main_ppo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "avrora, fop, jython, luindex,\n",
    "lusearch, lusearch-fix, pmd and xalan\n",
    "\n",
    "https://bergel.eu/MyPapers/Cana21a-JVMGlagsAndGA.pdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "! export WANDB_NOTEBOOK_NAME=\"main.ipynb\"\n",
    "import wandb\n",
    "\n",
    "PROJECT_NAME = \"Reinforcement Learning (DQN) - JVM-GC\"\n",
    "WANDB_KEY = \"4b077df3688052b0f43705d6b4d712c05fb979b7\"\n",
    "config = dict(\n",
    "    competition = PROJECT_NAME,\n",
    "    _wandb_kernel = 'lemon',\n",
    "    seed = 42,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import logging\n",
    "for handler in logging.root.handlers[:]:\n",
    "    logging.root.removeHandler(handler)\n",
    "\n",
    "logging.basicConfig(filename=\"gc-ml-ppo.log\",\n",
    "                    filemode='a',\n",
    "                    format='%(asctime)s,%(msecs)d %(name)s %(levelname)s %(message)s')\n",
    "\n",
    "\n",
    "logger = logging.getLogger()\n",
    "logger.setLevel(logging.DEBUG)\n",
    "# logger.setLevel(logging.INFO)\n",
    "# logger.error(\"test\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "from tqdm import tqdm\n",
    "\n",
    "from tf_agents.agents import PPOAgent\n",
    "from tf_agents.environments import tf_py_environment, BatchedPyEnvironment\n",
    "from tf_agents.networks import sequential\n",
    "from tf_agents.policies import random_tf_policy, policy_saver\n",
    "from tf_agents.replay_buffers import tf_uniform_replay_buffer\n",
    "from tf_agents.trajectories import trajectory\n",
    "from tf_agents.utils import common\n",
    "from tf_agents.specs import tensor_spec, from_spec\n",
    "from tf_agents import trajectories\n",
    "from tf_agents.networks.actor_distribution_network import ActorDistributionNetwork\n",
    "from tf_agents.networks.value_network import ValueNetwork\n",
    "from tf_agents.drivers.dynamic_episode_driver import DynamicEpisodeDriver\n",
    "\n",
    "\n",
    "# from env.PyEnvironments import CurveEnv, CurveMultipleEnv, JVMEnv\n",
    "# from env.PyEnvironments import JVMEnv\n",
    "from env.PyEnvironmentsTest import JVMEnv # !!!\n",
    "from util.plots_util import plot_dataset, plot_goal_heatmap\n",
    "\n",
    "tf.random.set_seed(42)\n",
    "np.random.seed(42) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka_2 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [1, 20, 0.00875, 0.9944960907170034, 0.9907217902495694, 0.0751173708920189, 0.0172935581495892, 0.0161681641767931, 0.9808917197452228, 1.0],\n",
      " Env. default goal value: 0.00875,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka_4 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [7, 8, 0.0102, 0.9957989346117848, 0.990684671313544, 0.7417840375586859, 0.3069606571552097, 0.3079898198519827, 0.9214437367303608, 0.5123893534258315],\n",
      " Env. default goal value: 0.0102,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka_8 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [10, 4, 0.01172, 0.990409976116464, 1.0, 0.8450704225352115, 0.7146562905317769, 0.717165574025787, 0.605095541401274, 0.3151704493246345],\n",
      " Env. default goal value: 0.01172,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka_16 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [4, 12, 0.01354, 0.990032042788194, 0.9991687697792734, 0.0751173708920189, 0.7276264591439688, 0.7303392963343571, 0.0403397027600849, 0.2466844313761527],\n",
      " Env. default goal value: 0.01354,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka_32 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [1, 12, 0.01053, 0.9106233378922056, 0.9352722028602332, 0.7089201877934275, 0.9801124081279724, 0.980083035790586, 0.4055201698513799, 0.2541578608839474],\n",
      " Env. default goal value: 0.01053,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora_2 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [7, 4, 0.00693, 0.9744866328328492, 0.9954249526515152, 0.5714285714285716, 0.5660377358490565, 0.5714176215843327, 0.5769230769230766, 0.0151724137931035],\n",
      " Env. default goal value: 0.00693,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora_4 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [10, 8, 0.00647, 0.5760536635598197, 0.5884297520661157, 0.1428571428571428, 0.150943396226415, 0.1428544053960831, 0.1442307692307691, 0.000689655172414],\n",
      " Env. default goal value: 0.00647,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora_8 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [10, 8, 0.00674, 0.7783191422679543, 0.7950413223140496, 0.4285714285714284, 0.4339622641509431, 0.42856321618825, 0.4326923076923075, 0.0372413793103452],\n",
      " Env. default goal value: 0.00674,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora_16 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [4, 12, 0.00928, 0.4859770502424289, 0.4964187327823691, 0.4285714285714284, 0.4339622641509431, 0.42856321618825, 0.4326923076923075, 0.0372413793103452],\n",
      " Env. default goal value: 0.00928,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora_32 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [7, 12, 0.00918, 0.8527517849665129, 0.8710743801652892, 0.8571428571428572, 0.8679245283018866, 0.8571455946039168, 0.8653846153846145, 0.96],\n",
      " Env. default goal value: 0.00918,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: h2 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [1, 20, 0.18954, 0.999979412534938, 0.9884644586588516, 1.0, 1.0, 0.9999987792999906, 1.0, 0.1472500455957914],\n",
      " Env. default goal value: 0.18954,\n",
      "\n"
     ]
    }
   ],
   "source": [
    "dataset_path = \"dataset/data\"\n",
    "tempdir = \"tmp\"\n",
    "checkpoint_dir = os.path.join(tempdir, 'checkpoint')\n",
    "policy_dir = os.path.join(tempdir, 'policy')\n",
    "\n",
    "env_args = {\n",
    "    \"jdk_path\": \"jdk-11.0.20.1.jdk\",\n",
    "    \"bm_path\": \"dacapo-bench.jar\",\n",
    "    \"gc_viewer_jar\": \"gcviewer-1.36.jar\",\n",
    "    \"callback_path\": \"callback/VMStatCallback.java\",\n",
    "    \"n\": 5,\n",
    "    \"goal\": \"avgGCPause\",\n",
    "    \"verbose\": False,\n",
    "}\n",
    "\n",
    "def get_tf_env(name, args):\n",
    "    env = JVMEnv(bm_name=name, **args)\n",
    "    tf_env = tf_py_environment.TFPyEnvironment(env, isolation=True) \n",
    "    return tf_env\n",
    "\n",
    "# Batch together multiple py environments and act as a single batch.\n",
    "names = [\"kafka_2\", \"kafka_4\", \"kafka_8\", \"kafka_16\", \"kafka_32\",\n",
    "         \"avrora_2\", \"avrora_4\", \"avrora_8\", \"avrora_16\", \"avrora_32\"]\n",
    "\n",
    "envs = [JVMEnv(bm_name=name, **env_args) for name in names]\n",
    "\n",
    "batched_env = BatchedPyEnvironment(envs)\n",
    "train_env = tf_py_environment.TFPyEnvironment(batched_env)\n",
    "\n",
    "test_env = get_tf_env(name=\"h2\", args=env_args)\n",
    "\n",
    "action_spec = from_spec(train_env.action_spec())\n",
    "observation_spec = from_spec(train_env.observation_spec())\n",
    "reward_spec = from_spec(train_env.reward_spec())\n",
    "time_step_spec = trajectories.time_step_spec(observation_spec, reward_spec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_q_net(action_spec, fc_layer_params = (100, 75, 50)):\n",
    "    action_tensor_spec = tensor_spec.from_spec(action_spec)\n",
    "    num_actions = action_tensor_spec.maximum - action_tensor_spec.minimum + 1\n",
    "\n",
    "    def dense_layer(num_units):\n",
    "        \"\"\"\n",
    "        Define a helper function to create Dense layers configured \n",
    "        with the right activation and kernel initializer.\n",
    "        \"\"\"\n",
    "        return tf.keras.layers.Dense(\n",
    "            num_units,\n",
    "            activation=tf.keras.activations.relu,\n",
    "            kernel_initializer=tf.keras.initializers.VarianceScaling(\n",
    "                scale=2.0, mode='fan_in', distribution='truncated_normal'))\n",
    "\n",
    "    # QNetwork consists of a sequence of Dense layers followed by a dense layer\n",
    "    # with `num_actions` units to generate one q_value per available action as\n",
    "    # its output.\n",
    "    dense_layers = [dense_layer(num_units) for num_units in fc_layer_params]\n",
    "    q_values_layer = tf.keras.layers.Dense(\n",
    "        num_actions,\n",
    "        activation=None,\n",
    "        kernel_initializer=tf.keras.initializers.RandomUniform(\n",
    "            minval=-0.03, maxval=0.03),\n",
    "        bias_initializer=tf.keras.initializers.Constant(-0.2))\n",
    "    q_net = sequential.Sequential(dense_layers + [q_values_layer])\n",
    "    return q_net\n",
    "\n",
    "def collect_step(environment, policy, replay_buffer):\n",
    "    time_step = environment.current_time_step()\n",
    "    action_step = policy.action(time_step)\n",
    "\n",
    "    next_time_step = environment.step(action_step.action)\n",
    "    # reward = next_time_step.reward\n",
    "    traj = trajectory.from_transition(time_step, action_step, next_time_step)\n",
    "    return traj\n",
    "\n",
    "def save_dataset(data, path):\n",
    "    step_counter = tf.Variable(0, trainable=False)\n",
    "    checkpoint_prefix = \"checkpoint\"\n",
    "    checkpoint_args = {\n",
    "        \"checkpoint_interval\": 50,\n",
    "        \"step_counter\": step_counter,\n",
    "        \"directory\": checkpoint_prefix,\n",
    "        \"max_to_keep\": 3,\n",
    "    }\n",
    "    try:\n",
    "        data.save(path, checkpoint_args=checkpoint_args)\n",
    "    except:\n",
    "        pass\n",
    "\n",
    "def save_rb(replay_buffer, path):\n",
    "    tf.train.Checkpoint(rb = replay_buffer).save(path)\n",
    "\n",
    "def restore_rb(replay_buffer, path):\n",
    "    tf.train.Checkpoint(rb = replay_buffer).restore(path)\n",
    "\n",
    "def get_dataset(\n",
    "    replay_buffer, \n",
    "    size, \n",
    "    batch_size, \n",
    "    collect_data_spec, \n",
    "    n_step_update, \n",
    "    create: bool = True, \n",
    "    save: bool = False):\n",
    "    \n",
    "    replay_buffer = tf_uniform_replay_buffer.TFUniformReplayBuffer(\n",
    "        data_spec=collect_data_spec, # agent.collect_data_spec\n",
    "        batch_size=1, # train_env.batch_size\n",
    "        max_length=size) # capacity\n",
    "\n",
    "    if create:\n",
    "        for _ in tqdm(range(size)):\n",
    "            traj = collect_step(train_env, random_policy, replay_buffer)\n",
    "            replay_buffer.add_batch(traj)\n",
    "\n",
    "        if save: save_rb(replay_buffer, dataset_path)\n",
    "    else: \n",
    "        restore_rb(replay_buffer, dataset_path+'-1')\n",
    "\n",
    "    dataset = replay_buffer.as_dataset(\n",
    "        num_parallel_calls=3, \n",
    "        sample_batch_size=batch_size,\n",
    "        num_steps=n_step_update + 1, \n",
    "        single_deterministic_pass=False).prefetch(3)\n",
    "    # iterator = iter(dataset)\n",
    "    return dataset\n",
    "    \n",
    "def compute_avg_return(environment, policy, num_episodes=50):\n",
    "    \"\"\"\n",
    "    Computes the average return of a policy, \n",
    "    given the policy, environment, and a number of episodes.\n",
    "\n",
    "    Note: for non-episodic tasks.\n",
    "    \"\"\"\n",
    "    total_return = 0.0\n",
    "    time_step = environment.reset()\n",
    "    for _ in tqdm(range(num_episodes)):\n",
    "        action_step = policy.action(time_step, seed=42)\n",
    "        time_step = environment.step(action_step.action)\n",
    "        total_return += time_step.reward\n",
    "\n",
    "    avg_return = total_return / num_episodes\n",
    "    return avg_return.numpy()[0]\n",
    "\n",
    "def get_env_state(environment):\n",
    "    return environment.current_time_step().observation.numpy().squeeze()[:3]\n",
    "\n",
    "def compute_avg_return_episodic(\n",
    "    environment, policy, num_episodes=10, patience=100,\n",
    "    print_info=False):\n",
    "    \"\"\"\n",
    "    Computes the average return of a policy, \n",
    "    given the policy, environment, and a number of episodes.\n",
    "\n",
    "    Note: for episodic tasks.\n",
    "    \"\"\"\n",
    "    total_return = 0.0\n",
    "    environment.reset()\n",
    "    # default_state = get_env_state(environment)\n",
    "    \n",
    "    for _ in range(num_episodes):\n",
    "        time_step = environment.reset()\n",
    "        episode_return = 0.0\n",
    "        i = 0\n",
    "        while not time_step.is_last():\n",
    "            if i >= patience:\n",
    "                break\n",
    "            action_step = policy.action(time_step)\n",
    "            time_step = environment.step(action_step.action)\n",
    "            obs = time_step.observation.numpy()[0]\n",
    "            rwd = time_step.reward.numpy()[0]\n",
    "            episode_return += time_step.reward\n",
    "            i += 1\n",
    "        total_return += episode_return / i\n",
    "    \n",
    "    # if print_info: print(\"default:\", default_state, \"current:\", get_env_state(environment))\n",
    "    avg_return = total_return / num_episodes\n",
    "    return avg_return.numpy()[0]\n",
    "\n",
    "def create_networks(observation_spec, action_spec, fc_layer_params):\n",
    "    actor_net = ActorDistributionNetwork(\n",
    "            observation_spec, # input\n",
    "            action_spec, # output\n",
    "            fc_layer_params=fc_layer_params,\n",
    "            activation_fn=tf.keras.activations.tanh)\n",
    "    \n",
    "    value_net = ValueNetwork(\n",
    "            observation_spec, # input\n",
    "            fc_layer_params=fc_layer_params,\n",
    "            activation_fn=tf.keras.activations.tanh)\n",
    "\n",
    "    return actor_net, value_net\n",
    "\n",
    "def plot_training(loss, rewards, num_steps, eval_interval):\n",
    "    fig, ax1 = plt.subplots()\n",
    "\n",
    "    steps = [step for step in range(0, num_steps, eval_interval)]\n",
    "    color = 'tab:red'\n",
    "    ax1.set_xlabel('steps')\n",
    "    ax1.set_ylabel('loss', color=color)\n",
    "    ax1.plot(steps, loss, color=color)\n",
    "    ax1.tick_params(axis='y', labelcolor=color)\n",
    "\n",
    "    ax2 = ax1.twinx()  # instantiate a second axes that shares the same x-axis\n",
    "\n",
    "    color = 'tab:blue'\n",
    "    ax2.set_ylabel('rewards', color=color)  # we already handled the x-label with ax1\n",
    "    ax2.plot(steps, rewards[:200], color=color)\n",
    "    ax2.tick_params(axis='y', labelcolor=color)\n",
    "\n",
    "    fig.tight_layout()  # otherwise the right y-label is slightly clipped\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_dataset_iter(envs, size, _agent, _train_episodes_per_iteration):\n",
    "    \"\"\"Get a dataset iterator with trajectories\n",
    "    collected from environments from list `envs`\n",
    "    using `_agent` collect policy. \n",
    "\n",
    "    Args:\n",
    "        envs (list(PyEnvironment)): List of PyEnvironments \n",
    "        to collect data from.\n",
    "        size (int):   A dataset size (number of trajectories\n",
    "        to collect.)\n",
    "        _agent (_type_): A Tensorflow Agent.\n",
    "        _train_episodes_per_iteration (int): \n",
    "            A number of trajectories per batch.\n",
    "\n",
    "    Returns:\n",
    "        iterator: A dataset iterator.\n",
    "    \"\"\"\n",
    "    from random import randrange\n",
    "    assert len(envs) >= 1, \"Environment list is empty!\"\n",
    "    num_iterations_to_reset = 50\n",
    "    replay_buffer = tf_uniform_replay_buffer.TFUniformReplayBuffer(\n",
    "        data_spec=_agent.collect_data_spec, # agent.collect_data_spec\n",
    "        batch_size=envs[0].batch_size, # train_env.batch_size\n",
    "        max_length=size) # capacity\n",
    "    \n",
    "    i = 0\n",
    "    for _ in tqdm(range(size)):\n",
    "        # Select a random environment from list.\n",
    "        indx = randrange(len(envs)) \n",
    "        env = envs[indx]\n",
    "        \n",
    "        if i >= num_iterations_to_reset:\n",
    "            env.reset()\n",
    "            \n",
    "        traj = collect_step(env, _agent.collect_policy, replay_buffer)\n",
