scff3y4zjWDdXlGEsdOt54fjaQebP98N2C/vu+++4brZ599dipzf5VjielZ855HY6pWjvnsk2zzfKdg2uG77757uM692G233TbUvfDCC1OZ+91N0MO/7IuIiIiIiIgsDF/2RURERERERBbGxobxJ10YF0OOGOKR4eCPPvroUJepxhgu+M4772x5j1WpF7ZLl65rk0LXsj0Y8sSwmkOHDm1Zl2GyGQJYNeqxV6kuNkmP7nkyzIjjge364IMPTuUMY+I9WNfpsVthgJuqRxeG2Y2Hqqof/ehHU5kWjAxXW0ePo318UI8MYc10O1WjjYLwe3IspcWi6vD1ZKtnO9L1dun0mFsobqdHpjmifYvWu6xnG2dYbFosqg4fd1s925Gut0tnX5rz+GBYaoawMj1u2iiqxjBlWiFTD6YdW0ePnbKpenAPlf38mWeeGeq++c1vDteZmu+5554b6l566aWpnGtJ1eFh5HvBJunR7WlzDnrkkUeGOlrv0iZG7XJscTww9dtesEl65PNwT5tz/Z133jnUcV3OtH3UI9dzjodOj71oq51YYf3LvoiIiIiIiMjC8GVfREREREREZGH4si8iIiIiIiKyMP7on7ZpKNiLdFl7RedVZCq+TAXHNDLpiWUzpW+Nfg163Haaei+fnSnrmBplzmQqK+qRHpmqsc0/+MEPDnXZBvRJpR700+yWHgmfe5P0yDHA1G70gWeqK46PhD6p/RgfyVL04DjnGEiPHz+b7cox0KUD6tLC7ZRN1iPblW3M1IipB+e2bGfqkXXUoxsfO4W/o0vXNTeyL7FfMTVizlf0/qcGnJOyf7Jt1GMkNeCaQD1yPeFvTA2oR36W/263zkRKNlmPbs/E+Sq147zTrRE5Brq0lVXq0enB1KG59nMMdGtEtvmq+Wk39OB701zTyx2J1IPzFfXJ/XC3RnA/06UK3YvzcajHdsaHf9kXERERERERWRi+7IuIiIiIiIgsjEWG8WcoBkMout/BEI8MjehSSbGOIRVZ3z0PpejC3zcpjCZDTlbpkW3AkKcMnelSSalHT6cHw/qzDRiimWF/66Ra60KO+O+OBj0yZJHP3enx0Y9+dKjL9DzqsXMy1HUdPTI1aNX29WA7dm21jh6blHqvowsFZzhj/q519OjaYx09OtRjaz3IdlPHrvPvyNGoR7YPQ/x3a3x0c1LH0ahH/i7qkVZI0u131WNkp3owxL9LP7lO+ved0r27buce/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYs/bs74bXZJ3Pdv+uS2exG6ktVrHqN+3HM+yUnfadzh/b/V712BuoRzI3PTbZY7ZduvFB9rs/rtJjv+fP/WCd9WO/ORr1mDNsf47lnK/UY+9ZlaI5fdDqsfes0qMbH+qzfbb7jkU9urS/6rFzdqqHnn0RERERERGRoxBf9kVEREREREQWhi/7IiIiIiIiIgvjT1Z/ZPNYJ494kvmVq8acjPye9FOsk9N6p2yy72W7uaB5vY4vqMs7qh4je6EHOej2Oej77xWpxyb5fI8GH99+zDs7ZdWzzOlZd4tN0uNoGB9zpttPHKle9pZV42GJZ/AcBNvt16vaXz12h53qsR38y76IiIiIiIjIwvBlX0RERERERGRhzDqMv0tllCGs/NwHP/jBqczQ4+OOO264PuaYY6YyQ1EyjP/nP//5UJfXv/71r4e6vQj5mkNI4nbDv/msaY+gHh/60Ie2vOb3pB6//OUvh7pf/epXU/k3v/nNULdUPbZLl8aGKW0+8IEPbHnd2Vyy/auqfvvb307l3/3ud0PdXoR8bZIeJOcyhuZzvKQG1CNtFRwDqRXtF5vUVvtBN5dRn24sZbtm+1eNY2A/QoYPInXuXrDqd3RjKenafD9C6peixyp2mtp4u3W7xdGiR7JJqULlYFGPebETPfzLvoiIiIiIiMjC8GVfREREREREZGH4si8iIiIiIiKyMDbGs995WdOjX1X10Y9+dCqfeuqpQ91pp502XJ900klTmb78t99+eyo/99xzQ1369Duv5ioO2hu1U6hHXtNLnD78E088caj7+Mc/PlyffPLJU5le77feemsqv/DCC0Ndfpb+2KXq0fl2sq7z5fMMC+qTevB+P/3pT6fyiy++ONT97Gc/m8rreMR32v5zT13VeVff//73T2WeYZFzWVXV8ccfP5Wp67vvvjuVX3/99aGOZyokR7uHn3rk2sK5jGdafOQjH5nK1CPPTcjxUHX43JbshYd/VYrNOZNrC9uY61COpe6MEZ5pkWNgP85Q2OQxtl0fftWoF88fyXZeZ83eDz02aXwkq56721Pv9NyK/dBjk8dLRzeW9jIN27ostf075nymhan3RERERERERMSXfREREREREZGlcaBh/OuEgDEk78Mf/vBUZqhrhuafeeaZQ91FF100XGeY8vPPPz/UZahfl76M4WkMj8pwNYbMbpeDSC22zm/O8FaGIp9wwglT+ayzzhrqrrzyyuH69NNPn8oM1X/88cen8muvvTbU5fOs0qML2dwuc0j11qWZynBWhh4fe+yxU5nj49prrx2uzzjjjKn8yiuvDHWPPfbYVE7LS9WYGpFtvI4e2a7btS3w3+0XXTrQLv1kZzvi+EjbC0P1n3nmmanM1JQZNr4qZDX1WCeccjdCEneT7nm6lHk5f6Vtoqrq3HPPHa5zbmOofs5R7Ndpueieu6rXY7vMISx5u6l0u7ks1/2qca2vGrXL9J9VVb/4xS+mclrCqg4fLx1duPN2mYMeCZ+nGzs5f6U2VYev/VnfpTZObarGvdeqPdNuzC1z04PsdC7jvjm16/Y+HDt5zX+3H3N91z8Pem0h3Vy26nqre9DmwuvkoNvjoEPc15nLVl1vVccx0KXSnQv+ZV9ERERERERkYfiyLyIiIiIiIrIwfNkXERERERERWRiz8ux3Kanoy8/r9E1WVZ199tlT+brrrhvq0hNetX1vGL1P6cvoPIb8LFMupRfqoNO/rKMHvZN5zXRuqceNN9441F1wwQVbPg894nn/Y445ZqjLtqMe1C51pvdpHW/zXrNKj/xdbI/0TnLspE+felx22WXDdfr9M9Ve1Zjykv3hzTffnMqrfGrpM1zHm5YcxBkW63hZUx/6WjMd6DXXXDPUXX755cN1annPPfdsWZcp4arGtKKrPJd5zTkx/+066Zj2w8dJunM8sl9z7OQ5Lmz/8847b7jONn/kkUeGuuy79PPnurNO23RewYNmnTN46C3e7hkjbH+ecZHz0EsvvTTUvfHGG1OZ80r280yreyS6Nt9puqz9WFt4j+2ek9Cl0mUqY641+W95LsI777wzldkfso5pEnej/Xfr370Xunm3SzXdpQPlXpjrUH4v2zH14ZkiuX7Qz78fc9B+pMMk3Rk83flVqQ/3RdQu4T1yL8qxk9d8pziIVHz7vf9a59yKbl9G7fJ3sC7XCKYyzjVjrmkj/cu+iIiIiIiIyMLwZV9ERERERERkYex7GP92QzGqxnCxDOWrGsP3zjnnnKHu/PPPn8oMS2Y4UKZze/HFF4e6DAPMMKaqMRyHoTkMC817MhwnQ0M6GwHbZrdCp7rQMV6nHgzVz5RgGbZfNaY7vPnmm4c6tt2TTz45lV9++eWhLvVhqp4MZevSzVSNYTWdHl3o2F7pkbyX8dHpcckll0zlL3zhC0Mdw78znRtTveU1Q19TD7Y/yXbl92w3pUmXzm+36MJgq8Zxz3DWnK8ynWFV1cUXXzyVP/vZzw51DMvMOYl9N0PFGTqW4Wtdajf+2852RDorzX6nY6oa+yD7depBa1fOVzfccMNQRz3SrpJzV9XYrhwDqQfbpgvLZPj5TtPAHYQeXXhr6pFzV9WY7pC2io997GPDdc4fnC9yzeAanes7xwP12G4Kpu2mqVz12Z3yXmxgqQfTG2b63FxLqg6f97I9MlVr1The2MY5t1GPdcbAdtflOeiRc0Ra5KqqTjnllKl84oknDnU5f9EWyXkvefbZZ4frXOuZ9jitFF37V/Vp2OZks1jHdkRrUY4JrglpbeG7CS18qTn3u08//fRUpiUp12Hq0f2u3WrHg9aD4fipB9Pl5thhqmmuQ6lzru1V43xFPVK7/UgVuhP8y76IiIiIiIjIwvBlX0RERERERGRh+LIvIiIiIiIisjAONPVel16kavS30FOWvqX0vFZVXXjhhVOZnr6HH354uH7iiSemMn1KmWqM3q98Nj43/Vb0lyfb9dfshQd51bPQ+57+Fnpd0jNz5ZVXDnVXXXXVVKZfmXqkr48eJp6bkHTeNOrTfc922Ss9tpt+p2rUgL8/fUrXX3/9UPfJT35yKuf5FlWH+47zTAt6mNIfy2dL7+aq9Hnd+Ngue5X+pztjpJuvOj2Y7jB9+jxfgXNSXjP9S7YB56A806E7p+JI37sT9sqXto4euWZ0Z1p8+tOfHuryHAv6xzknpe+VZ2qk/4/z5TopPpl6bCcchB5sj+yT3Zkv1OPzn//8VKZfmeeIpJeSXtpcz5laLPv8Kk8y/eXdZ9/r594Lq/TI/kk90nf8qU99aqjLc3cyTWXVmN6wamxztmt3llGuCWxvzvU5f83JE05WnfmSenDfmvumXL+rxvFCPTKFYdW4ZueaxHvy7JzcM3E+2uk5CQfNKj3SF04f+Cc+8YmpTD3ymnqwn2dbMbVx7gX+4R/+YajLz/Icnf1I/XYQZ1qkHlwH8mwE6pHn7lAP9uUuVWi+K/7n//yfh7q33357Kq/S46DwL/siIiIiIiIiC8OXfREREREREZGF4cu+iIiIiIiIyMLYc89+54tZx5PMXK/pmWEuy/Sb0XeRuUSrRh8G/U35fPTA5jVzZ/J3pVeNvrlsH/7+9POwHXfLF5XPymfjdadHapB5qllHP3DmEq0aPeL0VWY7M+9pth3PGqD/LK+pVeeZT+8N9VjlS98u3figHtke9CKlF/+KK64Y6lIf6vHUU08N1+lJ5m/MPsBzELJ/sq/Sm5ZjtPNtraPHbp2p0N2/04N+yGxznqGQuarZ5zk+0rPPXLM5zuhlzjbnXJZ+s6rt527vzhvZq7zVnR687vS47LLLpvJNN9001OWZL2xH6vHKK69MZZ4LkLpSq+yf9CR33uad5m7fDz1WeWDzDAXOVzlHpUe/quq8886bytneVVWHDh0art96662pTF9nrl+cH3L+4hki1Ge74+MgWGd85DzAsymuueaaqfzlL395qMv9Fc+wePHFF4frXGupeerBNs/zYTjPdnrMjXXGR+px+umnD3XpQ/6zP/uzoS77eZf/m8/AOTHPmeE4y+/hc3Msbcr4WOURz/mKZ03lmvHVr351qMuzDzgecn7iPXJc8Xv43pL7gJ/97GfVMTcNkp2uH2edddZQl2vGl770paEuz4+iHtxvZZufeuqpQ12ea/LQQw8NdblP7vZFB4l/2RcRERERERFZGL7