    "        replay_buffer.add_batch(traj)\n",
    "        i += 1\n",
    "            \n",
    "    dataset = replay_buffer.as_dataset(\n",
    "        sample_batch_size=batch_size,\n",
    "        num_steps=train_episodes_per_iteration+1,\n",
    "        num_parallel_calls=_train_episodes_per_iteration).prefetch(_train_episodes_per_iteration)\n",
    "    dataset_iter = iter(dataset)\n",
    "    return dataset_iter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "collect_steps_per_iteration = 1\n",
    "dataset_size = 10000\n",
    "# fc_layer_params = (100, 75, 50)\n",
    "# fc_layer_params = (200, 100)\n",
    "# fc_layer_params = (128, 128, 128)\n",
    "fc_layer_params = (200, 150, 100, 75, 50)\n",
    "\n",
    "batch_size = 24\n",
    "learning_rate = 1e-3\n",
    "# learning_rate = 0.0005\n",
    "\n",
    "num_eval_episodes = 10  # @param {type:\"integer\"}\n",
    "eval_interval = 100  # @param {type:\"integer\"}\n",
    "\n",
    "# train_episodes_per_iteration = 10\n",
    "train_episodes_per_iteration = 5\n",
    "n_step_update = 4\n",
    "actor_net, value_net = create_networks(observation_spec, action_spec, fc_layer_params)\n",
    "global_step = tf.compat.v1.train.get_or_create_global_step()\n",
    "\n",
    "agent_args = {\n",
    "    \"optimizer\": tf.compat.v1.train.AdamOptimizer(learning_rate=learning_rate),\n",
    "    \"actor_net\": actor_net,\n",
    "    \"value_net\": value_net,\n",
    "    \"train_step_counter\": global_step,\n",
    "    # \"train_step_counter\": global_step,\n",
    "    # \"importance_ratio_clipping\": 0.1,\n",
    "    # \"num_epochs\": 20,\n",
    "    \"entropy_regularization\": 0.01,\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "agent = PPOAgent(\n",
    "  time_step_spec,\n",
    "  action_spec,\n",
    "  **agent_args\n",
    ")\n",
    "\n",
    "agent.initialize()\n",
    "agent.train = common.function(agent.train, autograph=False)\n",
    "\n",
    "random_policy = random_tf_policy.RandomTFPolicy(\n",
    "  time_step_spec = time_step_spec,\n",
    "  action_spec = action_spec,\n",
    "  automatic_state_reset=True,\n",
    "  clip=False,\n",
    "  emit_log_probability=True\n",
    ")\n",
    "\n",
    "def get_cd_and_rb(_env, _agent, size, n_steps):\n",
    "  replay_buffer = tf_uniform_replay_buffer.TFUniformReplayBuffer(\n",
    "      data_spec=_agent.collect_data_spec, # agent.collect_data_spec\n",
    "      batch_size=_env.batch_size, # train_env.batch_size\n",
    "      max_length=size) # capacity\n",
    "\n",
    "  replay_buffer_observer = replay_buffer.add_batch\n",
    "\n",
    "  collect_driver = DynamicEpisodeDriver(\n",
    "      _env,\n",
    "      _agent.collect_policy,\n",
    "      observers=[replay_buffer_observer], # + train_metrics,\n",
    "      num_episodes=n_steps)\n",
    "  \n",
    "  return collect_driver, replay_buffer\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "policy_dir = os.path.join(tempdir, '0_policy')\n",
    "tf_policy_saver = policy_saver.PolicySaver(agent.policy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Agent Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(_agent, _env_train, _env_val, \n",
    "          collect_driver, \n",
    "          replay_buffer, \n",
    "          steps: int = 5000, \n",
    "          use_wandb: bool = False,\n",
    "          eval_interval: int=100):\n",
    "    \"\"\"\n",
    "    Train reinforcement learning agent and evaluate\n",
    "    performance on a separate environment.\n",
    "    \"\"\"\n",
    "    \n",
    "    _env_train.reset()\n",
    "    _env_val.reset()\n",
    "    _agent.train_step_counter.assign(0)\n",
    "    _agent.train = common.function(_agent.train)\n",
    "    time_step = None\n",
    "    policy_state = _agent.collect_policy.get_initial_state(_env_train.batch_size)\n",
    "\n",
    "    loss = []\n",
    "    observations = []\n",
    "    rewards = []\n",
    "    for step in tqdm(range(steps)):\n",
    "        \n",
    "        time_step, policy_state = collect_driver.run(\n",
    "            time_step=time_step,\n",
    "            policy_state=policy_state,\n",
    "            maximum_iterations=50,\n",
    "        )\n",
    "\n",
    "        experience = replay_buffer.gather_all()\n",
    "        train_loss = _agent.train(experience)\n",
    "        replay_buffer.clear()\n",
    "        \n",
    "        if step % eval_interval == 0:\n",
    "            avg_reward = compute_avg_return_episodic(_env_val, _agent.policy, num_episodes=20)\n",
    "\n",
    "            loss.append(train_loss.loss.numpy())\n",
    "            # observations.append(obs)\n",
    "            rewards.append(avg_reward)\n",
    "\n",
    "            # wandb logger for tuning hyperparameters\n",
    "            if use_wandb:\n",
    "                wandb.log({'loss': train_loss.loss, 'reward': avg_reward})\n",
    "\n",
    "            print(f\"step = {step}: loss = {train_loss.loss}, reward = {avg_reward}\")\n",
    "    return loss, observations, rewards\n",
    "\n",
    "def _train(_agent, _env_train, _env_val, data_iterator,\n",
    "          steps: int = 5000, \n",
    "          use_wandb: bool = False,\n",
    "          eval_interval: int=100):\n",
    "    \"\"\"\n",
    "    Train reinforcement learning agent and evaluate\n",
    "    performance on a separate environment.\n",
    "    \"\"\"\n",
    "    \n",
    "    _env_train.reset()\n",
    "    _env_val.reset()\n",
    "    _agent.train_step_counter.assign(0)\n",
    "    _agent.train = common.function(_agent.train)\n",
    "    \n",
    "    time_step = None\n",
    "    policy_state = _agent.collect_policy.get_initial_state(_env_train.batch_size)\n",
    "\n",
    "    loss = []\n",
    "    observations = []\n",
    "    rewards = []\n",
    "    for step in tqdm(range(steps)):\n",
    "        \n",
    "        experience, _ = next(data_iterator)\n",
    "        train_loss = _agent.train(experience)\n",
    "        \n",
    "        if step % eval_interval == 0:\n",
    "            avg_reward = compute_avg_return_episodic(_env_val, _agent.policy, num_episodes=20)\n",
    "\n",
    "            loss.append(train_loss.loss.numpy())\n",
    "            # observations.append(obs)\n",
    "            rewards.append(avg_reward)\n",
    "\n",
    "            # wandb logger for tuning hyperparameters\n",
    "            if use_wandb:\n",
    "                wandb.log({'loss': train_loss.loss, 'reward': avg_reward})\n",
    "\n",
    "            print(f\"step = {step}: loss = {train_loss.loss}, reward = {avg_reward}\")\n",
    "    \n",
    "    return loss, observations, rewards"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train Sequentially"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [4, 8, 0.00762, 0.2416649425551018, 0.241660319886844, 0.0, 0.0, 0.0, 0.0, 0.9883585213252486],\n",
      " Env. default goal value: 0.00762,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [1, 4, 0.02897, 0.9999987255117688, 0.9999982161835892, 0.9482758620689656, 0.9495590828924162, 0.948075874239938, 0.9489194499017684, 1.5521971566331838e-05],\n",
      " Env. default goal value: 0.02897,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [10, 8, 0.0274, 0.9999984332584332, 0.9999978065957856, 0.9770114942528738, 0.978835978835979, 0.9781394505509832, 0.9783889980353636, 1.3150559243703395e-05],\n",
      " Env. default goal value: 0.0274,\n",
      "\n"
     ]
    }
   ],
   "source": [
    "num_steps = 5000\n",
    "\n",
    "train_env_copy = get_tf_env(\"avrora\", env_args)\n",
    "\n",
    "train_env_2 = get_tf_env(\"kafka\", env_args)\n",
    "train_env_2_copy = get_tf_env(\"kafka\", env_args)\n",
    "\n",
    "# test_env = get_tf_env(\"test\", env_args)\n",
    "# test_env_copy = get_tf_env(\"test\", env_args)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/5000 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.66it/s]\n",
      "  0%|          | 1/5000 [00:28<39:46:25, 28.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 0: loss = 28.902631759643555, reward = 0.009563921019434929\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.67it/s]/s] \n",
      "  2%|▏         | 101/5000 [01:19<5:27:18,  4.01s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = -0.12217696011066437, reward = 0.10741065442562103\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.69it/s]/s]  \n",
      "  4%|▍         | 201/5000 [02:11<5:17:22,  3.97s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = -0.12155486643314362, reward = -0.054414473474025726\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.64it/s]/s]  \n",
      "  6%|▌         | 301/5000 [03:05<5:19:49,  4.08s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = -0.14047378301620483, reward = -0.1808028370141983\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]/s]  \n",
      "  8%|▊         | 401/5000 [03:55<4:51:41,  3.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = -0.12208157032728195, reward = 0.13025769591331482\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.67it/s]/s]  \n",
      " 10%|█         | 501/5000 [04:46<4:57:33,  3.97s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = -0.12516961991786957, reward = -0.047113142907619476\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.66it/s]/s]  \n",
      " 12%|█▏        | 601/5000 [05:40<4:58:12,  4.07s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = -0.13263054192066193, reward = 0.023856637999415398\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]/s]  \n",
      " 14%|█▍        | 701/5000 [06:31<4:36:40,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = -0.12135445326566696, reward = -0.09978403896093369\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]/s]  \n",
      " 16%|█▌        | 801/5000 [07:22<4:32:15,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = -0.12417945265769958, reward = -0.12913194298744202\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.74it/s]/s]  \n",
      " 18%|█▊        | 901/5000 [08:13<4:21:34,  3.83s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = -0.12012037634849548, reward = -0.0812155231833458\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:13<00:00,  1.51it/s]t/s] \n",
      " 20%|██        | 1001/5000 [09:06<4:55:30,  4.43s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = -0.12780579924583435, reward = 0.04057605192065239\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.68it/s]t/s]  \n",
      " 22%|██▏       | 1101/5000 [09:57<4:17:19,  3.96s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = -0.15671901404857635, reward = -0.04562435299158096\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.67it/s]t/s]  \n",
      " 24%|██▍       | 1201/5000 [10:53<4:18:10,  4.08s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = -0.12041907012462616, reward = 0.04573701694607735\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.70it/s]t/s]  \n",
      " 26%|██▌       | 1301/5000 [11:46<4:03:05,  3.94s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = -0.11761971563100815, reward = -0.0033242765348404646\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]t/s]  \n",
      " 28%|██▊       | 1401/5000 [12:39<3:53:54,  3.90s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = -0.11471748352050781, reward = 0.0010996530763804913\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:17<00:00,  1.13it/s]t/s]  \n",
      " 30%|███       | 1501/5000 [13:38<5:33:23,  5.72s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = -0.11971862614154816, reward = -0.04055100679397583\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:18<00:00,  1.05it/s]t/s]  \n",
      " 32%|███▏      | 1601/5000 [14:42<5:44:22,  6.08s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = -0.11644572764635086, reward = -0.15625840425491333\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]t/s]  \n",
      " 34%|███▍      | 1701/5000 [15:34<3:33:43,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = -0.11681109666824341, reward = -0.21186129748821259\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.67it/s]t/s]  \n",
      " 36%|███▌      | 1801/5000 [16:26<3:33:53,  4.01s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = -0.1096460372209549, reward = -0.12031445652246475\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]t/s]  \n",
      " 38%|███▊      | 1901/5000 [17:18<3:20:41,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = -0.07287707179784775, reward = -0.1658892035484314\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:14<00:00,  1.34it/s]t/s]  \n",
      " 40%|████      | 2001/5000 [18:16<4:03:25,  4.87s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2000: loss = -0.11350805312395096, reward = -0.08294306695461273\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]  \n",
      " 42%|████▏     | 2101/5000 [19:07<3:02:34,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2100: loss = -0.11182614415884018, reward = -0.20583422482013702\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]  \n",
      " 44%|████▍     | 2201/5000 [19:57<2:56:22,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2200: loss = -0.11087335646152496, reward = -0.12639972567558289\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.58it/s]t/s]  \n",
      " 46%|████▌     | 2301/5000 [20:49<3:13:17,  4.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2300: loss = -0.09479192644357681, reward = -0.05476980656385422\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.66it/s]t/s]  \n",
      " 48%|████▊     | 2401/5000 [21:41<2:54:56,  4.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2400: loss = -0.0014683204935863614, reward = -0.16155274212360382\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:13<00:00,  1.46it/s]t/s]  \n",
      " 50%|█████     | 2501/5000 [22:37<3:08:39,  4.53s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2500: loss = -0.03699420019984245, reward = -0.20467133820056915\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.75it/s]t/s]  \n",
      " 52%|█████▏    | 2601/5000 [23:29<2:36:24,  3.91s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2600: loss = -0.10927896946668625, reward = -0.31928008794784546\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]  \n",
      " 54%|█████▍    | 2701/5000 [24:19<2:24:42,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2700: loss = -0.10431338846683502, reward = -0.15599094331264496\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 56%|█████▌    | 2801/5000 [25:09<2:18:55,  3.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2800: loss = -0.11629707366228104, reward = -0.24302449822425842\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.70it/s]t/s]  \n",
      " 58%|█████▊    | 2901/5000 [26:01<2:17:37,  3.93s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2900: loss = -0.09800298511981964, reward = -0.4591086804866791\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]  \n",
      " 60%|██████    | 3001/5000 [26:51<2:09:36,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3000: loss = -0.10718017816543579, reward = -0.2882995009422302\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.78it/s]t/s]  \n",
      " 62%|██████▏   | 3101/5000 [27:41<1:59:01,  3.76s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3100: loss = -0.10472153127193451, reward = -0.29563093185424805\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:14<00:00,  1.41it/s]t/s]  \n",
      " 64%|██████▍   | 3201/5000 [28:37<2:20:58,  4.70s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3200: loss = -0.11518708616495132, reward = -0.15007063746452332\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:14<00:00,  1.39it/s]t/s]  \n",
      " 66%|██████▌   | 3301/5000 [29:41<2:15:06,  4.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3300: loss = -0.08996358513832092, reward = -0.20825421810150146\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]t/s]  \n",
      " 68%|██████▊   | 3401/5000 [30:35<1:42:30,  3.85s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3400: loss = -0.06932719796895981, reward = -0.09602121263742447\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.75it/s]t/s]  \n",
      " 70%|███████   | 3501/5000 [31:26<1:35:16,  3.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3500: loss = 0.08029784262180328, reward = -0.14908118546009064\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.69it/s]t/s]  \n",