si4iIiIiIiCyMA029x9BGhsNnuoszzzxzqMvQPobxZxgFw5oytVvVGC7G8LAMq2IIWj4bw++ZziFDaLuQJ4Y770U4VBdiwtB46pGpjNJGUTWGjZ977rlDXT57hrtUVd1///3D9WuvvTaVmcKiC3nKZ6WOTBmXYcvrhKDNTY8Mo+cY6MZHtg/Hw49+9KPhOsOe2Fb5PLR1ZPswbJ/6ZBhaF8JM9js8jXp06UCz/XnNELQMYeV4uOOOO4br7Lu8f4Zhds/KOZEWjAyb7eYkst960K5De0LO0RmaXzXaKhjCnKH7d95551DH8ZGhhlwjcr7k2EnrCkMJSc4R6+ixk8+9F6gH+2C2x6WXXjrUZbpWht8///zzU5l6PProo8N1hlpy3kt9cp3hPZg+bj/W5b2gS7VXNY4PWr2uu+66qUx7Slrt7rrrrqGOYbI5tqh5hqpz/ci9Idt/TmvCOqzSI8dHpiuuGq1fnOfuu+++qXzPPfcMdVx7c0wwvV+OO6at7OagTUm1R6gH30eyDWi9y3SU3P/nmv2Tn/xkqKOlNNch6pFzWYaiV23/vWGT6FIfVo3zFdPrpa2C33PLLbdM5XvvvXeo47p8+eWXT+UvfvGLQ11qwFS6XfrPuejhX/ZFREREREREFoYv+yIiIiIiIiILw5d9ERERERERkYWx7579LtUCPUzp46LP9YILLpjK9E9k+rDbb799qMtUe1WjX7VL9cAzA9L/R28505Z06XCyrvN68N/tFp0e9IinT55+yEwzRS9xpg353ve+N9Q9/PDDw3XqQc9O6nz22WcPdakHPclMTZLtSt/WVp/j9V7pkfAe7OeZDoYe8fQe0eOX4+M73/nOUEc90mNG31h6yqhH+sfpq3311VeH653qsR9046PTg+dWpF+VHr88N+HWW2/dsq5qbAOeW5HXHLvpx2SaK/qXu5RtW31uv+jmUv7mnLM5PtLDT8/jAw88MJV5ZgJT7+UYoNc81y+ubenTzzRKVf28M9e0PlV96qqqcXwwPWvqwzMM0of84x//eKjjXJLjjvfINYLreY6zvUpTuB908xXXgWyDiy++eKjLM3nYxnluBT3i1C7n82uvvXaoy/HBNHD5rKs8sJuiz6r9bp5hkPupqnFu57lHuWZw/eb5Udmu6TuvGjXgXJZr1lL0IJ0eHB/ZPg8++OBQl3uqPN+i6vA+kHsqzpc5Pujn794blgL16OarbEeu2d/61rem8pNPPjnU8ZyZPBegO+OEe79NOMfFv+yLiIiIiIiILAxf9kVEREREREQWxoGG8TOcNcMkqqpOPfXUqcww4QzrZ/qEDHNi2CVTX2SYGe+f9+jun6maqqreeuut2i7bTfXG37hbpB4MaWH6lS5MOUNs+DsydIZhTZmCqmrUg6ms0jqQNo6qMcUP07cx7C3Ddneaxmev0s2kHrQxsD0y1JJhyhn2R5tJhtUzTJwh3vkMDLXMPsCwwwxzoq2C4z779k712KtUiJ0eTBeVenC+yLA/hrpm6CXHBz+b4WoMtcwxyL6Sobi0VdA6kRrstF33IzUl+xHnq5yjaQPLdGJMtfbQQw9NZa4fHB95T94/9XjnnXeGupyT+DvIJoQIVh3ej9LiUDVqwDD6DJk8dOjQUJd6ZIq8qsPni2xL6pFzIq0TGUK7Kox/U/SgzYU2rNQj1/aqcU5m6GvOVy+//PJQx76cYeOcL3M9YQjzftjk9ptVtrxcMzKEvGq0NzJUP9dwWkjZrl066Rwv3AsuUY9VtqPcbzI9a64ZTK+XVmG+G3AM5j25v8r5c84pcHeLVbaj3G/mu2HVOJ8zHWjOX0zDTT3SCsjxkVrxPXIT1oTljWARERERERGRoxxf9kVEREREREQWhi/7IiIiIiIiIgtjzz379GGkr49+CXq60ofMNDr5PfRVpkeYqUeYwiJ9MfTlXH311VM5U2fxWXl/pqrJ3/n2228Pdb/61a+mMr0e6ZtjO+7Uw9/pkenSqg5vj/PPP38q06OdfmZ6/NJHxuemrzL1oY/wpptumspXXHHFUJeeT/pj6XFLL1DnTVsnpcxOfTrUIz2P9PRRjzy3gP0z/UXUI31k9OLRN5b68JyGm2++eSpzfKY3iuOa3vf8zV1qStbth28txwdTu3VnSjA1TGpJX36e8cF0L7xH9gH6Oq+55pqpzPNHch7kvEuv9XbTu62TBm639Mg+QI8lzzBIT3LOXVVjH6TvOP2x9A2y7VLz7gwHegXz/tR8lYd/TqQe/B1ca9MzzzMUEq6f6Y/k3MF75BzVrS1dKiu2P8dHngkztzR93RkjPEMhfa9cW/I38gyi3/3ud1OZ6ye/J/dwHB85tth3km5NmDtdv+JcknM92yrTB3MPmVqt0iP3DLxHl/Y26c61qTr4MdDR6cH9b7YP19NcM7jfTD1WnbuU+wTOVznOeO5S18abpEfC/te9H+aZO1Vj+uCf/vSnQ12nB8dA7qO5RuQYZDrvvTrDazfxL/siIiIiIiIiC8OXfREREREREZGFse9h/BmuxdCYCy+8cLjuUidl2ATDOfOaKdoYxpGh4gyLvfbaa6cyUz1kaAjDRvg9GbrDNGTJfqT46fRgSAv1yLZiarF8PuqaITcMN+fvyrAzpmfKMH4+a4beMvyH4b0ZrsXUKB37EQ7VpbpjaHjqkeHEVaPO7LupD+0QTCmSz8B7pM2FbZ4haF36n6rDx+RW7Ef7d7YK9iPOLTlH0fKQoai0p2QbM2SVqd7y3zLdYt4z279qtBHQjsB7bjeck+yHPvlsDOXjfJHzMPtuPmuG21eNafHYdxlOmRpwjcpxxjUi9eA9yJzClrsQXtqOOO90c3va2brxQc3Zd/N7uUZkiDPTs2a/WjUf5W+eW4hsPhv7FcOEs33SYlE19lfuZ1JXfmeXbpGfzWdN/atGPRiazs/OmfyNq2wuec2xk/2MYydtkvx31DXnOt4/27ULU+b6wLVmU2C/4tyS81m33+S83+0puWdIPbiHSusXrTRph+UcuAkh5UeCtgquJ7lv5DtGth31yLZj3+X7SO4TeP+cE2mlyTanHnuVNn1d/Mu+iIiIiIiIyMLwZV9ERERERERkYfiyLyIiIiIiIrIw9sSznz4leiTSJ0OPTOe17tIRMb1e+ms6Dxk/Sx9I1tGb98tf/nIq07PUeSz5O7pUb913ruMV7NKXpR5MPUI9Ui9+T/oc6VFJzx99z0xFkvdkXV7z97/77rtTOf1UVYfrk8/e6UFfZ95zt3y0nV+a3nZ6iPKaz5Na8h7p7ea5ALxH6sqxk1qyrdJvRn8Tx2v+5nXSwO2FHvyebDu2I/Xp0kelHvQEp4+PerB/pg+XZ2qkB5O+wfT+81wGknMC9cix3XkDdyv9T6cr5xm2VerBZ815hn0+fXw5rxzpHqkrz1BIrfg7Uo9uLuU1f/NenOuyU1Z5d7Mv0XedbUUvcZ4rwnvQd5trONePbGf+u/xeakw9si/xeQ5ag7w/59mu73DNzLHDOem6667b8v48JyC15F4sxxb/XT4r56BuL9j9xoMg78/zPrhm5jU987nWcB3uzjdgm6cenPdyTurOEaEenL+SOfvHqQevc/xQj5z36cPPM3HYjhxLuQ5Ru3wenrPT7em3u2eaG5xLOX/lOxfTHWa/vvzyy7es4zsNx1KOCfbzfB7qmhrMNTWlf9kXERERERERWRi+7IuIiIiIiIgsDF/2RURERERERBbGnnj22xuG34p+HnpW0ltKP016s+ipSy8x/X/0jGc9/V5vvPHGVKZ3Mz2H6U+uOjyncn6WPq1sA7bHXng76B9JXwrv1/ni2R7pKaIXLPMr03tED03+W3rz0lPGNn/11VenMnOSMld59qXuN3ft/1606TxV6f1hf+TvyP5JX3w+H9s49VjlQ0+96P9LfZ577rmh7uWXX57K1INeLI6JZD/02C58TrZ5+uRZl5pzTko9ulzUVaOWzLecPjLeP8cL/blsu+x3O/WbHYQenBOyD+ZYqRr7PdeEnK95jgv7bvrC6RHPdYlrQj47f0fX5nPx/x0Jei5zTq6qeuGFF6byiy++ONRlPnDqkf2R90gfZ1V/Xk9+T+dRJ5wj9+KskL2A/eq1114brg8dOjSVn3766aEuz9nhnNStURwfqQHbOLWkjtwnbHV/MufxQT1eeeWV4frZZ5+dyk888cRQ15171J1NwvGSXvPuPC2eqZH7ArZpNwbmpkfen32X4+OZZ56Zyo8++uhQl3pw3s82ph68zn7Otko9OD74zpPMrc23yyo9co6iHrl+5H6qalzD2f6cS3IPxXbL/S7PPerOEZmLHv5lX0RERERERGRh+LIvIiIiIiIisjD2PIyfIQsZRtGFiVeN6S6Y+iLDKJiGIcOTGDrG0KUMd2XYYT4fw6Ey5ITPxufpwpQzjKRLc/Ve6EKhu7Am/o4uVD61Yzhtl76Mz5PhlQxdylAy9pUM+WEdQ266PtilNOnSyKwTmrNdPdjnqEeGmTGULD/L35/9jJp3qaQY2pffQ4vB888/v+X9u9C2dfTYi7D+7t9xHLOfPf7441OZoa+Zcqcb16vCwzJMme2Rz8dw2hy7vEeXropa7Xdan3X0YB98+OGHp3K2W9U4Pjr7FOd2fjbDB5mOp5sv8v6dRY3P042P/QgP7OYr6sH54oEHHpjKXRq2LkybcyDbKv8t9cjnYT/OdadLiUbmFiLb6cHxcf/9909lWu+ynbt0qBwf5Oyzz57KnMu6uT3X/lWh0HNKP0k6Pbi/ST1oYzj//POnMu2uCe0p7OfnnHPOVO7GGetyzdgPu+l+wLmDeuR8RT0yzSrnmWxz7hE4ljINcgfD9rMvrbLfzpnu/WOn46NLG8k5kHNS6tq1I98r89nnqod/2RcRERERERFZGL7si4iIiIiIiCwMX/ZFREREREREFsaeePa361GgX4I+00yXxDQl6RuiFyp9ZPTI8NnSs5+ps6r6dCPpoaFPir+j86Hns++WJ7xjHT8J/S0vvfTSVE4/bNWYDoa+79Tj2GOPHeoyTUnV2B5d+hfW5bPSJ8U2T32o1X57b7p7sD/QH5mpejiWOq9ktg9T7fE6daWvM318TKWVevB8B46X1Jx9p/Pz77ce9JTxd6X3mtqlV5JtzHGW0H+W/5bnAqQemeasakz9xnmOeuTvPGh/7Drjg+kG2ZeSs846aypzDsp25Hd0aV7pc805iutXPmuXqrVq/8fAOnR60F+f6cTYVjmXcY3oUu9xfT/jjDOmMue9HJ/UI8+Z4frRnWMyZ6gH55lMZUUfeJ4Hw1ShnQ+d4yO9tNQj+3nuLarG8cEzRrozFOYM+w1/V44B6pFacXx0qRC5flx22WVTmWMnnyfTllaNY6JLH121OWco8Nm4F8w1lP06x8c6enCtueaaa6ZyNz6oR47l7oyXTYLPzXk32+Duu+8e6lIPppru9KB2119//VTu9OD5bjl25rZG/zP+ZV9ERERERERkYfiyLyIiIiIiIrIw9j31XhdqynQKmd6AdRmGybDYDCtiXZdGhiE2xx9//FRm6FKGXmbat6rDQ5ozTLZLs9SFfzCFym6lFsvQlEyXVnV4GEsXOpThemzjDKPpUkdVjaE7DPXMMPJOD4Y8UZ8M6+9C0jpbxV7RjQ8+a/ZPhgRmSBpD0DK0jyHDTOGX38N0I9k+tErkGGDIE8PGc0zsRTq990IX+ss2z3ZlKHCGzq+yTiQ5z/GeXd9NC1RV1XPPPTeVaT9g+GKX1uegyWdj2D5/R/Z7hspn2B/XltSR8xz1yPmK80WOCa4JObb5bF2a17np0c0B/B25nnAOeOqpp6ZypwfXT+rRpWxLWwHn1hwvnBPXSYV40KyTujXXV85ludZ2ViLOQWeeeeZwnfMg9ch5iHp0qSnnGiZ7JNZJhZh6PPTQQ0NdasB1OMcLNaceXcrR3Cd1trylhI3z93P9yH732GOPDXWpAfdX6+iRcz+fJ/cbtLnks67ap26KPqvC+HP85Ppd1Vsuuvnq9NNPH647PfK9gnvafNa5trd/2RcRERERERFZGL7si4iIiIiIiCwMX/ZFREREREREFsaee/bpe0hvWKa7qTrch5EeP6b9Sq9FprWqGr1P9JvRm5ZpGujVS+jXTQ/so48+OtRlmpSq0ZtG/9l2/bG75QPp9KCXl6TPkl779Hpn2req0VvOcxG676FnJ6/piU7PTqawqTrcs79OGpm9hrpm/6DHsjunoPMrf+ITnxjqUgOOK16fdtppU5npgNIzzTbO8cJUb0whmH2QfsCD1qPzZHdp2ehfzutTTjllqDvuuOOmMv2YvD711FO3rMt5h57L9CSzX3He69KBHrQfLZ+nO2ugajw7hL8x2yDnp6rR40f/Hz3iqQfPKsk5iXrkGOCzdR7lg25/0nmSOZZzXebYyTmZa0S2K/XgWEo9SOrB+aqbg3g9Z7pUoRzLufZyTsj5qjv3iOf6ZCrKqnH9IOmB5V6w88DuR4rivYDP1unTpQvmOpzacQ7i3JapKfk8OUexP8w5/edOWed3cK3J/skxkPtknjuVa33V+B7Dfp1naLE/bOoYWIdOH66RqQe1yvMm+L7BNH2dHvlewffIOa/R/4x/2RcRERERERFZGL7si4iIiIiIiCyMfQ/jz9AlhlswXVSG8zFFQoYGX3DBBUPdxz/+8an8sY99bKhjqGHekzaCDPVjaP4DDzwwlRk23oXEMXxxv0M+upA4hqYwtC5DYE444YShLkMkGTaeoXwMg6XNItuOoa+pR6Zqqqp65JFHpjItF12YLPvDnPRgqFI3lhhqeejQoanM0LHU56KLLhrqaInJ72WKsOwfHDt5f46HLrXY3EJmu7BYzl9Zz/C97JMMw8xQspy7qqouvvji4TrHEkM2cy5jGs3Uis/dherPLSSte7adhvvycxmKTFtLhsFWjbYj3j/XKKbq6VJZdc86Nz2SVXp0z54acE5ODdjnub7nXMd2TZsa57LOusLQzy6kec7sVA+2I+1DCS0YmYaMuqYNjGnouj6/W2mI50Y3J3V1GUbOtYX7q6znOtxZizqWqkeyjh6dzYX7tKynHmkH4N6cbd7VLVGPbi5jXe7FuH5wLsvPcv+d73Gsyzafa/v7l30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGHsuWefdJ5k+lwzvR29FelFomfmvPPOm8rnn3/+UEfvy5NPPjmV77nnnqEufZYPPvjgls/GlHW8R5e+66DpPMn0EKXvN39/1eidpG8v/XirPDPpO2abP/7441P5rrvuGurSM75OarG5+Gn+mc4LRs9jeojS/1g1elfp385xdtJJJ23576rGccazKVKfPDOhqk832Xn256ZHssoTnvpwDkiffHrxqsbzLugJZ9tlW/FsikwBmt9ZNY6BVb9jKZ7kzgeeXskupSVTV9ETmx7MLpUsPeKp4yqP3yZp0NHNbTkn8fdm300PeNXh81XqRR94jheeW5H35D6E6+JS9Eg6X3yXHpapEHmWT6ZGZPqwXLO6fZF69HXZHvSI55kiVb0eXWrlvF6lx9FO9mX2a6YG7fTIPQTXnU6Pub1jHDTZHmwr6pH74e69gfNefm93nkLVwc1X/mVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkY++7Z73x7nY+Oee7Tj8ec65mX99hjjx3q6C9KX37mca8aPUz05b/77rtTeVU+9Dl7YLv8lPRWZz3b6ic/+clUZk7prPvRj3401GWO8arR20oP7EsvvXTEMj9L79Mm6ZGs4xHvfOD0j7/88stTOXOBV1VddNFFw3WOF3pgH3744anM8Zk68v5L8Vyu4xFPfei1zzMmmIeXed5zbHHeyXGWGleN/jOO66V6/Lq5LfXhfJHtw/M/OLflfEa/bJ4xwu/Je6waD5s6PjrWGTvdnoHXOQ9xbcm5rsuTvMT2XsU6eavzujvDompcp7m/yz0U4VjqnvVoJ9uDZyDxjIs8E4n+5Vwj2P55TR1la6gH56Q8t4I+8GxnjrO85nwpIzm3s435fpj7tJ/97GdDXXdWSafHXPZX/mVfREREREREZGH4si8iIiIiIiKyMPY9jD/pUohUjWETTF2U/5bh3l0aBN4jQ4wZbpF1XVq6VeHWm0oX+sqw8Uypx7DYDAHLVIdVh4c5ZegSw2Ey5IlpMfKa7b+U1DBdP2NoXWrA0OMMKaZWDOtPXRmOn+kX+e/ye9n+1GcpYZn5O9h3c77g2MnP0h7zxBNPDNepM8PMMjUiQ2RzLl3V/kvRI+nCv7uQbvZ56pNpLfk9aaXg+OzSMy2x/UnX5zhf5HrO0GP285yHuC/Ium6NPhrGwzqwrXJNoB5dOjeOj9zTMWycKXrlX2B/zLZiu3EdSg04PnKuox2DYeRHO92ckBqwX3d7U1qFc89AO0bqwTF3tM9XJNdz6sH3w9SH1rscS0wvnnrw3WQu+Jd9ERERERERkYXhy76IiIiIiIjIwvBlX0RERERERGRhHKhnn3Qeftal34i+MfowuntsN8VM55deqkem+130gqX/jr6xvKYPpvPMdGcGdOktlnJmwiq6lFTpZaXvOD1F9MdSn6znPdIrxnscjXok7Lvp2e7OMMhzKaoOT2mYqaz4PTkndulA1WOcP1iX6wm99vTs33XXXVveM+fETvOlrh/rkO3BNaGb53j+Raaf5DrEsZXkPXmP7nmWSrYB2zF/P+vYzx988MGpfMwxxwx1qV23h2N7H4165G/s9rddysKqqkcffXQqM+1Y7guYWoz6yL/Q6cGzDtiOeYYV0/J1qd66c8mOxvHRke1BPTh/5blHHB85J/JMi/ye/dCj63Nb4QgWERERERERWRi+7IuIiIiIiIgsjFmF8ZMuHU9e7ySkYd1/e9ChMHMIzdmu5YEh9l0qq1UhL1vRpavaj7aZgx5Jl1qsS2nJNCEMa8qwwC68tQsbJ3vRVnPTo7MksW1yvLzyyitDHUOPP/zhD0/lLmVcl95vP1KLzU0P0oXRZx3D9hnWf+KJJ05lzm2pOXVM7Y7G+Yp0a33WMT3VvffeO1y/8cYbU5lzW9qOmHaM6au2uv9uMXc9km6+YkrkDNuvGtOHcb5KDTjvdemrdqut1rF7zokufRvn/ccee2y4zhRu/J5MFcp5j3PbXpB6dPbOudE9G+eVp556ari+5557pjL7XKY25vqR4+q9vP90bKoe66SoPnTo0HCd6wktMWm54LyXcyKtGruV+rvTYzv4l30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGH80T9t05y0V76QOTFnH92cn22voPdlv335HXy2OXuYyDpjOT/b/eb98IF3LEWPVSmO8rNd2qvuTAuyF1ptsh757KvSuqYGnR5d6r11tNopS9WDvspMrcS69IxTj/Rcduef7BarUtbNmW4OYoqwTFHFtFfpn00PMq+7NLu7xVL1YIqwj3zkI1OZeuQ5CTxfIbWiHnvRVhy7vOec6fTIMxOqqo477rip3OnBMxPyetV5SbuxnixVjzwDqWo8g4e/OfXgWQypB88F6M6g2SnsK+wDR8K/7IuIiIiIiIgsDF/2RURERERERBbGtsP4u5DqpTDnUPlVoddzelaSz77Oc8459HTOfWUVO9Vjp6GO+9FWS9WjG/e7FVqnHiM573TpJglDmBmavBWmQuzp9Oj2JQxhzvDKddpjt9adTU31RnaqB0OYM/S106NLpfVe2NTUYmSnejCEebvp9bq0su+Fo12PtFhU9Xp044N67HS/txQ9ut+/W3ps9R1VfSrybr/X1VGP7YxB/7IvIiIiIiIisjB82RcRERERERFZGL7si4iIiIiIiCyMbXv2RURERERERGQz8C/7IiIiIiIiIgvDl30RERERERGRheHLvoiIiIiIiMjC8GVfREREREREZGH4si8iIiIiIiKyMHzZFxEREREREVkYvuyLiIiIiIiILAxf9kVEREREREQWhi/7IiIiIiIiIgvj/we+J/b9lKOjegAAAABJRU5ErkJggg==",
+      "text/plain": [
+       "<Figure size 1000x500 with 50 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "# show interpolations\n",
+    "fig, axes = plt.subplots(nrows=5, ncols=10, figsize=(10, 5))\n",
+    "\n",
+    "for i in range(5):\n",
+    "    for j in range(10):\n",
+    "        axes[i][j].imshow(interpolations[i, j].cpu().squeeze(0), cmap='gray')\n",
+    "        axes[i][j].axis('off')\n",
+    "plt.tight_layout(pad=0.)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "interpreter": {
+   "hash": "95022f601a219c6b6d093149c9a9b9a061a4446d3680d89cef8a1f82970031f2"
+  },
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.14"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}
diff --git a/src/pythae/models/__init__.py b/src/pythae/models/__init__.py
index ba45191e..40284ac6 100755
--- a/src/pythae/models/__init__.py
+++ b/src/pythae/models/__init__.py