      " 72%|███████▏  | 3601/5000 [32:17<1:31:52,  3.94s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3600: loss = -0.10365701466798782, reward = -0.21505753695964813\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]  \n",
      " 74%|███████▍  | 3701/5000 [33:07<1:23:31,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3700: loss = -0.09262917190790176, reward = -0.19250598549842834\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.75it/s]t/s]  \n",
      " 76%|███████▌  | 3801/5000 [33:57<1:16:11,  3.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3800: loss = -0.09590969979763031, reward = -0.17102284729480743\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]t/s]  \n",
      " 78%|███████▊  | 3901/5000 [34:47<1:11:16,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3900: loss = -0.09566880762577057, reward = -0.2066037952899933\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.78it/s]t/s]  \n",
      " 80%|████████  | 4001/5000 [35:37<1:02:21,  3.75s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4000: loss = -0.09816298633813858, reward = -0.24682223796844482\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.74it/s]t/s]  \n",
      " 82%|████████▏ | 4101/5000 [36:29<57:54,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4100: loss = -0.10664721578359604, reward = -0.08116856217384338\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]\n",
      " 84%|████████▍ | 4201/5000 [37:21<51:33,  3.87s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4200: loss = -0.07868582010269165, reward = -0.04766662418842316\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:13<00:00,  1.46it/s]t/s]\n",
      " 86%|████████▌ | 4301/5000 [38:15<52:52,  4.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4300: loss = -0.07521019876003265, reward = -0.09636393934488297\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.68it/s]t/s]\n",
      " 88%|████████▊ | 4401/5000 [39:07<40:02,  4.01s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4400: loss = -0.10130394250154495, reward = -0.1365879327058792\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.69it/s]t/s]\n",
      " 90%|█████████ | 4501/5000 [39:59<32:47,  3.94s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4500: loss = -0.10128697752952576, reward = -0.011847374960780144\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]\n",
      " 92%|█████████▏| 4601/5000 [40:49<25:52,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4600: loss = -0.10102717578411102, reward = -0.12869268655776978\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]\n",
      " 94%|█████████▍| 4701/5000 [41:42<19:20,  3.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4700: loss = -0.10895377397537231, reward = 0.032198309898376465\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.63it/s]t/s]\n",
      " 96%|█████████▌| 4801/5000 [42:33<13:27,  4.06s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4800: loss = -0.10902483761310577, reward = -0.06275109946727753\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.74it/s]t/s]\n",
      " 98%|█████████▊| 4901/5000 [43:26<06:18,  3.82s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4900: loss = -0.12051510065793991, reward = -0.05913602560758591\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5000/5000 [44:05<00:00,  1.89it/s]\n",
      "100%|██████████| 50/50 [00:28<00:00,  1.77it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AVG RETURN - KAFKA: 0.0030280044\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Train on first\n",
    "collect_driver, replay_buffer = get_cd_and_rb(train_env, agent)\n",
    "loss, observations, rewards = train(\n",
    "    agent, train_env, train_env_copy, collect_driver, replay_buffer, steps = num_steps, eval_interval=100)\n",
    "\n",
    "print(\"AVG RETURN - KAFKA:\", \n",
    "      compute_avg_return_episodic(train_env_2, agent.policy, num_episodes=50))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training(loss, rewards, num_steps, 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:13<00:00,  1.51it/s]\n",
      "  0%|          | 1/5000 [00:18<26:03:03, 18.76s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 0: loss = 2.3935961723327637, reward = -0.08524542301893234\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]/s] \n",
      "  2%|▏         | 101/5000 [01:09<5:17:43,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = -0.12164812535047531, reward = -0.09026005119085312\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]/s]  \n",
      "  4%|▍         | 201/5000 [01:59<5:08:48,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = -0.12125816941261292, reward = -0.02092362567782402\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]/s]  \n",
      "  6%|▌         | 301/5000 [02:51<5:02:07,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = -0.12100689858198166, reward = -0.019010009244084358\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.66it/s]/s]  \n",
      "  8%|▊         | 401/5000 [03:42<5:07:32,  4.01s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = -0.12203968316316605, reward = -0.1057734340429306\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.64it/s]/s]  \n",
      " 10%|█         | 501/5000 [04:35<5:10:42,  4.14s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = 0.5449185967445374, reward = 0.20654818415641785\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]/s]  \n",
      " 12%|█▏        | 601/5000 [05:27<4:42:49,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = 0.12020786851644516, reward = -0.08972353488206863\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]/s]  \n",
      " 14%|█▍        | 701/5000 [06:17<4:36:43,  3.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = 0.1178484857082367, reward = 0.04620600491762161\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.63it/s]/s]  \n",
      " 16%|█▌        | 801/5000 [07:09<4:44:21,  4.06s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = 0.11423169821500778, reward = 0.05797123908996582\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:13<00:00,  1.50it/s]/s]  \n",
      " 18%|█▊        | 901/5000 [08:09<5:00:31,  4.40s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = 0.11202795803546906, reward = 0.08857231587171555\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.61it/s]t/s] \n",
      " 20%|██        | 1001/5000 [09:05<4:46:14,  4.29s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = 0.1183023676276207, reward = 0.06645048409700394\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:27<00:00,  1.38s/it]3s/it]\n",
      " 22%|██▏       | 1101/5000 [10:26<10:29:55,  9.69s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = 0.11778132617473602, reward = 0.084475576877594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.68it/s]t/s]   \n",
      " 24%|██▍       | 1201/5000 [11:28<4:33:47,  4.32s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = 0.11349839717149734, reward = 0.23287606239318848\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.64it/s]t/s]  \n",
      " 26%|██▌       | 1301/5000 [12:21<4:08:57,  4.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = 0.09798666089773178, reward = 0.15252044796943665\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]t/s]  \n",
      " 28%|██▊       | 1401/5000 [13:14<3:51:58,  3.87s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = 0.10044006258249283, reward = 0.2212250679731369\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.68it/s]t/s]  \n",
      " 30%|███       | 1501/5000 [14:09<3:53:43,  4.01s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = 0.08438204228878021, reward = 0.21633855998516083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:15<00:00,  1.27it/s]t/s]  \n",
      " 32%|███▏      | 1601/5000 [15:05<4:49:45,  5.11s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = 0.06959390640258789, reward = 0.2215852439403534\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.68it/s]t/s]  \n",
      " 34%|███▍      | 1701/5000 [15:59<3:39:43,  4.00s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = 0.07805465906858444, reward = 0.2126867026090622\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]  \n",
      " 36%|███▌      | 1801/5000 [16:50<3:21:09,  3.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = 0.11215945333242416, reward = 0.13990293443202972\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.67it/s]t/s]  \n",
      " 38%|███▊      | 1901/5000 [17:42<3:25:01,  3.97s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = 0.1106598749756813, reward = 0.24889183044433594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.75it/s]t/s]  \n",
      " 40%|████      | 2001/5000 [18:31<3:10:37,  3.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2000: loss = 0.07175104320049286, reward = 0.22520026564598083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]t/s]  \n",
      " 42%|████▏     | 2101/5000 [19:21<3:05:44,  3.84s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2100: loss = 0.05647712200880051, reward = 0.2317294329404831\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]  \n",
      " 44%|████▍     | 2201/5000 [20:11<2:56:22,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2200: loss = -0.12164752930402756, reward = 0.2467513382434845\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 46%|████▌     | 2301/5000 [21:01<2:50:55,  3.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2300: loss = 0.05813376232981682, reward = 0.2371426820755005\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.74it/s]t/s]  \n",
      " 48%|████▊     | 2401/5000 [21:50<2:46:00,  3.83s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2400: loss = 0.10486495494842529, reward = 0.28368085622787476\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 50%|█████     | 2501/5000 [22:40<2:39:08,  3.82s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2500: loss = 0.10581547021865845, reward = 0.21570351719856262\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]  \n",
      " 52%|█████▏    | 2601/5000 [23:30<2:34:59,  3.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2600: loss = 0.03669300675392151, reward = 0.2747664451599121\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.75it/s]t/s]  \n",
      " 54%|█████▍    | 2701/5000 [24:20<2:26:09,  3.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2700: loss = 0.049320586025714874, reward = 0.26521965861320496\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.70it/s]t/s]  \n",
      " 56%|█████▌    | 2801/5000 [25:10<2:23:52,  3.93s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2800: loss = 0.10751616209745407, reward = 0.03046208620071411\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 58%|█████▊    | 2901/5000 [26:00<2:12:47,  3.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2900: loss = 0.028574436902999878, reward = -0.078873410820961\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 60%|██████    | 3001/5000 [26:49<2:06:03,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3000: loss = 0.11305896937847137, reward = -0.0822506695985794\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]  \n",
      " 62%|██████▏   | 3101/5000 [27:39<2:02:51,  3.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3100: loss = 0.10693422704935074, reward = -0.13296742737293243\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 64%|██████▍   | 3201/5000 [28:29<1:53:45,  3.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3200: loss = 0.1026078313589096, reward = -0.054457150399684906\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.70it/s]t/s]  \n",
      " 66%|██████▌   | 3301/5000 [29:21<1:50:31,  3.90s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3300: loss = 0.09237673878669739, reward = -0.13591960072517395\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.74it/s]t/s]  \n",
      " 68%|██████▊   | 3401/5000 [30:11<1:42:14,  3.84s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3400: loss = 0.10058660060167313, reward = -0.06709173321723938\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 70%|███████   | 3501/5000 [31:00<1:34:43,  3.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3500: loss = 0.0451681949198246, reward = -0.06473610550165176\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]t/s]  \n",
      " 72%|███████▏  | 3601/5000 [31:50<1:30:36,  3.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3600: loss = 0.11888808757066727, reward = -0.14535395801067352\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 74%|███████▍  | 3701/5000 [32:40<1:22:02,  3.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3700: loss = 0.022169478237628937, reward = -0.06088108569383621\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 76%|███████▌  | 3801/5000 [33:30<1:15:42,  3.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3800: loss = -0.11868000030517578, reward = -0.07580452412366867\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.76it/s]t/s]  \n",
      " 78%|███████▊  | 3901/5000 [34:20<1:09:30,  3.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3900: loss = -0.10230109095573425, reward = -0.07970742881298065\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.71it/s]t/s]  \n",
      " 80%|████████  | 4001/5000 [35:10<1:04:59,  3.90s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4000: loss = 0.07089538872241974, reward = -0.0261381808668375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]  \n",
      " 82%|████████▏ | 4101/5000 [36:00<56:37,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4100: loss = -0.022438040003180504, reward = -0.09630493819713593\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]\n",
      " 84%|████████▍ | 4201/5000 [36:50<50:22,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4200: loss = 0.07430949807167053, reward = -0.061931073665618896\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.73it/s]t/s]\n",
      " 86%|████████▌ | 4301/5000 [37:40<45:08,  3.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4300: loss = 0.012303875759243965, reward = -0.1478734314441681\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.77it/s]t/s]\n",
      " 88%|████████▊ | 4401/5000 [38:29<37:45,  3.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4400: loss = -0.11188575625419617, reward = -0.1391969472169876\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.66it/s]t/s]\n",
      " 90%|█████████ | 4501/5000 [39:20<33:38,  4.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4500: loss = -0.055052660405635834, reward = -0.1591639518737793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.72it/s]t/s]\n",
      " 92%|█████████▏| 4601/5000 [40:12<25:47,  3.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4600: loss = -0.0020119748078286648, reward = -0.11100932210683823\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:12<00:00,  1.60it/s]t/s]\n",
      " 94%|█████████▍| 4701/5000 [41:04<20:39,  4.14s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4700: loss = -0.008959325030446053, reward = -0.08426783978939056\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:16<00:00,  1.25it/s]t/s]\n",
      " 96%|█████████▌| 4801/5000 [42:03<18:24,  5.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4800: loss = 0.004117509815841913, reward = -0.03693928197026253\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:11<00:00,  1.67it/s]t/s]\n",
      " 98%|█████████▊| 4901/5000 [42:54<06:33,  3.97s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4900: loss = 0.05862385034561157, reward = -0.06138558313250542\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5000/5000 [43:33<00:00,  1.91it/s]\n"
     ]
    }
   ],
   "source": [
    "loss, observations, rewards = train(\n",
    "    agent, train_env, train_env_copy, collect_driver, replay_buffer, steps = num_steps, eval_interval=100)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:32<00:00,  1.55it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AVG RETURN - KAFKA: -0.026877487\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "ename": "",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31mThe Kernel crashed while executing code in the the current cell or a previous cell. Please review the code in the cell(s) to identify a possible cause of the failure. Click <a href='https://aka.ms/vscodeJupyterKernelCrash'>here</a> for more info. View Jupyter <a href='command:jupyter.viewOutput'>log</a> for further details."