@@ -24,6 +24,7 @@
 from .disentangled_beta_vae import DisentangledBetaVAE, DisentangledBetaVAEConfig
 from .factor_vae import FactorVAE, FactorVAEConfig
 from .hvae import HVAE, HVAEConfig
+from .hrq_vae import HRQVAE, HRQVAEConfig
 from .info_vae import INFOVAE_MMD, INFOVAE_MMD_Config
 from .iwae import IWAE, IWAEConfig
 from .miwae import MIWAE, MIWAEConfig
@@ -96,4 +97,6 @@
     "MIWAEConfig",
     "PIWAE",
     "PIWAEConfig",
+    "HRQVAE",
+    "HRQVAEConfig",
 ]
diff --git a/src/pythae/models/auto_model/auto_config.py b/src/pythae/models/auto_model/auto_config.py
index 3a2d55a2..fae89d18 100644
--- a/src/pythae/models/auto_model/auto_config.py
+++ b/src/pythae/models/auto_model/auto_config.py
@@ -60,6 +60,11 @@ def from_json_file(cls, json_path):
 
             model_config = HVAEConfig.from_json_file(json_path)
 
+        elif config_name == "HRQVAEConfig":
+            from ..hrq_vae import HRQVAEConfig
+
+            model_config = HRQVAEConfig.from_json_file(json_path)
+
         elif config_name == "INFOVAE_MMD_Config":
             from ..info_vae import INFOVAE_MMD_Config
 
diff --git a/src/pythae/models/auto_model/auto_model.py b/src/pythae/models/auto_model/auto_model.py
index 7d875c0d..d103ac29 100644
--- a/src/pythae/models/auto_model/auto_model.py
+++ b/src/pythae/models/auto_model/auto_model.py
@@ -74,6 +74,11 @@ def load_from_folder(cls, dir_path: str):
 
             model = HVAE.load_from_folder(dir_path=dir_path)
 
+        elif model_name == "HRQVAEConfig":
+            from ..hrq_vae import HRQVAE
+
+            model = HRQVAE.load_from_folder(dir_path=dir_path)
+
         elif model_name == "INFOVAE_MMD_Config":
             from ..info_vae import INFOVAE_MMD
 
diff --git a/src/pythae/models/hrq_vae/__init__.py b/src/pythae/models/hrq_vae/__init__.py
new file mode 100755
index 00000000..770b4a93
--- /dev/null
+++ b/src/pythae/models/hrq_vae/__init__.py
@@ -0,0 +1,19 @@
+"""This module is the implementation of the Vector Quantized VAE proposed in 
+(https://arxiv.org/abs/1711.00937).
+
+Available samplers
+-------------------
+
+Normalizing flows sampler to come.
+
+.. autosummary::
+    ~pythae.samplers.GaussianMixtureSampler
+    ~pythae.samplers.MAFSampler
+    ~pythae.samplers.IAFSampler
+    :nosignatures:
+"""
+
+from .hrq_vae_config import HRQVAEConfig
+from .hrq_vae_model import HRQVAE
+
+__all__ = ["HRQVAE", "HRQVAEConfig"]
diff --git a/src/pythae/models/hrq_vae/hrq_vae_config.py b/src/pythae/models/hrq_vae/hrq_vae_config.py
new file mode 100755
index 00000000..51aa0131
--- /dev/null
+++ b/src/pythae/models/hrq_vae/hrq_vae_config.py
@@ -0,0 +1,37 @@
+from pydantic.dataclasses import dataclass
+from typing import Optional
+
+from ..ae import AEConfig
+
+
+@dataclass
+class HRQVAEConfig(AEConfig):
+    r"""
+    Hierarchical Residual Quantization VAE model config config class
+
+    Parameters:
+        input_dim (tuple): The input_data dimension.
+        latent_dim (int): The latent space dimension. Default: 10.
+        num_embedding (int): The number of embedding points. Default: 64
+        num_levels (int): Depth of hierarchy. Default: 4
+        kl_weight (float): Weighting of the KL term. Default: 0.1
+        init_scale (float): Magnitude of the embedding initialisation, should roughly match the encoder. Default: 1.0
+        init_decay_weight (float): Factor by which the magnitude of each successive levels is multiplied. Default: 1.5
+        norm_loss_weight (float): Weighting of the norm loss term. Default: 0.5
+        norm_loss_scale (float): Scale for the norm loss. Default: 1.5
+        temp_schedule_gamma (float): Decay constant for the Gumbel temperature - will be (epoch/gamma). Default: 33.333
+        depth_drop_rate (float): Probability of dropping each level during training. Default: 0.1
+    """
+
+    num_embeddings: int = 64
+    num_levels: int = 4
+    kl_weight: float = 0.1
+    init_scale: float = 1.0
+    init_decay_weight: float = 0.5
+    norm_loss_weight: Optional[float] = 0.5
+    norm_loss_scale: float = 1.5
+    temp_schedule_gamma: float = 33.333
+    depth_drop_rate: float = 0.1
+
+    def __post_init__(self):
+        super().__post_init__()
diff --git a/src/pythae/models/hrq_vae/hrq_vae_model.py b/src/pythae/models/hrq_vae/hrq_vae_model.py
new file mode 100755
index 00000000..59459335
--- /dev/null
+++ b/src/pythae/models/hrq_vae/hrq_vae_model.py
@@ -0,0 +1,128 @@
+from typing import Optional, Any, Dict
+
+import torch
+import torch.nn.functional as F
+
+from ...data.datasets import BaseDataset
+from ..ae import AE
+from ..base.base_utils import ModelOutput
+from ..nn import BaseDecoder, BaseEncoder
+from .hrq_vae_config import HRQVAEConfig
+from .hrq_vae_utils import HierarchicalResidualQuantizer
+
+
+class HRQVAE(AE):
+    r"""
+    Hierarchical Residual Quantization-VAE model. Introduced in https://aclanthology.org/2022.acl-long.178/ (Hosking et al., ACL 2022)
+
+    Args:
+        model_config (HRQVAEConfig): The Variational Autoencoder configuration setting the main
+            parameters of the model.
+
+        encoder (BaseEncoder): An instance of BaseEncoder (inheriting from `torch.nn.Module` which
+            plays the role of encoder. This argument allows you to use your own neural networks
+            architectures if desired. If None is provided, a simple Multi Layer Preception
+            (https://en.wikipedia.org/wiki/Multilayer_perceptron) is used. Default: None.