     ]
    }
   ],
   "source": [
    "print(\"AVG RETURN - KAFKA:\", \n",
    "      compute_avg_return_episodic(train_env_2, agent.policy, num_episodes=50))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.78it/s]\n",
      "  0%|          | 1/2000 [00:11<6:11:57, 11.16s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 0: loss = 9.476568222045898, reward = 0.08902159333229065\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.82it/s]/s]\n",
      "  5%|▌         | 101/2000 [01:47<1:35:29,  3.02s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = 0.059642937034368515, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.81it/s]/s]  \n",
      " 10%|█         | 201/2000 [03:26<1:31:07,  3.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = 0.0535656213760376, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.81it/s]/s]  \n",
      " 15%|█▌        | 301/2000 [05:03<1:25:42,  3.03s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = 0.04861774295568466, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.68it/s]/s]  \n",
      " 20%|██        | 401/2000 [06:41<1:23:49,  3.15s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = 0.04700877517461777, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.76it/s]/s]  \n",
      " 25%|██▌       | 501/2000 [08:21<1:17:13,  3.09s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = 0.02722451649606228, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.79it/s]/s]  \n",
      " 30%|███       | 601/2000 [10:00<1:11:00,  3.05s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = 0.047364652156829834, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.80it/s]/s]  \n",
      " 35%|███▌      | 701/2000 [11:37<1:05:47,  3.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = 0.04358407109975815, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.63it/s]/s]  \n",
      " 40%|████      | 801/2000 [13:19<1:03:25,  3.17s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = 0.035567134618759155, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.64it/s]/s]  \n",
      " 45%|████▌     | 901/2000 [15:03<59:11,  3.23s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = 0.04098467528820038, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.56it/s]t/s]\n",
      " 50%|█████     | 1001/2000 [16:45<54:24,  3.27s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = 0.0423898883163929, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.76it/s]t/s]\n",
      " 55%|█████▌    | 1101/2000 [18:25<45:51,  3.06s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = 0.0068428427912294865, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.76it/s]t/s]\n",
      " 60%|██████    | 1201/2000 [20:06<41:46,  3.14s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = 0.006304117850959301, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:08<00:00,  2.42it/s]t/s]\n",
      " 65%|██████▌   | 1301/2000 [21:48<39:23,  3.38s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = -0.049190741032361984, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.65it/s]t/s]\n",
      " 70%|███████   | 1401/2000 [23:27<31:34,  3.16s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = -0.05680543929338455, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.77it/s]t/s]\n",
      " 75%|███████▌  | 1501/2000 [25:04<25:30,  3.07s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = 0.03054279275238514, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.71it/s]t/s]\n",
      " 80%|████████  | 1601/2000 [26:44<20:48,  3.13s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = 0.005212224554270506, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.79it/s]t/s]\n",
      " 85%|████████▌ | 1701/2000 [28:25<15:17,  3.07s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = -0.012469688430428505, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.71it/s]t/s]\n",
      " 90%|█████████ | 1801/2000 [30:03<10:17,  3.10s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = -0.022902732715010643, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:07<00:00,  2.75it/s]t/s]\n",
      " 95%|█████████▌| 1901/2000 [31:42<05:06,  3.10s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = -3.552382986526936e-05, reward = 0.08900009095668793\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2000/2000 [33:16<00:00,  1.00it/s]\n",
      "100%|██████████| 50/50 [00:19<00:00,  2.58it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AVG RETURN - KAFKA: 0.08900006\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Train on second\n",
    "collect_driver, replay_buffer = get_cd_and_rb(train_env_2, agent)\n",
    "loss, observations, rewards = train(\n",
    "    agent, train_env_2, train_env_2_copy, collect_driver, replay_buffer, steps = num_steps, eval_interval=100)\n",
    "\n",
    "print(\"AVG RETURN - KAFKA:\", \n",
    "      compute_avg_return_episodic(train_env_2, agent.policy, num_episodes=50))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"AVG RETURN - TEST:\", \n",
    "      compute_avg_return_episodic(test_env, agent.policy, num_episodes=50))\n",
    "\n",
    "# Test environment\n",
    "collect_driver, replay_buffer = get_cd_and_rb(test_env, agent)\n",
    "loss, observations, rewards = train(\n",
    "    agent, test_env, test_env_copy, collect_driver, replay_buffer, steps = num_steps, eval_interval=100)\n",
    "\n",
    "print(\"AVG RETURN - TEST:\", \n",
    "      compute_avg_return_episodic(test_env, agent.policy, num_episodes=50))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Final results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 74.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AVG RETURN - AVRORA: 1.2897495\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:28<00:00,  1.78it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AVG RETURN - KAFKA: -0.6872998\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:27<00:00,  1.79it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AVG RETURN - TEST: 0.15185018\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"AVG RETURN - AVRORA:\", \n",
    "      compute_avg_return_episodic(train_env, agent.policy, num_episodes=50))\n",
    "\n",
    "print(\"AVG RETURN - KAFKA:\", \n",
    "      compute_avg_return_episodic(train_env_2, agent.policy, num_episodes=50))\n",
    "\n",
    "print(\"AVG RETURN - TEST:\", \n",
    "      compute_avg_return_episodic(test_env, agent.policy, num_episodes=50))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train on a single dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: avrora (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [13, 16, 0.01032, 0.0220530411853552, 0.0220509416656523, 0.0, 0.0, 0.0, 0.0, 0.9816754502341878],\n",
      " Env. default goal value: 0.01032,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [4, 20, 0.03266, 0.9999994550464116, 0.9999992366373232, 0.8793103448275863, 0.8807760141093476, 0.8774827845478718, 0.8781925343811396, 2.3067374411080144e-05],\n",
      " Env. default goal value: 0.03266,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: kafka (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [4, 12, 0.02978, 0.9999992902418988, 0.9999990063540516, 0.9655172413793104, 0.970194003527337, 0.9693582431571556, 0.9685658153241652, 2.0695962088444764e-05],\n",
      " Env. default goal value: 0.02978,\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50000/50000 [04:46<00:00, 174.50it/s]\n"
     ]
    }
   ],
   "source": [
    "num_steps = 5000\n",
    "\n",
    "\n",
    "# Collect trajectories from both benchmarks.\n",
    "dataset_iter = get_dataset_iter([train_env, train_env_2], 50000, agent, train_episodes_per_iteration)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/20000 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 4/20000 [00:11<12:12:23,  2.20s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 0: loss = 0.24985742568969727, reward = 0.03627775236964226\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 104/20000 [00:22<3:04:36,  1.80it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = 0.18949705362319946, reward = 0.03435734659433365\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 204/20000 [00:35<4:31:45,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = 0.15754052996635437, reward = 0.02616189420223236\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 304/20000 [00:47<4:28:42,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = 0.13620947301387787, reward = -0.019210290163755417\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 404/20000 [01:00<4:26:59,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = 0.2568551301956177, reward = 0.023504946380853653\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 504/20000 [01:13<4:25:48,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = 0.1418035924434662, reward = 0.018958551809191704\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 604/20000 [01:26<4:25:47,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = 0.35291072726249695, reward = 0.029474308714270592\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▎         | 704/20000 [01:39<4:24:15,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = 0.2931548058986664, reward = 0.017287787050008774\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 804/20000 [01:52<4:21:30,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = 0.3352110981941223, reward = 0.03280129283666611\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 904/20000 [02:05<4:20:48,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = 0.33797353506088257, reward = 0.026038503274321556\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 1004/20000 [02:18<4:21:05,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = 0.35232457518577576, reward = -0.037331633269786835\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 1102/20000 [02:31<6:10:08,  1.18s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = 0.3840436637401581, reward = 0.030994366854429245\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 1204/20000 [02:45<4:30:35,  1.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = 0.2831926941871643, reward = 0.023907102644443512\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 1304/20000 [02:58<4:21:50,  1.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = 0.32109829783439636, reward = 0.007971655577421188\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 1404/20000 [03:11<4:25:23,  1.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = 0.2960289716720581, reward = -0.01567075029015541\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 1504/20000 [03:25<4:24:35,  1.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = 0.1903543323278427, reward = 0.039607446640729904\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 1604/20000 [03:38<4:12:57,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = 0.14620988070964813, reward = 0.026723291724920273\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▊         | 1704/20000 [03:51<4:22:07,  1.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = 0.36498910188674927, reward = 0.02997639775276184\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 1804/20000 [04:04<4:10:39,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = 0.2329278588294983, reward = 0.022604594007134438\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|▉         | 1902/20000 [04:18<6:34:18,  1.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = 0.37557756900787354, reward = 0.018153684213757515\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|█         | 2002/20000 [04:34<6:46:35,  1.36s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2000: loss = 0.20745645463466644, reward = 0.047425199300050735\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 11%|█         | 2104/20000 [04:48<4:30:19,  1.10it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2100: loss = 0.12628722190856934, reward = 0.032698556780815125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 11%|█         | 2204/20000 [05:02<4:04:21,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2200: loss = 0.28112223744392395, reward = 0.032847389578819275\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 2304/20000 [05:15<3:59:44,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2300: loss = 0.23477435111999512, reward = 0.0321074053645134\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 2404/20000 [05:28<3:58:34,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2400: loss = 0.26505008339881897, reward = 0.01781424880027771\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 2502/20000 [05:41<5:29:08,  1.13s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2500: loss = 0.21987807750701904, reward = 0.02736610732972622\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 2604/20000 [05:54<3:55:44,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2600: loss = 0.21958446502685547, reward = -0.05722207576036453\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 14%|█▎        | 2704/20000 [06:07<3:54:58,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2700: loss = 0.40489062666893005, reward = 0.033309366554021835\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 14%|█▍        | 2804/20000 [06:20<3:52:06,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2800: loss = 0.21243752539157867, reward = -0.0046446374617516994\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 15%|█▍        | 2904/20000 [06:34<3:51:48,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2900: loss = 0.24446099996566772, reward = 0.03422443941235542\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 15%|█▌        | 3004/20000 [06:47<3:48:46,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3000: loss = 0.28121286630630493, reward = 0.03498591110110283\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 3104/20000 [07:00<3:48:32,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3100: loss = 0.2877977192401886, reward = -0.061251796782016754\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 3204/20000 [07:13<3:48:54,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3200: loss = 0.14211082458496094, reward = 0.022305238991975784\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 17%|█▋        | 3304/20000 [07:26<3:49:13,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3300: loss = 0.25767552852630615, reward = 0.03099989891052246\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 17%|█▋        | 3404/20000 [07:40<3:44:06,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3400: loss = 0.32213884592056274, reward = 0.015603247098624706\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 3504/20000 [07:53<3:42:42,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3500: loss = 0.20584708452224731, reward = 0.012885654345154762\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 3604/20000 [08:06<3:43:10,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3600: loss = 0.16325877606868744, reward = 0.002983792219310999\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 19%|█▊        | 3704/20000 [08:19<3:42:41,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3700: loss = 0.13512112200260162, reward = 0.028231313452124596\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 19%|█▉        | 3804/20000 [08:32<3:41:42,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3800: loss = 0.2544684410095215, reward = 0.01989850588142872\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 20%|█▉        | 3904/20000 [08:46<3:49:47,  1.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3900: loss = 0.1507875770330429, reward = 0.042065300047397614\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 20%|██        | 4004/20000 [08:59<3:39:01,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4000: loss = 0.11149635910987854, reward = -0.008994200266897678\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 21%|██        | 4104/20000 [09:12<3:37:33,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4100: loss = 0.2606261074542999, reward = 0.047553565353155136\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 21%|██        | 4204/20000 [09:26<3:39:37,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4200: loss = 0.3898237943649292, reward = 0.01164049468934536\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 4304/20000 [09:39<3:32:44,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4300: loss = 0.2230168581008911, reward = 0.026527341455221176\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 4404/20000 [09:52<3:30:50,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4400: loss = 0.1936362385749817, reward = 0.027814308181405067\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 23%|██▎       | 4504/20000 [10:05<3:29:42,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4500: loss = 0.2133648693561554, reward = 0.004976757802069187\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 23%|██▎       | 4604/20000 [10:18<3:30:23,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4600: loss = 0.3311423361301422, reward = 0.007316400296986103\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 24%|██▎       | 4702/20000 [10:32<4:53:40,  1.15s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4700: loss = 0.11626237630844116, reward = 0.038581907749176025\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 24%|██▍       | 4804/20000 [10:45<3:26:50,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4800: loss = 0.24157986044883728, reward = 0.022277645766735077\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|██▍       | 4904/20000 [10:58<3:23:26,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4900: loss = 0.29611992835998535, reward = 0.04321884363889694\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|██▌       | 5004/20000 [11:11<3:22:57,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5000: loss = 0.24908176064491272, reward = 0.023435909301042557\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▌       | 5104/20000 [11:24<3:22:51,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5100: loss = 0.4302345812320709, reward = 0.03838800638914108\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▌       | 5204/20000 [11:37<3:18:48,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5200: loss = 0.1361045390367508, reward = 0.010508047416806221\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 27%|██▋       | 5304/20000 [11:50<3:17:20,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5300: loss = 0.32617735862731934, reward = 0.001993041019886732\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 27%|██▋       | 5404/20000 [12:02<3:15:12,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5400: loss = 0.2851839065551758, reward = 0.01948023959994316\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 5504/20000 [12:15<3:15:47,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5500: loss = 0.43152138590812683, reward = 0.04435548558831215\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 5604/20000 [12:28<3:14:49,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5600: loss = 0.29081177711486816, reward = 0.022282103076577187\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 29%|██▊       | 5704/20000 [12:41<3:12:57,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5700: loss = 0.25021618604660034, reward = -0.06722939759492874\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 29%|██▉       | 5803/20000 [12:53<3:10:59,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5800: loss = 0.38924163579940796, reward = -0.014260141178965569\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 30%|██▉       | 5903/20000 [13:06<3:08:58,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5900: loss = 0.21488197147846222, reward = 0.01917889341711998\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 30%|███       | 6003/20000 [13:19<3:10:57,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6000: loss = 0.3656989634037018, reward = 0.02787599340081215\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 6103/20000 [13:31<3:06:15,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6100: loss = 0.24007469415664673, reward = 0.03764120116829872\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 6203/20000 [13:44<3:06:48,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6200: loss = 0.16221477091312408, reward = 0.010355299338698387\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 32%|███▏      | 6303/20000 [13:57<3:04:38,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6300: loss = 0.22436459362506866, reward = 0.03099079802632332\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 32%|███▏      | 6403/20000 [14:10<3:03:19,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6400: loss = 0.24129751324653625, reward = 0.03142649307847023\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 6503/20000 [14:22<3:01:36,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6500: loss = 0.25571560859680176, reward = 0.009862711653113365\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 6603/20000 [14:35<3:02:35,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6600: loss = 0.2787809371948242, reward = 0.019480206072330475\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▎      | 6703/20000 [14:48<3:00:37,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6700: loss = 0.2985495626926422, reward = 0.011412251740694046\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▍      | 6803/20000 [15:01<2:56:37,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6800: loss = 0.27407902479171753, reward = 0.036545880138874054\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 35%|███▍      | 6903/20000 [15:13<2:55:44,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6900: loss = 0.16550034284591675, reward = 0.030543338507413864\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 35%|███▌      | 7003/20000 [15:26<2:55:06,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7000: loss = 0.23837843537330627, reward = 0.0327099934220314\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 36%|███▌      | 7103/20000 [15:39<2:52:32,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7100: loss = 0.3921396732330322, reward = -0.02555229887366295\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 36%|███▌      | 7203/20000 [15:51<2:53:59,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7200: loss = 0.32735878229141235, reward = 0.03397664800286293\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 37%|███▋      | 7303/20000 [16:04<2:50:02,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7300: loss = 0.3671906888484955, reward = 0.03368739038705826\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 37%|███▋      | 7403/20000 [16:17<2:50:00,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7400: loss = 0.26604023575782776, reward = 0.017673656344413757\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 7503/20000 [16:30<2:47:31,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7500: loss = 0.24808797240257263, reward = 0.007826892659068108\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 7603/20000 [16:42<2:46:10,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7600: loss = 0.120560422539711, reward = 0.02314900793135166\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 39%|███▊      | 7703/20000 [16:55<2:45:45,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7700: loss = 0.32504144310951233, reward = 0.04141543060541153\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 39%|███▉      | 7803/20000 [17:08<2:44:22,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7800: loss = 0.5370333790779114, reward = 0.028494015336036682\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 40%|███▉      | 7903/20000 [17:20<2:41:54,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7900: loss = 0.1283600926399231, reward = 0.035373248159885406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 40%|████      | 8003/20000 [17:33<2:40:32,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8000: loss = 0.16220442950725555, reward = 0.016758259385824203\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 41%|████      | 8103/20000 [17:46<2:40:12,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8100: loss = 0.14994856715202332, reward = 0.021713051944971085\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 41%|████      | 8203/20000 [17:59<2:40:16,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8200: loss = 0.3100902736186981, reward = 0.02561871148645878\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 8303/20000 [18:11<2:36:57,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8300: loss = 0.28485986590385437, reward = 0.03194400668144226\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 8403/20000 [18:24<2:35:41,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8400: loss = 0.2703417241573334, reward = 0.0242760069668293\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 8503/20000 [18:37<2:33:42,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8500: loss = 0.20987364649772644, reward = 0.04921314865350723\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 8603/20000 [18:50<2:34:43,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8600: loss = 0.30480578541755676, reward = 0.045444004237651825\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████▎     | 8703/20000 [19:03<2:33:21,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8700: loss = 0.24081099033355713, reward = 0.019481603056192398\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████▍     | 8803/20000 [19:16<2:30:44,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8800: loss = 0.2606129050254822, reward = 0.022872157394886017\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 45%|████▍     | 8903/20000 [19:30<2:34:57,  1.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8900: loss = 0.2125171720981598, reward = -0.03956236317753792\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 45%|████▌     | 9003/20000 [19:43<2:29:08,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9000: loss = 0.2571754455566406, reward = 0.050605904310941696\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 46%|████▌     | 9103/20000 [19:56<2:28:38,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9100: loss = 0.17493905127048492, reward = 0.048502154648303986\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 46%|████▌     | 9203/20000 [20:09<2:25:08,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9200: loss = 0.3369608223438263, reward = 0.029967159032821655\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 9303/20000 [20:22<2:26:23,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9300: loss = 0.21062473952770233, reward = 0.02399105206131935\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 9403/20000 [20:35<2:24:00,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9400: loss = 0.22874122858047485, reward = 0.022599248215556145\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 48%|████▊     | 9503/20000 [20:50<2:34:25,  1.13it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9500: loss = 0.17614218592643738, reward = 0.010355794802308083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 48%|████▊     | 9603/20000 [21:03<2:23:57,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9600: loss = 0.35222911834716797, reward = 0.03868580237030983\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 49%|████▊     | 9703/20000 [21:16<2:24:42,  1.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9700: loss = 0.17193174362182617, reward = 0.03527269884943962\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 49%|████▉     | 9803/20000 [21:30<2:18:52,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9800: loss = 0.29037564992904663, reward = -0.010149320587515831\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|████▉     | 9903/20000 [21:43<2:15:56,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9900: loss = 0.12818235158920288, reward = 0.013634102419018745\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████     | 10003/20000 [21:56<2:14:27,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10000: loss = 0.18994712829589844, reward = 0.02275080420076847\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 51%|█████     | 10103/20000 [22:09<2:16:36,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10100: loss = 0.16500890254974365, reward = 0.0206805020570755\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 51%|█████     | 10203/20000 [22:22<2:18:18,  1.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10200: loss = 0.2985037863254547, reward = -0.014398030936717987\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 52%|█████▏    | 10303/20000 [22:36<2:11:03,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10300: loss = 0.2481953501701355, reward = 0.034351855516433716\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 52%|█████▏    | 10403/20000 [22:49<2:10:27,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10400: loss = 0.15936757624149323, reward = -0.010402058251202106\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 53%|█████▎    | 10503/20000 [23:02<2:08:55,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10500: loss = 0.2935214042663574, reward = -0.018190553411841393\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 53%|█████▎    | 10603/20000 [23:15<2:06:43,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10600: loss = 0.4175000488758087, reward = 0.02903285250067711\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 54%|█████▎    | 10703/20000 [23:28<2:06:07,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10700: loss = 0.3340311646461487, reward = 0.03699815273284912\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 54%|█████▍    | 10803/20000 [23:41<2:05:04,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10800: loss = 0.24288594722747803, reward = 0.01725880429148674\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 55%|█████▍    | 10903/20000 [23:55<2:02:40,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 10900: loss = 0.221414715051651, reward = 0.04057825356721878\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 55%|█████▌    | 11003/20000 [24:08<2:02:00,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11000: loss = 0.35102277994155884, reward = 0.027450602501630783\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|█████▌    | 11103/20000 [24:21<2:02:46,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11100: loss = 0.2549188435077667, reward = 0.03528924658894539\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|█████▌    | 11203/20000 [24:34<1:59:33,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11200: loss = 0.1860245019197464, reward = 0.01647549495100975\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 57%|█████▋    | 11303/20000 [24:47<1:57:49,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11300: loss = 0.42202672362327576, reward = 0.02129724994301796\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 57%|█████▋    | 11403/20000 [25:01<1:58:00,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11400: loss = 0.3794873058795929, reward = 0.039244603365659714\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 11503/20000 [25:14<1:58:13,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11500: loss = 0.42235517501831055, reward = 0.016923243179917336\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 11603/20000 [25:28<1:59:05,  1.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11600: loss = 0.31331300735473633, reward = -0.016771525144577026\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 59%|█████▊    | 11703/20000 [25:41<1:54:42,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11700: loss = 0.21873454749584198, reward = 0.01584719866514206\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 59%|█████▉    | 11803/20000 [25:54<1:52:59,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11800: loss = 0.339395135641098, reward = 0.031111741438508034\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 60%|█████▉    | 11903/20000 [26:08<1:50:02,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 11900: loss = 0.19148913025856018, reward = 0.011176547966897488\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 60%|██████    | 12003/20000 [26:21<1:49:57,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12000: loss = 0.4209326505661011, reward = 0.024248290807008743\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 61%|██████    | 12103/20000 [26:34<1:47:50,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12100: loss = 0.26758265495300293, reward = 0.04665274918079376\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 61%|██████    | 12203/20000 [26:48<1:49:44,  1.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12200: loss = 0.26957276463508606, reward = 0.018956348299980164\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 12303/20000 [27:01<1:46:40,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12300: loss = 0.3854345977306366, reward = 0.028666352853178978\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 12403/20000 [27:15<1:45:22,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12400: loss = 0.19581261277198792, reward = 0.020709004253149033\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 63%|██████▎   | 12503/20000 [27:28<1:45:26,  1.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12500: loss = 0.2152058482170105, reward = 0.030046189203858376\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 63%|██████▎   | 12603/20000 [27:42<1:45:17,  1.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12600: loss = 0.2161836475133896, reward = 0.034942347556352615\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 64%|██████▎   | 12703/20000 [27:56<1:44:20,  1.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12700: loss = 0.19305600225925446, reward = 0.018813790753483772\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 64%|██████▍   | 12803/20000 [28:09<1:39:23,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12800: loss = 0.1213333010673523, reward = 0.013052955269813538\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 65%|██████▍   | 12903/20000 [28:23<1:37:50,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 12900: loss = 0.24266910552978516, reward = 0.004309503361582756\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 65%|██████▌   | 13003/20000 [28:36<1:37:32,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13000: loss = 0.20264410972595215, reward = 0.015783360227942467\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▌   | 13103/20000 [28:51<1:39:10,  1.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13100: loss = 0.10965907573699951, reward = 0.02822929061949253\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▌   | 13203/20000 [29:04<1:33:21,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13200: loss = 0.18908382952213287, reward = 0.020571347326040268\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 13303/20000 [29:18<1:35:55,  1.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13300: loss = 0.29961344599723816, reward = 0.03018525242805481\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 13403/20000 [29:32<1:35:38,  1.15it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13400: loss = 0.4527949094772339, reward = 0.05688738822937012\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 68%|██████▊   | 13503/20000 [29:46<1:29:30,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13500: loss = 0.5011105537414551, reward = 0.010356593877077103\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 68%|██████▊   | 13603/20000 [29:59<1:27:36,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13600: loss = 0.23930104076862335, reward = 0.022572344169020653\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 69%|██████▊   | 13703/20000 [30:13<1:26:53,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13700: loss = 0.4600016474723816, reward = 0.03431164100766182\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 69%|██████▉   | 13803/20000 [30:26<1:24:48,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13800: loss = 0.205466166138649, reward = 0.03564384579658508\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|██████▉   | 13903/20000 [30:40<1:24:23,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 13900: loss = 0.23823849856853485, reward = 0.024357499554753304\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|███████   | 14003/20000 [30:53<1:22:08,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14000: loss = 0.21844594180583954, reward = 0.03459470346570015\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 71%|███████   | 14103/20000 [31:07<1:20:15,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14100: loss = 0.32898181676864624, reward = 0.01147269643843174\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 71%|███████   | 14203/20000 [31:20<1:19:29,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14200: loss = 0.27995678782463074, reward = 0.017521541565656662\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 14303/20000 [31:34<1:18:04,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14300: loss = 0.20431207120418549, reward = 0.015330428257584572\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 14403/20000 [31:47<1:19:58,  1.