+
+        decoder (BaseDecoder): An instance of BaseDecoder (inheriting from `torch.nn.Module` which
+            plays the role of encoder. This argument allows you to use your own neural networks
+            architectures if desired. If None is provided, a simple Multi Layer Preception
+            (https://en.wikipedia.org/wiki/Multilayer_perceptron) is used. Default: None.
+
+    """
+
+    def __init__(
+        self,
+        model_config: HRQVAEConfig,
+        encoder: Optional[BaseEncoder] = None,
+        decoder: Optional[BaseDecoder] = None,
+    ):
+        AE.__init__(self, model_config=model_config, encoder=encoder, decoder=decoder)
+
+        self._set_quantizer(model_config)
+
+        self.model_name = "HRQVAE"
+
+    def _set_quantizer(self, model_config):
+        if model_config.input_dim is None:
+            raise AttributeError(
+                "No input dimension provided !"
+                "'input_dim' parameter of HRQVAEConfig instance must be set to 'data_shape' where "
+                "the shape of the data is (C, H, W ..). Unable to set quantizer."
+            )
+
+        x = torch.randn((2,) + self.model_config.input_dim)
+        z = self.encoder(x).embedding
+        if len(z.shape) == 2:
+            z = z.reshape(z.shape[0], 1, 1, -1)
+
+        z = z.permute(0, 2, 3, 1)
+
+        self.model_config.embedding_dim = z.shape[-1]
+
+        self.quantizer = HierarchicalResidualQuantizer(model_config=model_config)
+
+    def forward(self, inputs: Dict[str, Any], **kwargs) -> ModelOutput:
+        """
+        The VAE model
+
+        Args:
+            inputs (dict): A dict of samples
+
+        Returns:
+            ModelOutput: An instance of ModelOutput containing all the relevant parameters
+
+        """
+
+        x = inputs["data"]
+        uses_ddp = kwargs.pop("uses_ddp", False)
+        epoch = kwargs.pop("epoch", 0)
+
+        encoder_output = self.encoder(x)
+
+        embeddings = encoder_output.embedding
+
+        reshape_for_decoding = False
+
+        if len(embeddings.shape) == 2:
+            embeddings = embeddings.reshape(embeddings.shape[0], 1, 1, -1)
+            reshape_for_decoding = True
+
+        embeddings = embeddings.permute(0, 2, 3, 1)
+
+        quantizer_output = self.quantizer(embeddings, epoch=epoch, uses_ddp=uses_ddp)
+
+        quantized_embed = quantizer_output.quantized_vector
+
+        if reshape_for_decoding:
+            quantized_embed = quantized_embed.reshape(embeddings.shape[0], -1)
+
+        recon_x = self.decoder(quantized_embed).reconstruction
+
+        loss, recon_loss, hrq_loss = self.loss_function(recon_x, x, quantizer_output)
+
+        output = ModelOutput(
+            loss=loss,
+            recon_loss=recon_loss,
+            hrq_loss=hrq_loss,
+            recon_x=recon_x,
+            z=quantized_embed,
+            z_orig=quantizer_output.z_orig,
+            quantized_indices=quantizer_output.quantized_indices,
+            probs=quantizer_output.probs,
+        )
+
+        return output
+
+    def loss_function(self, recon_x, x, quantizer_output):
+        recon_loss = F.mse_loss(
+            recon_x.reshape(x.shape[0], -1), x.reshape(x.shape[0], -1), reduction="none"
+        ).sum(dim=-1)
+
+        hrq_loss = quantizer_output.loss
+
+        return (
+            (recon_loss + hrq_loss).mean(dim=0),
+            recon_loss.mean(dim=0),
+            hrq_loss.mean(dim=0),
+        )
diff --git a/src/pythae/models/hrq_vae/hrq_vae_utils.py b/src/pythae/models/hrq_vae/hrq_vae_utils.py
new file mode 100755
index 00000000..346c68ea
--- /dev/null
+++ b/src/pythae/models/hrq_vae/hrq_vae_utils.py
@@ -0,0 +1,153 @@
+import torch
+import torch.distributed as dist
+import torch.nn as nn
+import torch.nn.functional as F
+from math import sqrt
+
+from ..base.base_utils import ModelOutput
+from .hrq_vae_config import HRQVAEConfig
+
+
+class HierarchicalResidualQuantizer(nn.Module):
+    def __init__(self, model_config: HRQVAEConfig):
+        nn.Module.__init__(self)
+
+        self.model_config = model_config
+
+        self.embedding_dim = model_config.embedding_dim
+        self.num_embeddings = model_config.num_embeddings
+        self.num_levels = model_config.num_levels
+
+        self.embeddings = nn.ModuleList(
+            [
+                nn.Embedding(self.num_embeddings, self.embedding_dim)
+                for hix in range(self.num_levels)
+            ]
+        )
+
+        init_scale = model_config.init_scale
+        init_decay_weight = model_config.init_decay_weight
+        for hix, embedding in enumerate(self.embeddings):
+            scale = init_scale * init_decay_weight**hix / sqrt(self.embedding_dim)
+            embedding.weight.data.uniform_(-1.0 * scale, scale)
+
+            # Normalise onto sphere
+            embedding.weight.data = (
+                embedding.weight.data
+                / torch.linalg.vector_norm(embedding.weight, dim=1, keepdim=True)
+                * init_scale
+                * init_decay_weight**hix
+            )
+
+    def forward(self, z: torch.Tensor, epoch: int, uses_ddp: bool = False):
+        if uses_ddp:
+            raise Exception("HRQVAE doesn't currently support DDP :(")
+
+        input_shape = z.shape
+
+        z = z.reshape(-1, self.embedding_dim)
+
+        loss = torch.zeros(z.shape[0]).to(z.device)
+
+        resid_error = z
+
+        quantized = []
+        codes = []
+        all_probs = []
+
+        for head_ix, embedding in enumerate(self.embeddings):
+            if head_ix > 0:
+                resid_error = z - torch.sum(torch.cat(quantized, dim=1), dim=1)
+
+            distances = -1.0 * (
+                torch.sum(resid_error**2, dim=-1, keepdim=True)
+                + torch.sum(embedding.weight**2, dim=-1)
+                - 2 * torch.matmul(resid_error, embedding.weight.T)
+            )
+
+            gumbel_sched_weight = torch.exp(
+                -torch.tensor(float(epoch))
+                / float(self.model_config.temp_schedule_gamma * 1.5**head_ix)
+            )
+            gumbel_temp = max(gumbel_sched_weight, 0.5)
+
+            if self.training:
+
+                sample_onehot = F.gumbel_softmax(
+                    distances, tau=gumbel_temp, hard=True, dim=-1
+                )
+            else:
+                indices = torch.argmax(distances, dim=-1)
+                sample_onehot = F.one_hot(
+                    indices, num_classes=self.num_embeddings
+                ).float()
+
+            probs = F.softmax(distances / gumbel_temp, dim=-1)
+
+            # KL loss
+            prior = (
+                torch.ones_like(distances).detach()
+                / torch.ones_like(distances).sum(-1, keepdim=True).detach()
+            )
+            kl_loss = torch.nn.KLDivLoss(reduction="none")
+            kl = kl_loss(nn.functional.log_softmax(distances, dim=-1), prior).sum(
+                dim=-1
+            )
+            loss += kl * self.model_config.kl_weight
+
+            # quantization
+            this_quantized = sample_onehot @ embedding.weight
+            this_quantized = this_quantized.reshape_as(z)
+
+            quantized.append(this_quantized.unsqueeze(-2))
+            codes.append(torch.argmax(sample_onehot, dim=-1).unsqueeze(-1))
+            all_probs.append(probs.unsqueeze(-2))
+
+        quantized = torch.cat(quantized, dim=-2)
+        quantized_indices = torch.cat(codes, dim=-1)
+        all_probs = torch.cat(all_probs, dim=-2)
+
+        # Calculate the norm loss
+        if self.model_config.norm_loss_weight is not None:
+            upper_norms = torch.linalg.vector_norm(quantized[:, :-1, :], dim=-1)
+            lower_norms = torch.linalg.vector_norm(quantized[:, 1:, :], dim=-1)
+            norm_loss = (
+                torch.max(
+                    lower_norms / upper_norms * self.model_config.norm_loss_scale,
+                    torch.ones_like(lower_norms),
+                )
+                - 1.0
+            ) ** 2
+
+            loss += norm_loss.mean(dim=1) * self.model_config.norm_loss_weight
+
+        # Depth drop out
+        if self.training:
+
+            drop_dist = torch.distributions.Bernoulli(
+                1 - self.model_config.depth_drop_rate
+            )
+
+            mask = drop_dist.sample(sample_shape=(*quantized.shape[:-1], 1))
+
+            mask = torch.cumprod(mask, dim=1).to(quantized.device)
+            quantized = quantized * mask
+
+        quantized = quantized.sum(dim=-2).reshape(*input_shape)
+        quantized = quantized.permute(0, 3, 1, 2)
+
+        loss = loss.reshape(input_shape[0], -1).mean(dim=1)
+
+        quantized_indices = quantized_indices.reshape(
+            *input_shape[:-1], self.num_levels
+        )
+
+        output = ModelOutput(
+            z_orig=z,
+            quantized_vector=quantized,
+            quantized_indices=quantized_indices,
+            loss=loss,
+            probs=all_probs,
+        )
+
+        return output
diff --git a/src/pythae/models/nn/benchmarks/mnist/resnets.py b/src/pythae/models/nn/benchmarks/mnist/resnets.py
index 4176f5a9..d7afc0ea 100644
--- a/src/pythae/models/nn/benchmarks/mnist/resnets.py
+++ b/src/pythae/models/nn/benchmarks/mnist/resnets.py
@@ -578,6 +578,117 @@ def forward(self, x: torch.Tensor, output_layer_levels: List[int] = None):
         return output
 
 
+class Encoder_ResNet_HRQVAE_MNIST(BaseEncoder):
+    """
+    A ResNet encoder suited for MNIST and VQ- or HRQ- VAE models. It differs from the VQVAE ResNet in that it outputs only a single embedding vector.
+
+    It can be built as follows:
+
+    .. code-block::
+
+        >>> from pythae.models.nn.benchmarks.mnist import Encoder_ResNet_HRQVAE_MNIST
+        >>> from pythae.models import HRQVAEConfig
+        >>> model_config = HRQVAEConfig(input_dim=(1, 28, 28), latent_dim=16)
+        >>> encoder = Encoder_ResNet_HRQVAE_MNIST(model_config)
+
+
+    and then passed to a :class:`pythae.models` instance
+
+        >>> from pythae.models import HRQVAE
+        >>> model = HRQVAE(model_config=model_config, encoder=encoder)
+        >>> model.encoder == encoder
+        ... True
+
+    .. note::
+
+        Please note that this encoder is only suitable for Autoencoder based models since it only
+        outputs the embeddings of the input data under the key `embedding`.