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14400: loss = 0.35720497369766235, reward = 0.0158720463514328\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 14503/20000 [32:01<1:14:49,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14500: loss = 0.13140425086021423, reward = 0.01639840006828308\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 14603/20000 [32:14<1:14:02,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14600: loss = 0.15935814380645752, reward = -0.012200744822621346\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 74%|███████▎  | 14703/20000 [32:28<1:19:27,  1.11it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14700: loss = 0.4161395728588104, reward = 0.013149097561836243\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 74%|███████▍  | 14803/20000 [32:42<1:10:40,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14800: loss = 0.20379342138767242, reward = 0.03221284970641136\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 75%|███████▍  | 14903/20000 [32:55<1:10:30,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 14900: loss = 0.30539557337760925, reward = 0.007336760871112347\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 75%|███████▌  | 15003/20000 [33:08<1:08:23,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15000: loss = 0.2678513526916504, reward = 0.0048437402583658695\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 76%|███████▌  | 15103/20000 [33:22<1:07:52,  1.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15100: loss = 0.2429802119731903, reward = -0.026663202792406082\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 76%|███████▌  | 15203/20000 [33:36<1:09:31,  1.15it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15200: loss = 0.2407996654510498, reward = 0.04678025096654892\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 15303/20000 [33:49<1:04:09,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15300: loss = 0.39074310660362244, reward = 0.009758999571204185\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 15403/20000 [34:04<1:08:39,  1.12it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15400: loss = 0.29942965507507324, reward = 0.03654450923204422\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 15504/20000 [34:17<1:01:22,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15500: loss = 0.36401963233947754, reward = 0.019089559093117714\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 15603/20000 [34:29<59:50,  1.22it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15600: loss = 0.2666255235671997, reward = 0.039123497903347015\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 79%|███████▊  | 15704/20000 [34:42<57:17,  1.25it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15700: loss = 0.30016130208969116, reward = 0.03198104351758957\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 79%|███████▉  | 15804/20000 [34:55<57:33,  1.22it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15800: loss = 0.4818044900894165, reward = 0.029416244477033615\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 80%|███████▉  | 15904/20000 [35:07<55:33,  1.23it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 15900: loss = 0.5133304595947266, reward = 0.031092459335923195\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 80%|████████  | 16004/20000 [35:20<54:08,  1.23it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16000: loss = 0.15663743019104004, reward = 0.020326707512140274\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████  | 16103/20000 [35:33<53:08,  1.22it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16100: loss = 0.2674199342727661, reward = 0.028164546936750412\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████  | 16204/20000 [35:45<51:54,  1.22it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16200: loss = 0.33699843287467957, reward = 0.01498035341501236\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 16303/20000 [35:58<50:00,  1.23it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16300: loss = 0.48910778760910034, reward = 0.023824337869882584\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 16404/20000 [36:10<48:52,  1.23it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16400: loss = 0.37667104601860046, reward = 0.018844105303287506\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 83%|████████▎ | 16504/20000 [36:23<49:09,  1.19it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16500: loss = 0.2992458939552307, reward = 0.020283302292227745\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 83%|████████▎ | 16604/20000 [36:36<45:41,  1.24it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16600: loss = 0.22386230528354645, reward = 0.021562248468399048\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 84%|████████▎ | 16704/20000 [36:48<44:25,  1.24it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16700: loss = 0.31734248995780945, reward = 0.008787089958786964\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 84%|████████▍ | 16804/20000 [37:01<43:09,  1.23it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16800: loss = 0.3774816691875458, reward = -0.005967739969491959\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 85%|████████▍ | 16904/20000 [37:14<42:34,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 16900: loss = 0.2948073148727417, reward = -0.027173522859811783\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 85%|████████▌ | 17003/20000 [37:26<39:48,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17000: loss = 0.2935958504676819, reward = -0.016064997762441635\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▌ | 17103/20000 [37:39<39:13,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17100: loss = 0.2580260634422302, reward = 0.009974196553230286\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▌ | 17204/20000 [37:52<38:21,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17200: loss = 0.2459775060415268, reward = 0.031969159841537476\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 87%|████████▋ | 17303/20000 [38:04<36:14,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17300: loss = 0.22677096724510193, reward = -0.0015499459113925695\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 87%|████████▋ | 17404/20000 [38:16<34:16,  1.26it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17400: loss = 0.29247766733169556, reward = 0.024155784398317337\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 17503/20000 [38:29<32:16,  1.29it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17500: loss = 0.2435913383960724, reward = 0.023485349491238594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 17603/20000 [38:41<31:56,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17600: loss = 0.27358850836753845, reward = 0.033354613929986954\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 89%|████████▊ | 17704/20000 [38:54<30:25,  1.26it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17700: loss = 0.36078697443008423, reward = 0.011208396404981613\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 89%|████████▉ | 17804/20000 [39:06<29:43,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17800: loss = 0.2712183892726898, reward = 0.033273398876190186\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 90%|████████▉ | 17903/20000 [39:19<28:53,  1.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 17900: loss = 0.41096603870391846, reward = 0.04305361211299896\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 90%|█████████ | 18004/20000 [39:31<27:18,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18000: loss = 0.19135867059230804, reward = 0.018954003229737282\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 91%|█████████ | 18103/20000 [39:44<25:22,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18100: loss = 0.203397735953331, reward = 0.03420184180140495\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 91%|█████████ | 18204/20000 [39:56<23:53,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18200: loss = 0.12625373899936676, reward = 0.037343043833971024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 18303/20000 [40:09<22:59,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18300: loss = 0.394363671541214, reward = 0.030170589685440063\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 18404/20000 [40:21<21:26,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18400: loss = 0.35190069675445557, reward = 0.02852560579776764\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 93%|█████████▎| 18503/20000 [40:33<19:12,  1.30it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18500: loss = 0.5232994556427002, reward = -0.0395599901676178\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 93%|█████████▎| 18603/20000 [40:46<18:45,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18600: loss = 0.30945855379104614, reward = 0.03961283713579178\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▎| 18703/20000 [40:58<17:18,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18700: loss = 0.3504294157028198, reward = 0.0208230409771204\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▍| 18804/20000 [41:11<16:22,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18800: loss = 0.141177698969841, reward = 0.008496701717376709\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 95%|█████████▍| 18904/20000 [41:23<14:40,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 18900: loss = 0.1896134912967682, reward = 0.032260406762361526\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 95%|█████████▌| 19003/20000 [41:36<13:32,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19000: loss = 0.28229570388793945, reward = -0.037842459976673126\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▌| 19104/20000 [41:49<12:30,  1.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19100: loss = 0.375541090965271, reward = 0.023044347763061523\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▌| 19203/20000 [42:01<10:45,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19200: loss = 0.2777419686317444, reward = 0.023801807314157486\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 97%|█████████▋| 19304/20000 [42:14<10:05,  1.15it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19300: loss = 0.21851059794425964, reward = 0.0218031108379364\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 97%|█████████▋| 19404/20000 [42:27<07:44,  1.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19400: loss = 0.3905331790447235, reward = 0.024327492341399193\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 19504/20000 [42:40<06:27,  1.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19500: loss = 0.2196487933397293, reward = 0.024008430540561676\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 19604/20000 [42:52<05:25,  1.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19600: loss = 0.16009560227394104, reward = 0.03513675183057785\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 99%|█████████▊| 19704/20000 [43:05<03:59,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19700: loss = 0.30739906430244446, reward = 0.03955841436982155\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 99%|█████████▉| 19803/20000 [43:17<02:40,  1.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19800: loss = 0.2997934818267822, reward = 0.026838252320885658\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████▉| 19903/20000 [43:30<01:10,  1.37it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 19900: loss = 0.25723883509635925, reward = 0.03693921118974686\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20000/20000 [43:35<00:00,  7.65it/s]\n"
     ]
    }
   ],
   "source": [
    "# Train on Avrora, Evaluate on Kafka\n",
    "loss, observations, rewards = _train(\n",
    "    agent, train_env, train_env_2, dataset_iter, steps = 20000, eval_interval=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: h2 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [13, 8, 0.14861, 0.9999572476715284, 0.9883840216515078, 0.983277591973244, 0.9828115408225904, 0.999997114709069, 1.0, 0.1393387985045442],\n",
      " Env. default goal value: 0.14861,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: h2 (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16, 'step': 3}, 'ParallelGCThreads': {'min': 4, 'max': 24, 'step': 4}},\n",
      " Env. default state: [1, 12, 0.15362, 0.999997745946094, 0.988448062333557, 0.983277591973244, 0.981379169224473, 0.9999992231909032, 1.0, 0.1402957711681965],\n",
      " Env. default goal value: 0.15362,\n",
      "\n"
     ]
    }
   ],
   "source": [
    "test_env= get_tf_env(\"h2\", env_args)\n",
    "test_env_copy = get_tf_env(\"h2\", env_args)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "default: [1.00e+01 1.60e+01 9.83e-03] current: [1.000e+00 1.600e+01 1.116e-02]\n",
      "AVG RETURN - AVRORA: 0.11686779\n",
      "default: [ 4.      12.       0.02978] current: [1.      4.      0.02897]\n",
      "AVG RETURN - KAFKA: 0.020553958\n",
      "default: [ 4.      16.       0.13649] current: [1.      4.      0.18607]\n",
      "AVG RETURN - H2: 0.06736073\n"
     ]
    }
   ],
   "source": [
    "print(\"AVG RETURN - AVRORA:\", \n",
    "      compute_avg_return_episodic(train_env, agent.policy, num_episodes=50, print_info=True))\n",
    "\n",
    "print(\"AVG RETURN - KAFKA:\", \n",
    "      compute_avg_return_episodic(train_env_2, agent.policy, num_episodes=50, print_info=True))\n",
    "\n",
    "print(\"AVG RETURN - H2:\", \n",
    "      compute_avg_return_episodic(test_env, agent.policy, num_episodes=50, print_info=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_checkpointer.save(global_step)\n",
    "tf_policy_saver.save(policy_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training(loss, rewards, num_steps, 100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Common collect driver"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_steps = 10000\n",
    "collect_driver, replay_buffer = get_cd_and_rb(train_env, agent, dataset_size, train_episodes_per_iteration)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/10000 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 1/10000 [00:18<50:22:50, 18.14s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 0: loss = 48.937225341796875, reward = -0.17753641307353973\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 101/10000 [01:01<7:09:21,  2.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = -0.029841536656022072, reward = -0.4798048138618469\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 201/10000 [01:52<7:38:07,  2.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = -0.04589027166366577, reward = 0.0646543949842453\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 301/10000 [02:54<8:11:22,  3.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = -0.0024730374570935965, reward = 0.1006183847784996\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 401/10000 [03:46<7:03:52,  2.65s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = 0.008433199487626553, reward = 0.0948130339384079\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 501/10000 [04:38<6:48:37,  2.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = -0.005747173447161913, reward = 0.09036823362112045\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 601/10000 [05:30<6:52:01,  2.63s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = -0.015521685592830181, reward = 0.06981039047241211\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 701/10000 [06:30<8:06:04,  3.14s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = -0.0054346658289432526, reward = 0.07592859119176865\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 801/10000 [07:30<7:48:58,  3.06s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = -0.009554612450301647, reward = 0.06064216047525406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 901/10000 [08:27<7:47:50,  3.09s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = -0.013438339345157146, reward = 0.13207818567752838\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|█         | 1001/10000 [09:21<6:25:21,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = -0.02279972843825817, reward = 0.07570149004459381\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 11%|█         | 1101/10000 [10:12<6:18:53,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = -0.02009693719446659, reward = 0.06107410043478012\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 1201/10000 [11:01<6:13:26,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = -0.03956253081560135, reward = -0.14187034964561462\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 1301/10000 [11:52<6:09:56,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = 0.006894805934280157, reward = -0.08655361086130142\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 14%|█▍        | 1401/10000 [12:41<6:05:52,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = -0.011612087488174438, reward = 0.08446063101291656\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 15%|█▌        | 1501/10000 [13:32<6:04:33,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = 0.13603658974170685, reward = -0.004821706563234329\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 1601/10000 [14:24<6:06:23,  2.62s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = -0.026009418070316315, reward = 0.095827117562294\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 17%|█▋        | 1701/10000 [15:13<5:53:48,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = -0.002900458872318268, reward = 0.02800840511918068\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 1801/10000 [16:02<5:48:56,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = -0.0037551005370914936, reward = 0.07382015883922577\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 19%|█▉        | 1901/10000 [16:53<6:16:09,  2.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = 0.003202388994395733, reward = 0.04558534920215607\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 20%|██        | 2001/10000 [17:45<5:43:12,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2000: loss = -0.01820695959031582, reward = 0.07326477766036987\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 21%|██        | 2101/10000 [18:38<5:53:28,  2.68s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2100: loss = -0.020956963300704956, reward = 0.05151586979627609\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 2201/10000 [19:28<5:44:00,  2.65s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2200: loss = -0.02764793112874031, reward = 0.08506140112876892\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 23%|██▎       | 2301/10000 [20:18<5:59:03,  2.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2300: loss = 0.001491648145020008, reward = 0.10345141589641571\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 24%|██▍       | 2401/10000 [21:08<5:45:57,  2.73s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2400: loss = -0.0023741128388792276, reward = 0.05721893161535263\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|██▌       | 2501/10000 [21:58<5:30:18,  2.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2500: loss = -0.026749931275844574, reward = -0.3260703980922699\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▌       | 2601/10000 [22:49<5:48:27,  2.83s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2600: loss = -0.009807603433728218, reward = -0.4977230131626129\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 27%|██▋       | 2701/10000 [23:40<5:23:21,  2.66s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2700: loss = -0.003918954636901617, reward = -0.5366655588150024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 2801/10000 [24:32<5:22:57,  2.69s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2800: loss = -0.007426843512803316, reward = -0.515765368938446\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 29%|██▉       | 2901/10000 [25:23<5:12:43,  2.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2900: loss = -0.027111032977700233, reward = -0.6060086488723755\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 30%|███       | 3001/10000 [26:14<5:01:33,  2.59s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3000: loss = -0.009513258934020996, reward = 0.05808001756668091\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 3101/10000 [27:06<5:15:35,  2.74s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3100: loss = -0.09422209858894348, reward = -0.010376416146755219\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 32%|███▏      | 3201/10000 [28:00<5:23:22,  2.85s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3200: loss = -0.022492051124572754, reward = -0.1794758141040802\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 3301/10000 [28:54<5:09:37,  2.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3300: loss = -0.03663545101881027, reward = -0.08150073885917664\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▍      | 3401/10000 [29:49<4:57:58,  2.71s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3400: loss = -0.014627097174525261, reward = -0.14371761679649353\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 35%|███▌      | 3501/10000 [30:45<4:53:07,  2.71s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3500: loss = -0.006813403218984604, reward = -0.16940483450889587\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 36%|███▌      | 3601/10000 [31:53<11:02:02,  6.21s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3600: loss = -0.0007351728272624314, reward = -0.1557752788066864\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 37%|███▋      | 3701/10000 [32:53<5:00:30,  2.86s/it] "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3700: loss = 0.0016316949622705579, reward = -0.10929298400878906\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 3801/10000 [33:48<4:42:03,  2.73s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3800: loss = -0.006584027782082558, reward = -0.1447393149137497\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 39%|███▉      | 3901/10000 [34:41<4:28:01,  2.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3900: loss = -0.039411235600709915, reward = -0.12452354282140732\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 40%|████      | 4001/10000 [35:35<4:22:33,  2.63s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4000: loss = -0.014827650040388107, reward = -0.1095043420791626\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 41%|████      | 4101/10000 [36:29<4:50:42,  2.96s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4100: loss = -0.013007811270654202, reward = -0.14809484779834747\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 4201/10000 [37:38<4:22:12,  2.71s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4200: loss = -0.006307572592049837, reward = -0.13430342078208923\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 4301/10000 [38:30<4:16:43,  2.70s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4300: loss = 0.003515255870297551, reward = -0.12775585055351257\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████▍     | 4401/10000 [39:23<4:21:52,  2.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4400: loss = 0.004743673373013735, reward = -0.11923233419656754\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 45%|████▌     | 4501/10000 [40:19<4:01:43,  2.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4500: loss = 0.005955316126346588, reward = -0.12080593407154083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 46%|████▌     | 4601/10000 [41:11<3:53:44,  2.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4600: loss = -0.05429302901029587, reward = -0.08684021234512329\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 4701/10000 [42:00<3:44:05,  2.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4700: loss = 0.005476806312799454, reward = 0.005218320991843939\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 48%|████▊     | 4801/10000 [42:51<3:50:34,  2.66s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4800: loss = 0.002454745350405574, reward = -0.02282392606139183\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 49%|████▉     | 4901/10000 [43:41<3:38:04,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4900: loss = 0.00216989335604012, reward = 0.008491285145282745\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████     | 5001/10000 [44:32<3:33:03,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5000: loss = 0.007778072264045477, reward = -0.05301349237561226\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 51%|█████     | 5101/10000 [45:22<3:27:46,  2.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5100: loss = -0.10312110930681229, reward = 0.1318182647228241\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 52%|█████▏    | 5201/10000 [46:12<3:24:23,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5200: loss = -0.00267457845620811, reward = -0.1441565603017807\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 53%|█████▎    | 5301/10000 [47:02<3:23:29,  2.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5300: loss = -0.009113558568060398, reward = -0.008346456103026867\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 54%|█████▍    | 5401/10000 [47:52<3:16:34,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5400: loss = -0.004117089323699474, reward = -0.005369010381400585\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 55%|█████▌    | 5501/10000 [48:43<3:11:44,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5500: loss = -0.0033366698771715164, reward = 0.006968374364078045\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|█████▌    | 5601/10000 [49:33<3:07:15,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5600: loss = 0.0053539699874818325, reward = -0.3954353928565979\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 57%|█████▋    | 5701/10000 [50:23<3:01:32,  2.53s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5700: loss = -0.005777641199529171, reward = -0.4003916382789612\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 5801/10000 [51:13<2:58:54,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5800: loss = -0.0013245362788438797, reward = -0.08402419835329056\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 59%|█████▉    | 5901/10000 [52:03<2:53:27,  2.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 5900: loss = 0.004345765803009272, reward = -0.11660643666982651\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 60%|██████    | 6001/10000 [52:53<2:50:32,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6000: loss = -0.00141683139372617, reward = -0.37293437123298645\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 61%|██████    | 6101/10000 [53:43<2:46:44,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6100: loss = 0.0020304021891206503, reward = -0.46715623140335083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 6201/10000 [54:33<2:41:45,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6200: loss = 0.002308701630681753, reward = -0.4524999260902405\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 63%|██████▎   | 6301/10000 [55:23<2:36:05,  2.53s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6300: loss = 0.000125030055642128, reward = -0.5659657120704651\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 64%|██████▍   | 6401/10000 [56:13<2:32:57,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6400: loss = -0.0006637836340814829, reward = -0.5557823777198792\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 65%|██████▌   | 6501/10000 [57:03<2:28:50,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6500: loss = -0.003973644692450762, reward = 0.05257822945713997\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▌   | 6601/10000 [57:53<2:24:07,  2.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6600: loss = -0.007424875628203154, reward = 0.02047843299806118\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 6701/10000 [58:43<2:20:26,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6700: loss = -0.03125167265534401, reward = -0.08490550518035889\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 68%|██████▊   | 6801/10000 [59:34<2:20:06,  2.63s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6800: loss = -0.006715892814099789, reward = -0.20727285742759705\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 69%|██████▉   | 6901/10000 [1:00:26<2:12:41,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 6900: loss = -0.002833651378750801, reward = -0.14743776619434357\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|███████   | 7001/10000 [1:01:17<2:11:12,  2.62s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7000: loss = -0.0018620153423398733, reward = -0.03974051773548126\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 71%|███████   | 7101/10000 [1:02:21<2:05:20,  2.59s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7100: loss = -0.006395635660737753, reward = -0.13103936612606049\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 7201/10000 [1:03:15<2:01:14,  2.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7200: loss = -0.0013959211064502597, reward = -0.09121858328580856\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 7301/10000 [1:04:08<1:56:31,  2.59s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7300: loss = -0.030369849875569344, reward = -0.13139879703521729\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 74%|███████▍  | 7401/10000 [1:05:01<1:52:57,  2.61s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7400: loss = -0.0003544760402292013, reward = -0.17020411789417267\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 75%|███████▌  | 7501/10000 [1:05:54<1:47:44,  2.59s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7500: loss = -0.0017358472105115652, reward = -0.09534154832363129\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 76%|███████▌  | 7601/10000 [1:06:47<1:44:21,  2.61s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7600: loss = -0.00711114052683115, reward = -0.09150321781635284\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 7701/10000 [1:07:40<1:38:42,  2.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7700: loss = -0.01760304719209671, reward = 0.01841781474649906\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 7801/10000 [1:08:33<1:34:56,  2.59s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7800: loss = -0.0004201991541776806, reward = 0.11232483386993408\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 79%|███████▉  | 7901/10000 [1:09:25<1:30:09,  2.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 7900: loss = 0.0022978587076067924, reward = 0.0885215550661087\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 80%|████████  | 8001/10000 [1:10:18<1:27:39,  2.63s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8000: loss = 0.0017010057345032692, reward = 0.06793919205665588\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████  | 8101/10000 [1:11:10<1:21:48,  2.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8100: loss = 0.0019339965656399727, reward = 0.08732522279024124\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 8201/10000 [1:12:02<1:17:01,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8200: loss = -0.001657668617554009, reward = 0.023821990936994553\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 83%|████████▎ | 8301/10000 [1:15:04<1:13:48,  2.61s/it] "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8300: loss = -0.002437570597976446, reward = 0.10907639563083649\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 84%|████████▍ | 8401/10000 [1:15:55<1:08:07,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8400: loss = -0.002247120952233672, reward = 0.038752634078264236\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 85%|████████▌ | 8501/10000 [1:16:46<1:04:17,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8500: loss = -0.0022757744882255793, reward = 0.06408777832984924\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▌ | 8601/10000 [1:17:36<1:00:02,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8600: loss = -0.004113011993467808, reward = -0.0400715097784996\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 87%|████████▋ | 8701/10000 [1:18:27<56:11,  2.60s/it]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8700: loss = -0.002262270776554942, reward = -0.05166323855519295\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 8801/10000 [1:19:18<51:27,  2.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8800: loss = -0.0010613177437335253, reward = -0.09124276041984558\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 89%|████████▉ | 8901/10000 [1:20:09<46:44,  2.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 8900: loss = -0.002044474007561803, reward = -0.5163190364837646\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 90%|█████████ | 9001/10000 [1:21:00<42:35,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9000: loss = -0.0025240506511181593, reward = -0.572773814201355\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 91%|█████████ | 9101/10000 [1:21:51<38:26,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9100: loss = -0.0029183807782828808, reward = -0.5777910947799683\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 9201/10000 [1:22:43<34:21,  2.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9200: loss = -0.0010590116726234555, reward = -0.5694754123687744\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 93%|█████████▎| 9301/10000 [1:23:34<29:59,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9300: loss = -0.00102526496630162, reward = 0.052893705666065216\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▍| 9401/10000 [1:24:25<25:38,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9400: loss = -0.0010977387428283691, reward = -0.520660936832428\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 95%|█████████▌| 9501/10000 [1:25:17<21:17,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9500: loss = -0.016466550529003143, reward = -0.010863653384149075\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▌| 9601/10000 [1:26:08<16:59,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9600: loss = -0.004927287343889475, reward = 0.09239820390939713\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 97%|█████████▋| 9701/10000 [1:26:59<12:47,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9700: loss = -0.05658155307173729, reward = 0.10042295604944229\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 9801/10000 [1:27:50<08:30,  2.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9800: loss = -0.007920180447399616, reward = 0.06870749592781067\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 99%|█████████▉| 9901/10000 [1:28:41<04:14,  2.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 9900: loss = -0.0016685246955603361, reward = 0.05374101921916008\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 10000/10000 [1:29:25<00:00,  1.86it/s]\n"
     ]
    }
   ],
   "source": [
    "loss, observations, rewards = train(\n",
    "    agent, train_env, test_env, collect_driver, replay_buffer, steps = num_steps, eval_interval=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/ellkrauze/projects/gc-ml/gc-ml-env/lib/python3.8/site-packages/tensorflow/python/saved_model/nested_structure_coder.py:475: UserWarning: Encoding a StructuredValue with type tfp.distributions.Deterministic_ACTTypeSpec; loading this StructuredValue will require that this type be imported and registered.\n",
      "  warnings.warn(\"Encoding a StructuredValue with type %s; loading this \"\n"
     ]
    }
   ],
   "source": [
    "tf_policy_saver.save(policy_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training(loss, rewards, num_steps, 100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test(env, policy, patience: int=50, verbose: bool=False, num_flags: int=2):\n",