+
+        .. code-block::
+
+            >>> import torch
+            >>> input = torch.rand(2, 1, 28, 28)
+            >>> out = encoder(input)
+            >>> out.embedding.shape
+            ... torch.Size([2, 16, 1,  1])
+
+    """
+
+    def __init__(self, args: BaseAEConfig):
+        BaseEncoder.__init__(self)
+
+        self.input_dim = (1, 28, 28)
+        self.latent_dim = args.latent_dim
+        self.n_channels = 1
+
+        layers = nn.ModuleList()
+
+        layers.append(nn.Sequential(nn.Conv2d(self.n_channels, 64, 4, 2, padding=1)))
+
+        layers.append(nn.Sequential(nn.Conv2d(64, 128, 4, 2, padding=1)))
+
+        layers.append(nn.Sequential(nn.Conv2d(128, 128, 3, 2, padding=1)))
+
+        layers.append(
+            nn.Sequential(
+                ResBlock(in_channels=128, out_channels=32),
+                ResBlock(in_channels=128, out_channels=32),
+            )
+        )
+
+        # Additional Conv layer to squeeze down to a single embedding
+        # layers.append(nn.Sequential(nn.Conv2d(128, 128, 4, 1, padding=0)))
+
+        self.layers = layers
+        self.depth = len(layers)
+
+        self.pre_qantized = nn.Conv2d(128, self.latent_dim, 4, 1)
+
+    def forward(self, x: torch.Tensor, output_layer_levels: List[int] = None):
+        """Forward method
+
+        Args:
+            x (torch.Tensor): A batch of inputs.  
+            output_layer_levels (List[int]): The levels of the layers where the outputs are
+                extracted. If None, the last layer's output is returned. Default: None.
+
+        Returns:
+            ModelOutput: An instance of ModelOutput containing the embeddings of the input data
+            under the key `embedding`. Optional: The outputs of the layers specified in
+            `output_layer_levels` arguments are available under the keys `embedding_layer_i` where
+            i is the layer's level."""
+        output = ModelOutput()
+
+        max_depth = self.depth
+
+        if output_layer_levels is not None:
+            assert all(
+                self.depth >= levels > 0 or levels == -1
+                for levels in output_layer_levels
+            ), (
+                f"Cannot output layer deeper than depth ({self.depth})."
+                f"Got ({output_layer_levels})."
+            )
+
+            if -1 in output_layer_levels:
+                max_depth = self.depth
+            else:
+                max_depth = max(output_layer_levels)
+
+        out = x
+
+        for i in range(max_depth):
+            out = self.layers[i](out)
+
+            if output_layer_levels is not None:
+                if i + 1 in output_layer_levels:
+                    output[f"embedding_layer_{i+1}"] = out
+            if i + 1 == self.depth:
+                output["embedding"] = self.pre_qantized(out)
+
+        return output
+
+
 class Decoder_ResNet_AE_MNIST(BaseDecoder):
     """
     A ResNet decoder suited for MNIST and Autoencoder-based
@@ -884,3 +995,122 @@ def forward(self, z: torch.Tensor, output_layer_levels: List[int] = None):
                 output["reconstruction"] = out
 
         return output
+
+
+class Decoder_ResNet_HRQVAE_MNIST(BaseDecoder):
+    """
+    A ResNet decoder suited for MNIST and VQ- or HRQ- VAE models. It differs from the VQVAE ResNet in that it expects only a single embedding vector as input.
+
+    .. code-block::
+
+        >>> from pythae.models.nn.benchmarks.mnist import Decoder_ResNet_HRQVAE_MNIST
+        >>> from pythae.models import HRQVAEConfig
+        >>> model_config = HRQVAEConfig(input_dim=(1, 28, 28), latent_dim=16)
+        >>> decoder = Decoder_ResNet_HRQVAE_MNIST(model_config)
+
+
+    and then passed to a :class:`pythae.models` instance
+
+        >>> from pythae.models import HRQVAE
+        >>> model = HRQVAE(model_config=model_config, decoder=decoder)
+        >>> model.decoder == decoder
+        ... True
+
+    .. note::
+
+        Please note that this decoder is suitable for **all** models.
+
+        .. code-block::
+
+            >>> import torch
+            >>> input = torch.randn(2, 16, 1, 1)
+            >>> out = decoder(input)
+            >>> out.reconstruction.shape
+            ... torch.Size([2, 1, 28, 28])
+    """
+
+    def __init__(self, args: BaseAEConfig):
+        BaseDecoder.__init__(self)
+
+        self.input_dim = (1, 28, 28)
+        self.latent_dim = args.latent_dim
+        self.n_channels = 1
+
+        layers = nn.ModuleList()
+
+        layers.append(nn.ConvTranspose2d(self.latent_dim, 128, 4, 1))
+
+        layers.append(nn.ConvTranspose2d(128, 128, 3, 2, padding=1))
+
+        layers.append(
+            nn.Sequential(
+                ResBlock(in_channels=128, out_channels=32),
+                ResBlock(in_channels=128, out_channels=32),
+                nn.ReLU(),
+            )
+        )
+
+        layers.append(
+            nn.Sequential(
+                nn.ConvTranspose2d(128, 64, 3, 2, padding=1, output_padding=1),
+                nn.ReLU(),
+            )
+        )
+
+        layers.append(
+            nn.Sequential(
+                nn.ConvTranspose2d(
+                    64, self.n_channels, 3, 2, padding=1, output_padding=1
+                ),
+                nn.Sigmoid(),
+            )
+        )
+
+        self.layers = layers
+        self.depth = len(layers)
+
+    def forward(self, z: torch.Tensor, output_layer_levels: List[int] = None):
+        """Forward method
+
+        Args:
+            z (torch.Tensor): A batch of embeddings.  
+            output_layer_levels (List[int]): The levels of the layers where the outputs are
+                extracted. If None, the last layer's output is returned. Default: None.
+
+        Returns:
+            ModelOutput: An instance of ModelOutput containing the reconstruction of the latent code
+            under the key `reconstruction`. Optional: The outputs of the layers specified in
+            `output_layer_levels` arguments are available under the keys `reconstruction_layer_i`
+            where i is the layer's level.
+        """
+        output = ModelOutput()
+
+        max_depth = self.depth
+
+        if output_layer_levels is not None:
+            assert all(
+                self.depth >= levels > 0 or levels == -1
+                for levels in output_layer_levels
+            ), (
+                f"Cannot output layer deeper than depth ({self.depth})."
+                f"Got ({output_layer_levels})"
+            )
+
+            if -1 in output_layer_levels:
+                max_depth = self.depth
+            else:
+                max_depth = max(output_layer_levels)
+
+        out = z
+
+        for i in range(max_depth):
+            out = self.layers[i](out)
+
+            if output_layer_levels is not None:
+                if i + 1 in output_layer_levels:
+                    output[f"reconstruction_layer_{i+1}"] = out
+
+            if i + 1 == self.depth:
+                output["reconstruction"] = out
+
+        return output
diff --git a/tests/test_HRQVAE.py b/tests/test_HRQVAE.py
new file mode 100755
index 00000000..262dc715
--- /dev/null
+++ b/tests/test_HRQVAE.py
@@ -0,0 +1,760 @@
+import os
+from copy import deepcopy
+
+import pytest
+import torch
+from pydantic import ValidationError
+
+from pythae.customexception import BadInheritanceError
+from pythae.models import HRQVAE, AutoModel, HRQVAEConfig
+from pythae.models.base.base_utils import ModelOutput
+from pythae.models.vq_vae.vq_vae_utils import Quantizer, QuantizerEMA
+from pythae.pipelines import GenerationPipeline, TrainingPipeline
+from pythae.samplers import PixelCNNSamplerConfig
+from pythae.trainers import BaseTrainer, BaseTrainerConfig
+from tests.data.custom_architectures import (
+    Decoder_AE_Conv,
+    Encoder_AE_Conv,
+    NetBadInheritance,
+)
+
+PATH = os.path.dirname(os.path.abspath(__file__))
+
+
+@pytest.fixture(params=[HRQVAEConfig(), HRQVAEConfig(latent_dim=4)])
+def model_configs_no_input_dim(request):
+    return request.param
+
+
+@pytest.fixture(
+    params=[
+        HRQVAEConfig(
+            input_dim=(1, 28, 28), latent_dim=16, num_embeddings=10
+        ),  #  ! Needs squared latent_dim !
+        HRQVAEConfig(
+            input_dim=(1, 28, 28),
+            latent_dim=16,
+            num_embeddings=10,
+            num_levels = 8,
+            kl_weight = 0.01,
+            init_scale = 1.0,
+            init_decay_weight = 0.5,
+            norm_loss_weight = 0.5,
+            norm_loss_scale = 1.5,
+            temp_schedule_gamma=10,
+        ),
+    ]
+)
+def model_configs(request):
+    return request.param
+
+
+@pytest.fixture
+def custom_encoder(model_configs):
+    return Encoder_AE_Conv(model_configs)
+
+
+@pytest.fixture
+def custom_decoder(model_configs):
+    return Decoder_AE_Conv(model_configs)
+
+
+class Test_Model_Building:
+    @pytest.fixture()
+    def bad_net(self):
+        return NetBadInheritance()
+
+    def test_build_model(self, model_configs):
+        model = HRQVAE(model_configs)
+        assert all(
+            [
+                model.input_dim == model_configs.input_dim,
+                model.latent_dim == model_configs.latent_dim,
+            ]
+        )
+
+
+    def build_quantizer(self, model_configs):
+        model = HRQVAE(model_configs)
+
+        if model.use_ema:
+            assert isinstance(model.quantizer, QuantizerEMA)
+
+        else:
+            assert isinstance(model.quantizer, Quantizer)
+
+    def test_raises_bad_inheritance(self, model_configs, bad_net):
+        with pytest.raises(BadInheritanceError):
+            model = HRQVAE(model_configs, encoder=bad_net)
+
+        with pytest.raises(BadInheritanceError):
+            model = HRQVAE(model_configs, decoder=bad_net)
+
+    def test_raises_no_input_dim(
+        self, model_configs_no_input_dim, custom_encoder, custom_decoder
+    ):
+        with pytest.raises(AttributeError):
+            model = HRQVAE(model_configs_no_input_dim)
+
+        with pytest.raises(AttributeError):
+            model = HRQVAE(model_configs_no_input_dim, encoder=custom_encoder)
+
+        with pytest.raises(AttributeError):
+            model = HRQVAE(model_configs_no_input_dim, decoder=custom_decoder)
+
+    def test_build_custom_arch(self, model_configs, custom_encoder, custom_decoder):
+        model = HRQVAE(model_configs, encoder=custom_encoder, decoder=custom_decoder)
+
+        assert model.encoder == custom_encoder
+        assert not model.model_config.uses_default_encoder
+        assert model.decoder == custom_decoder
+        assert not model.model_config.uses_default_decoder
+
+        model = HRQVAE(model_configs, encoder=custom_encoder)
+
+        assert model.encoder == custom_encoder
+        assert not model.model_config.uses_default_encoder
+        assert model.model_config.uses_default_decoder
+
+        model = HRQVAE(model_configs, decoder=custom_decoder)
+
+        assert model.model_config.uses_default_encoder
+        assert model.decoder == custom_decoder
+        assert not model.model_config.uses_default_decoder
+
+
+class Test_Model_Saving:
+    def test_default_model_saving(self, tmpdir, model_configs):
+        tmpdir.mkdir("dummy_folder")
+        dir_path = dir_path = os.path.join(tmpdir, "dummy_folder")
+
+        model = HRQVAE(model_configs)
+
+        model.state_dict()["encoder.layers.0.0.weight"][0] = 0
+
+        model.save(dir_path=dir_path)
+
+        assert set(os.listdir(dir_path)) == set(
+            ["model_config.json", "model.pt", "environment.json"]
+        )
+
+        # reload model
+        model_rec = AutoModel.load_from_folder(dir_path)
+
+        # check configs are the same
+        assert model_rec.model_config.__dict__ == model.model_config.__dict__
+
+        assert all(
+            [
+                torch.equal(model_rec.state_dict()[key], model.state_dict()[key])
+                for key in model.state_dict().keys()
+            ]
+        )
+
+    def test_custom_encoder_model_saving(self, tmpdir, model_configs, custom_encoder):
+        tmpdir.mkdir("dummy_folder")
+        dir_path = dir_path = os.path.join(tmpdir, "dummy_folder")
+
+        model = HRQVAE(model_configs, encoder=custom_encoder)
+
+        model.state_dict()["encoder.layers.0.0.weight"][0] = 0
+
+        model.save(dir_path=dir_path)
+
+        assert set(os.listdir(dir_path)) == set(
+            ["model_config.json", "model.pt", "encoder.pkl", "environment.json"]
+        )
+
+        # reload model
+        model_rec = AutoModel.load_from_folder(dir_path)
+
+        # check configs are the same
+        assert model_rec.model_config.__dict__ == model.model_config.__dict__
+
+        assert all(
+            [
+                torch.equal(model_rec.state_dict()[key], model.state_dict()[key])
+                for key in model.state_dict().keys()
+            ]
+        )
+
+    def test_custom_decoder_model_saving(self, tmpdir, model_configs, custom_decoder):
+        tmpdir.mkdir("dummy_folder")
+        dir_path = dir_path = os.path.join(tmpdir, "dummy_folder")
+
+        model = HRQVAE(model_configs, decoder=custom_decoder)
+
+        model.state_dict()["encoder.layers.0.0.weight"][0] = 0
+
+        model.save(dir_path=dir_path)
+
+        assert set(os.listdir(dir_path)) == set(
+            ["model_config.json", "model.pt", "decoder.pkl", "environment.json"]
+        )
+
+        # reload model
+        model_rec = AutoModel.load_from_folder(dir_path)
+
+        # check configs are the same
+        assert model_rec.model_config.__dict__ == model.model_config.__dict__
+
+        assert all(
+            [
+                torch.equal(model_rec.state_dict()[key], model.state_dict()[key])
+                for key in model.state_dict().keys()
+            ]
+        )
+
+    def test_full_custom_model_saving(
+        self, tmpdir, model_configs, custom_encoder, custom_decoder
+    ):
+        tmpdir.mkdir("dummy_folder")
+        dir_path = dir_path = os.path.join(tmpdir, "dummy_folder")
+
+        model = HRQVAE(model_configs, encoder=custom_encoder, decoder=custom_decoder)
+
+        model.state_dict()["encoder.layers.0.0.weight"][0] = 0
+
+        model.save(dir_path=dir_path)
+
+        assert set(os.listdir(dir_path)) == set(
+            [
+                "model_config.json",
+                "model.pt",
+                "encoder.pkl",
+                "decoder.pkl",
+                "environment.json",
+            ]
+        )
+
+        # reload model
+        model_rec = AutoModel.load_from_folder(dir_path)
+
+        # check configs are the same
+        assert model_rec.model_config.__dict__ == model.model_config.__dict__
+
+        assert all(
+            [
+                torch.equal(model_rec.state_dict()[key], model.state_dict()[key])
+                for key in model.state_dict().keys()
+            ]
+        )
+
+    def test_raises_missing_files(
+        self, tmpdir, model_configs, custom_encoder, custom_decoder
+    ):
+        tmpdir.mkdir("dummy_folder")
+        dir_path = dir_path = os.path.join(tmpdir, "dummy_folder")
+
+        model = HRQVAE(model_configs, encoder=custom_encoder, decoder=custom_decoder)
+
+        model.state_dict()["encoder.layers.0.0.weight"][0] = 0
+
+        model.save(dir_path=dir_path)
+
+        os.remove(os.path.join(dir_path, "decoder.pkl"))
+
+        # check raises decoder.pkl is missing
+        with pytest.raises(FileNotFoundError):
+            model_rec = AutoModel.load_from_folder(dir_path)
+
+        os.remove(os.path.join(dir_path, "encoder.pkl"))
+
+        # check raises encoder.pkl is missing
+        with pytest.raises(FileNotFoundError):
+            model_rec = AutoModel.load_from_folder(dir_path)
+
+        os.remove(os.path.join(dir_path, "model.pt"))
+
+        # check raises encoder.pkl is missing
+        with pytest.raises(FileNotFoundError):
+            model_rec = AutoModel.load_from_folder(dir_path)
+
+        os.remove(os.path.join(dir_path, "model_config.json"))
+
+        # check raises encoder.pkl is missing
+        with pytest.raises(FileNotFoundError):
+            model_rec = AutoModel.load_from_folder(dir_path)
+
+
+class Test_Model_forward:
+    @pytest.fixture
+    def demo_data(self):
+        data = torch.load(os.path.join(PATH, "data/mnist_clean_train_dataset_sample"))[
+            :
+        ]
+        return data  # This is an extract of 3 data from MNIST (unnormalized) used to test custom architecture
+
+    @pytest.fixture
+    def vae(self, model_configs, demo_data):
+        model_configs.input_dim = tuple(demo_data["data"][0].shape)
+        return HRQVAE(model_configs)
+
+    def test_model_train_output(self, vae, demo_data):
+        vae.train()
+
+        out = vae(demo_data)
+
+        assert isinstance(out, ModelOutput)
+
+        assert set(
+            ["loss", "recon_loss", "hrq_loss", "recon_x", "z", "z_orig", "quantized_indices", "probs"]
+        ) == set(out.keys())
+
+        print(out.z.shape, demo_data['data'].shape)
+
+        assert out.z.shape[0] == demo_data["data"].shape[0]
+        assert out.recon_x.shape == demo_data["data"].shape
+
+
+class Test_Model_interpolate:
+    @pytest.fixture(
+        params=[
+            torch.rand(3, 2, 3, 1),
+            torch.rand(3, 2, 2),
+            torch.load(os.path.join(PATH, "data/mnist_clean_train_dataset_sample"))[:][
+                "data"
+            ],
+        ]
+    )
+    def demo_data(self, request):
+        return request.param
+
+    @pytest.fixture()
+    def granularity(self):
+        return int(torch.randint(1, 10, (1,)))
+
+    @pytest.fixture
+    def ae(self, model_configs, demo_data):
+        model_configs.input_dim = tuple(demo_data[0].shape)
+        return HRQVAE(model_configs)
+
+    def test_interpolate(self, ae, demo_data, granularity):
+        with pytest.raises(AssertionError):
+            ae.interpolate(demo_data, demo_data[1:], granularity)
+
+        interp = ae.interpolate(demo_data, demo_data, granularity)
+
+        assert tuple(interp.shape) == (
+            demo_data.shape[0],
+            granularity,
+        ) + (demo_data.shape[1:])
+
+
+class Test_Model_reconstruct:
+    @pytest.fixture(
+        params=[
+            torch.rand(3, 2, 3, 1),
+            torch.rand(3, 2, 2),
+            torch.load(os.path.join(PATH, "data/mnist_clean_train_dataset_sample"))[:][
+                "data"
+            ],
+        ]
+    )
+    def demo_data(self, request):
+        return request.param
+
+    @pytest.fixture
+    def ae(self, model_configs, demo_data):
+        model_configs.input_dim = tuple(demo_data[0].shape)
+        return HRQVAE(model_configs)
+
+    def test_reconstruct(self, ae, demo_data):
+        recon = ae.reconstruct(demo_data)
+        assert tuple(recon.shape) == demo_data.shape
+
+
+@pytest.mark.slow
+class Test_HRQVAETraining:
+    @pytest.fixture
+    def train_dataset(self):
+        return torch.load(os.path.join(PATH, "data/mnist_clean_train_dataset_sample"))
+
+    @pytest.fixture(
+        params=[BaseTrainerConfig(num_epochs=3, steps_saving=2, learning_rate=1e-5)]
+    )
+    def training_configs(self, tmpdir, request):
+        tmpdir.mkdir("dummy_folder")
+        dir_path = os.path.join(tmpdir, "dummy_folder")
+        request.param.output_dir = dir_path
+        return request.param
+
+    @pytest.fixture(
+        params=[
+            torch.rand(1),
+            torch.rand(1),
+            torch.rand(1),
+            torch.rand(1),
+            torch.rand(1),
+        ]
+    )
+    def vae(self, model_configs, custom_encoder, custom_decoder, request):
+        # randomized
+
+        alpha = request.param
+
+        if alpha < 0.25:
+            model = HRQVAE(model_configs)
+
+        elif 0.25 <= alpha < 0.5:
+            model = HRQVAE(model_configs, encoder=custom_encoder)
+
+        elif 0.5 <= alpha < 0.75:
+            model = HRQVAE(model_configs, decoder=custom_decoder)
+
+        else:
+            model = HRQVAE(model_configs, encoder=custom_encoder, decoder=custom_decoder)
+
+        return model
+
+    @pytest.fixture
+    def trainer(self, vae, train_dataset, training_configs):
+        trainer = BaseTrainer(
+            model=vae,
+            train_dataset=train_dataset,
+            eval_dataset=train_dataset,
+            training_config=training_configs,
+        )
+
+        trainer.prepare_training()
+
+        return trainer
+
+    def test_vae_train_step(self, trainer):
+        start_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        step_1_loss = trainer.train_step(epoch=1)
+
+        step_1_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        # check that weights were updated
+        assert not all(
+            [
+                torch.equal(start_model_state_dict[key], step_1_model_state_dict[key])
+                for key in start_model_state_dict.keys()
+            ]
+        )
+
+    def test_vae_eval_step(self, trainer):
+        start_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        step_1_loss = trainer.eval_step(epoch=1)
+
+        step_1_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        # check that weights were not updated
+        assert all(
+            [
+                torch.equal(start_model_state_dict[key], step_1_model_state_dict[key])
+                for key in start_model_state_dict.keys()
+            ]
+        )
+
+    def test_vae_predict_step(self, trainer, train_dataset):
+        start_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        inputs, recon, generated = trainer.predict(trainer.model)
+
+        step_1_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        # check that weights were not updated
+        assert all(
+            [
+                torch.equal(start_model_state_dict[key], step_1_model_state_dict[key])
+                for key in start_model_state_dict.keys()
+            ]
+        )
+
+        assert inputs.cpu() in train_dataset.data
+        assert recon.shape == inputs.shape
+        assert generated.shape == inputs.shape
+
+    def test_vae_main_train_loop(self, trainer):
+        start_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        trainer.train()
+
+        step_1_model_state_dict = deepcopy(trainer.model.state_dict())
+
+        # check that weights were updated
+        assert not all(
+            [
+                torch.equal(start_model_state_dict[key], step_1_model_state_dict[key])
+                for key in start_model_state_dict.keys()
+            ]
+        )
+
+    def test_checkpoint_saving(self, vae, trainer, training_configs):
+        dir_path = training_configs.output_dir
+
+        # Make a training step
+        step_1_loss = trainer.train_step(epoch=1)
+
+        model = deepcopy(trainer.model)
+        optimizer = deepcopy(trainer.optimizer)
+
+        trainer.save_checkpoint(dir_path=dir_path, epoch=0, model=model)
+
+        checkpoint_dir = os.path.join(dir_path, "checkpoint_epoch_0")
+
+        assert os.path.isdir(checkpoint_dir)
+
+        files_list = os.listdir(checkpoint_dir)
+
+        assert set(["model.pt", "optimizer.pt", "training_config.json"]).issubset(
+            set(files_list)
+        )
+
+        # check pickled custom decoder
+        if not vae.model_config.uses_default_decoder:
+            assert "decoder.pkl" in files_list
+
+        else:
+            assert not "decoder.pkl" in files_list
+
+        # check pickled custom encoder
+        if not vae.model_config.uses_default_encoder:
+            assert "encoder.pkl" in files_list
+
+        else:
+            assert not "encoder.pkl" in files_list
+
+        model_rec_state_dict = torch.load(os.path.join(checkpoint_dir, "model.pt"))[
+            "model_state_dict"
+        ]
+
+        assert all(
+            [
+                torch.equal(
+                    model_rec_state_dict[key].cpu(), model.state_dict()[key].cpu()
+                )
+                for key in model.state_dict().keys()
+            ]
+        )
+
+        # check reload full model
+        model_rec = AutoModel.load_from_folder(os.path.join(checkpoint_dir))
+
+        assert all(
+            [
+                torch.equal(
+                    model_rec.state_dict()[key].cpu(), model.state_dict()[key].cpu()
+                )
+                for key in model.state_dict().keys()
+            ]
+        )
+
+        assert type(model_rec.encoder.cpu()) == type(model.encoder.cpu())
+        assert type(model_rec.decoder.cpu()) == type(model.decoder.cpu())
+
+        optim_rec_state_dict = torch.load(os.path.join(checkpoint_dir, "optimizer.pt"))
+
+        assert all(
+            [
+                dict_rec == dict_optimizer
+                for (dict_rec, dict_optimizer) in zip(
+                    optim_rec_state_dict["param_groups"],
+                    optimizer.state_dict()["param_groups"],
+                )
+            ]
+        )
+
+        assert all(
+            [
+                dict_rec == dict_optimizer
+                for (dict_rec, dict_optimizer) in zip(
+                    optim_rec_state_dict["state"], optimizer.state_dict()["state"]
+                )
+            ]
+        )
+
+    def test_checkpoint_saving_during_training(self, vae, trainer, training_configs):
+        #
+        target_saving_epoch = training_configs.steps_saving
+
+        dir_path = training_configs.output_dir
+
+        model = deepcopy(trainer.model)
+
+        trainer.train()
+
+        training_dir = os.path.join(
+            dir_path, f"HRQVAE_training_{trainer._training_signature}"
+        )
+        assert os.path.isdir(training_dir)
+
+        checkpoint_dir = os.path.join(
+            training_dir, f"checkpoint_epoch_{target_saving_epoch}"
+        )
+
+        assert os.path.isdir(checkpoint_dir)
+
+        files_list = os.listdir(checkpoint_dir)
+
+        # check files
+        assert set(["model.pt", "optimizer.pt", "training_config.json"]).issubset(
+            set(files_list)
+        )
+
+        # check pickled custom decoder
+        if not vae.model_config.uses_default_decoder:
+            assert "decoder.pkl" in files_list
+
+        else:
+            assert not "decoder.pkl" in files_list
+
+        # check pickled custom encoder
+        if not vae.model_config.uses_default_encoder:
+            assert "encoder.pkl" in files_list
+
+        else:
+            assert not "encoder.pkl" in files_list
+
+        model_rec_state_dict = torch.load(os.path.join(checkpoint_dir, "model.pt"))[
+            "model_state_dict"
+        ]
+
+        assert not all(
+            [
+                torch.equal(model_rec_state_dict[key], model.state_dict()[key])
+                for key in model.state_dict().keys()
+            ]
+        )
+
+    def test_final_model_saving(self, vae, trainer, training_configs):
+        dir_path = training_configs.output_dir
+
+        trainer.train()
+
+        model = deepcopy(trainer._best_model)
+
+        training_dir = os.path.join(
+            dir_path, f"HRQVAE_training_{trainer._training_signature}"
+        )
+        assert os.path.isdir(training_dir)
+
+        final_dir = os.path.join(training_dir, f"final_model")
+        assert os.path.isdir(final_dir)
+
+        files_list = os.listdir(final_dir)
+
+        assert set(["model.pt", "model_config.json", "training_config.json"]).issubset(
+            set(files_list)
+        )
+
+        # check pickled custom decoder
+        if not vae.model_config.uses_default_decoder:
+            assert "decoder.pkl" in files_list
+
+        else:
+            assert not "decoder.pkl" in files_list
+
+        # check pickled custom encoder
+        if not vae.model_config.uses_default_encoder:
+            assert "encoder.pkl" in files_list
+
+        else:
+            assert not "encoder.pkl" in files_list
+
+        # check reload full model
+        model_rec = AutoModel.load_from_folder(os.path.join(final_dir))
+
+        assert all(
+            [
+                torch.equal(
+                    model_rec.state_dict()[key].cpu(), model.state_dict()[key].cpu()
+                )
+                for key in model.state_dict().keys()
+            ]
+        )
+
+        assert type(model_rec.encoder.cpu()) == type(model.encoder.cpu())
+        assert type(model_rec.decoder.cpu()) == type(model.decoder.cpu())
+
+    def test_vae_training_pipeline(self, vae, train_dataset, training_configs):
+        dir_path = training_configs.output_dir
+
+        # build pipeline
+        pipeline = TrainingPipeline(model=vae, training_config=training_configs)
+
+        # Launch Pipeline
+        pipeline(
+            train_data=train_dataset.data,  # gives tensor to pipeline
+            eval_data=train_dataset.data,  # gives tensor to pipeline
+        )
+
+        model = deepcopy(pipeline.trainer._best_model)
+
+        training_dir = os.path.join(
+            dir_path, f"HRQVAE_training_{pipeline.trainer._training_signature}"
+        )
+        assert os.path.isdir(training_dir)
+
+        final_dir = os.path.join(training_dir, f"final_model")
+        assert os.path.isdir(final_dir)
+
+        files_list = os.listdir(final_dir)
+
+        assert set(["model.pt", "model_config.json", "training_config.json"]).issubset(
+            set(files_list)
+        )
+
+        # check pickled custom decoder
+        if not vae.model_config.uses_default_decoder:
+            assert "decoder.pkl" in files_list
+
+        else:
+            assert not "decoder.pkl" in files_list
+
+        # check pickled custom encoder
+        if not vae.model_config.uses_default_encoder:
+            assert "encoder.pkl" in files_list
+
+        else:
+            assert not "encoder.pkl" in files_list
+
+        # check reload full model
+        model_rec = AutoModel.load_from_folder(os.path.join(final_dir))
+
+        assert all(
+            [
+                torch.equal(
+                    model_rec.state_dict()[key].cpu(), model.state_dict()[key].cpu()
+                )
+                for key in model.state_dict().keys()
+            ]
+        )
+
+        assert type(model_rec.encoder.cpu()) == type(model.encoder.cpu())
+        assert type(model_rec.decoder.cpu()) == type(model.decoder.cpu())
+
+
+class Test_HRQVAE_Generation:
+    @pytest.fixture
+    def train_data(self):
+        return torch.load(
+            os.path.join(PATH, "data/mnist_clean_train_dataset_sample")
+        ).data
+
+    @pytest.fixture()
+    def ae_model(self):
+        return HRQVAE(HRQVAEConfig(input_dim=(1, 28, 28), latent_dim=4))
+
+    @pytest.fixture(params=[PixelCNNSamplerConfig()])
+    def sampler_configs(self, request):
+        return request.param
+
+    @pytest.mark.skip(reason="Sampling not currently supported for HRQVAE")
+    def test_fits_in_generation_pipeline(self, ae_model, sampler_configs, train_data):
+        pipeline = GenerationPipeline(model=ae_model, sampler_config=sampler_configs)
+        gen_data = pipeline(
+            num_samples=11,
+            batch_size=7,
+            output_dir=None,
+            return_gen=True,
+            train_data=train_data,
+            eval_data=train_data,
+            training_config=BaseTrainerConfig(num_epochs=1),
+        )
+
+        assert gen_data.shape[0] == 11