    "    i = 0\n",
    "    _observations = np.array([])\n",
    "    _actions = np.array([])\n",
    "    \n",
    "    time_step = env.reset()\n",
    "    default_state = get_env_state(env)\n",
    "    \n",
    "    while not time_step.is_last():\n",
    "        if i >= patience:\n",
    "            break\n",
    "        \n",
    "        _observations = np.append(_observations, time_step.observation.numpy().squeeze()[:num_flags])\n",
    "        \n",
    "        action_step = policy.action(time_step)\n",
    "        time_step = env.step(action_step.action)\n",
    "        \n",
    "        _actions = np.append(_actions, action_step.action.numpy()[0])\n",
    "\n",
    "        i += 1\n",
    "        \n",
    "    if verbose:\n",
    "        print(\"start:\", *default_state, \n",
    "          \"recommendation:\", *get_env_state(env))\n",
    "    \n",
    "    return _observations, _actions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Getting 5000 samples...\n"
     ]
    }
   ],
   "source": [
    "# observations = np.array([])\n",
    "# actions = np.array([])\n",
    "observations = []\n",
    "actions = []\n",
    "# flags_num = 2\n",
    "flags_num = 10\n",
    "num_episodes = 500\n",
    "# num_episodes = 1\n",
    "envs = [*batched_env.envs] # , JVMEnv(bm_name=\"h2\", **env_args)\n",
    "\n",
    "print(f\"Getting {num_episodes * len(envs)} samples...\")\n",
    "\n",
    "for env in envs:\n",
    "    # print(env._bm)\n",
    "    env = tf_py_environment.TFPyEnvironment(env)\n",
    "    \n",
    "    for _ in range(num_episodes):\n",
    "        # Get the agent's next recommendation (patience=1)\n",
    "        # based on the input state.\n",
    "        obs, act = test(env, agent.policy, patience=1, verbose=False, num_flags=flags_num)\n",
    "        observations.append(obs)\n",
    "        actions.append(act)\n",
    "        # observations = np.vstack([observations, obs])\n",
    "        # actions = np.vstack([actions, act])\n",
    "    \n",
    "# print(\"Test environments\")\n",
    "# obs, act = test(test_env, agent.policy, patience=1)\n",
    "# observations.append(obs)\n",
    "# actions.append(act)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>observations</th>\n",
       "      <th>actions</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[13.0, 16.0, 0.009589999914169312, 0.994306325...</td>\n",
       "      <td>[2.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[10.0, 20.0, 0.008620000444352627, 0.992601871...</td>\n",
       "      <td>[2.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[13.0, 16.0, 0.009589999914169312, 0.994306325...</td>\n",
       "      <td>[2.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[7.0, 16.0, 0.008170000277459621, 0.9922335147...</td>\n",
       "      <td>[0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[13.0, 12.0, 0.00880999956279993, 0.9933167695...</td>\n",
       "      <td>[0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>[7.0, 4.0, 0.010130000300705433, 0.99206787347...</td>\n",
       "      <td>[2.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>[10.0, 8.0, 0.00839999970048666, 0.99285924434...</td>\n",
       "      <td>[0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>[1.0, 12.0, 0.0072200000286102295, 0.992177724...</td>\n",
       "      <td>[0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>[13.0, 8.0, 0.009399999864399433, 0.9920404553...</td>\n",
       "      <td>[0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>[13.0, 20.0, 0.010130000300705433, 0.992835521...</td>\n",
       "      <td>[2.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                        observations actions\n",
       "0  [13.0, 16.0, 0.009589999914169312, 0.994306325...   [2.0]\n",
       "1  [10.0, 20.0, 0.008620000444352627, 0.992601871...   [2.0]\n",
       "2  [13.0, 16.0, 0.009589999914169312, 0.994306325...   [2.0]\n",
       "3  [7.0, 16.0, 0.008170000277459621, 0.9922335147...   [0.0]\n",
       "4  [13.0, 12.0, 0.00880999956279993, 0.9933167695...   [0.0]\n",
       "5  [7.0, 4.0, 0.010130000300705433, 0.99206787347...   [2.0]\n",
       "6  [10.0, 8.0, 0.00839999970048666, 0.99285924434...   [0.0]\n",
       "7  [1.0, 12.0, 0.0072200000286102295, 0.992177724...   [0.0]\n",
       "8  [13.0, 8.0, 0.009399999864399433, 0.9920404553...   [0.0]\n",
       "9  [13.0, 20.0, 0.010130000300705433, 0.992835521...   [2.0]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = pd.DataFrame(data={\n",
    "    \"observations\": observations,\n",
    "    \"actions\": actions,\n",
    "})\n",
    "# df[[\"MaxTenuringThreshold\", \"ParallelGCThreads\"]] = df[\"observations\"].apply(pd.Series)\n",
    "# df = df.drop(columns=[\"observations\"])\n",
    "display(df[:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we transform the input data X by PCA into Xt. We consider only the first two columns, which contain the most information, and plot it in two dimensional. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.manifold import TSNE\n",
    "from sklearn.cluster import KMeans\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "fig = plt.figure(figsize=(5, 5))\n",
    "X, y = df[\"observations\"].apply(pd.Series), df[\"actions\"]\n",
    "\n",
    "Xt = StandardScaler().fit_transform(X) # normalizing the features\n",
    "\n",
    "Xt = TSNE().fit_transform(X)\n",
    "\n",
    "plot = plt.scatter(Xt[:,0], Xt[:,1], c=y)\n",
    "\n",
    "# plot = plt.scatter(X[0], X[1], c=y)\n",
    "target_names = [\n",
    "    \"decrease_MaxTenuringThreshold\", \n",
    "    \"increase_MaxTenuringThreshold\", \n",
    "    \"decrease_ParallelGCThreads\", \n",
    "    \"increase_ParallelGCThreads\"\n",
    "]\n",
    "# plt.legend(handles=plot.legend_elements()[0], labels=target_names)\n",
    "plt.legend(\n",
    "    handles=plot.legend_elements()[0], labels=target_names,  bbox_to_anchor=(1.1, 1.0))\n",
    "# plt.xlabel(\"MaxTenuringThreshold\")\n",
    "# plt.ylabel(\"ParallelGCThreads\")\n",
    "plt.xlabel(\"Principal Component 1\")\n",
    "plt.ylabel(\"Principal Component 2\")\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "fig = plt.figure(figsize=(5, 5))\n",
    "X, y = df[\"observations\"].apply(pd.Series), df[\"actions\"]\n",
    "\n",
    "X = StandardScaler().fit_transform(X) # normalizing the features\n",
    "\n",
    "# pca = PCA()\n",
    "# Xt = pca.fit_transform(X)\n",
    "\n",
    "# plot = plt.scatter(Xt[:,0], Xt[:,1], c=y)\n",
    "# plot = plt.scatter(X[0], X[1], c=y)\n",
    "plot = plt.scatter(X[:, 0], X[:, 1], c=y)\n",
    "target_names = [\n",
    "    \"decrease_MaxTenuringThreshold\", \n",
    "    \"increase_MaxTenuringThreshold\", \n",
    "    \"decrease_ParallelGCThreads\", \n",
    "    \"increase_ParallelGCThreads\"\n",
    "]\n",
    "\n",
    "plt.legend(\n",
    "    handles=plot.legend_elements()[0], labels=target_names,  bbox_to_anchor=(1.1, 1.0))\n",
    "plt.xlabel(\"MaxTenuringThreshold\")\n",
    "plt.ylabel(\"ParallelGCThreads\")\n",
    "# plt.xlabel(\"Principal Component 1\")\n",
    "# plt.ylabel(\"Principal Component 2\")\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "(slice(None, None, None), 0)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[154], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[43mX\u001b[49m\u001b[43m[\u001b[49m\u001b[43m:\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m0\u001b[39;49m\u001b[43m]\u001b[49m\n",
      "File \u001b[0;32m~/projects/gc-ml/gc-ml-env/lib/python3.8/site-packages/pandas/core/frame.py:3761\u001b[0m, in \u001b[0;36mDataFrame.__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   3759\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mcolumns\u001b[38;5;241m.\u001b[39mnlevels \u001b[38;5;241m>\u001b[39m \u001b[38;5;241m1\u001b[39m:\n\u001b[1;32m   3760\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_getitem_multilevel(key)\n\u001b[0;32m-> 3761\u001b[0m indexer \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcolumns\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mget_loc\u001b[49m\u001b[43m(\u001b[49m\u001b[43mkey\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3762\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m is_integer(indexer):\n\u001b[1;32m   3763\u001b[0m     indexer \u001b[38;5;241m=\u001b[39m [indexer]\n",
      "File \u001b[0;32m~/projects/gc-ml/gc-ml-env/lib/python3.8/site-packages/pandas/core/indexes/range.py:349\u001b[0m, in \u001b[0;36mRangeIndex.get_loc\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m    347\u001b[0m         \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m(key) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01merr\u001b[39;00m\n\u001b[1;32m    348\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(key, Hashable):\n\u001b[0;32m--> 349\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m(key)\n\u001b[1;32m    350\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_check_indexing_error(key)\n\u001b[1;32m    351\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mKeyError\u001b[39;00m(key)\n",
      "\u001b[0;31mKeyError\u001b[0m: (slice(None, None, None), 0)"
     ]
    }
   ],
   "source": [
    "X[:, 0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.011308405\n"
     ]
    }
   ],
   "source": [
    "print(compute_avg_return_episodic(test_env, agent.policy, num_episodes=50))\n",
    "# print(compute_avg_return_episodic(test_env_2, agent.policy, num_episodes=50))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### PCA Visualization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the progress.\n",
    "train_checkpointer.save(global_step)\n",
    "tf_policy_saver.save(policy_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_checkpointer.initialize_or_restore()\n",
    "global_step = tf.compat.v1.train.get_global_step()\n",
    "saved_policy = tf.saved_model.load(policy_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: test (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16}, 'ParallelGCThreads': {'min': 4, 'max': 24}},\n",
      " Env. default state: [list([7, 12]) 0.57],\n",
      " Env. default goal value: 0.57,\n",
      "\n",
      "Successfully initialized a JVM Environment!\n",
      " JDK: jdk-11.0.20.1.jdk/bin,\n",
      " Benchmark: test (dacapo-bench.jar),\n",
      " Number of iterations: 5,\n",
      " Goal: avgGCPause,\n",
      " Number of JVM options: 2,\n",
      " JVM options: {'MaxTenuringThreshold': {'min': 1, 'max': 16}, 'ParallelGCThreads': {'min': 4, 'max': 24}},\n",
      " Env. default state: [list([7, 12]) 0.57],\n",
      " Env. default goal value: 0.57,\n",
      "\n"
     ]
    }
   ],
   "source": [
    "test_env = get_tf_env(\"test\", env_args)\n",
    "test_env_copy = get_tf_env(\"test\", env_args)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:16<00:00,  3.06it/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "-0.122350894"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "compute_avg_return_episodic(test_env, agent.policy, num_episodes=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.02it/s]\n",
      "  0%|          | 3/5000 [00:06<2:29:07,  1.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 0: loss = -0.08275873959064484, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]s] \n",
      "  2%|▏         | 103/5000 [00:20<1:01:50,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = 0.027595680207014084, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.07it/s]/s]  \n",
      "  4%|▍         | 203/5000 [00:33<1:00:14,  1.33it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = 0.01597478985786438, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.03it/s]/s]  \n",
      "  6%|▌         | 303/5000 [00:47<59:38,  1.31it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = 0.07849355041980743, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]/s]\n",
      "  8%|▊         | 403/5000 [01:00<57:59,  1.32it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = 0.07866153120994568, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]/s]\n",
      " 10%|█         | 503/5000 [01:13<56:36,  1.32it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = 0.0788775384426117, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]/s]\n",
      " 12%|█▏        | 603/5000 [01:27<55:18,  1.33it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = 0.03862088546156883, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.03it/s]/s]\n",
      " 14%|█▍        | 703/5000 [01:40<54:29,  1.31it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = 0.020812472328543663, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]/s]\n",
      " 16%|█▌        | 803/5000 [01:54<52:54,  1.32it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = 0.030953237786889076, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]/s]\n",
      " 18%|█▊        | 903/5000 [02:07<51:42,  1.32it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = -0.07765475660562515, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.88it/s]/s]\n",
      " 20%|██        | 1003/5000 [02:21<53:21,  1.25it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = 0.06647924333810806, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.98it/s]t/s]\n",
      " 22%|██▏       | 1103/5000 [02:34<50:21,  1.29it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = 0.07829032093286514, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.08it/s]t/s]\n",
      " 24%|██▍       | 1203/5000 [02:48<47:26,  1.33it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = 0.07940496504306793, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.09it/s]t/s]\n",
      " 26%|██▌       | 1303/5000 [03:01<45:59,  1.34it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = 0.03141622990369797, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.09it/s]t/s]\n",
      " 28%|██▊       | 1403/5000 [03:14<44:50,  1.34it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = 0.055435724556446075, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.10it/s]t/s]\n",
      " 30%|███       | 1503/5000 [03:27<43:22,  1.34it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = 0.06776073575019836, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.10it/s]t/s]\n",
      " 32%|███▏      | 1603/5000 [03:41<42:07,  1.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = 0.0697450116276741, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.94it/s]t/s]\n",
      " 34%|███▍      | 1703/5000 [03:54<43:14,  1.27it/s]  "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = 0.0339801087975502, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.95it/s]t/s]\n",
      " 36%|███▌      | 1803/5000 [04:08<41:34,  1.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = 0.06431885808706284, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.86it/s]t/s]\n",
      " 38%|███▊      | 1903/5000 [04:22<41:33,  1.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = 0.0638260543346405, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.88it/s]t/s]\n",
      " 40%|████      | 2003/5000 [04:36<39:58,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2000: loss = -0.06647807359695435, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.97it/s]t/s]\n",
      " 42%|████▏     | 2103/5000 [04:50<37:27,  1.29it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2100: loss = 0.06989308446645737, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 44%|████▍     | 2203/5000 [05:03<35:25,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2200: loss = 0.07144589722156525, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 46%|████▌     | 2303/5000 [05:16<34:00,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2300: loss = 0.017992911860346794, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 48%|████▊     | 2403/5000 [05:30<32:49,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2400: loss = 0.06617488712072372, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 50%|█████     | 2503/5000 [05:43<31:27,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2500: loss = 0.06551690399646759, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 52%|█████▏    | 2603/5000 [05:57<30:18,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2600: loss = 0.06771502643823624, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.98it/s]t/s]\n",
      " 54%|█████▍    | 2703/5000 [06:10<29:36,  1.29it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2700: loss = 0.01253617275506258, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 56%|█████▌    | 2803/5000 [06:24<27:46,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2800: loss = 0.0662756860256195, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.02it/s]t/s]\n",
      " 58%|█████▊    | 2903/5000 [06:37<26:43,  1.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2900: loss = 0.06570452451705933, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 60%|██████    | 3003/5000 [06:50<25:14,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3000: loss = 0.0676717460155487, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]t/s]\n",
      " 62%|██████▏   | 3103/5000 [07:04<23:52,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3100: loss = 0.015165931545197964, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]t/s]\n",
      " 64%|██████▍   | 3203/5000 [07:17<22:37,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3200: loss = 0.06071428209543228, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 66%|██████▌   | 3303/5000 [07:31<21:27,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3300: loss = 0.05883081629872322, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 68%|██████▊   | 3403/5000 [07:44<20:11,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3400: loss = 0.05168354883790016, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]t/s]\n",
      " 70%|███████   | 3503/5000 [07:57<18:50,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3500: loss = 0.017617255449295044, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.03it/s]t/s]\n",
      " 72%|███████▏  | 3603/5000 [08:11<17:45,  1.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3600: loss = 0.054672833532094955, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 74%|███████▍  | 3703/5000 [08:24<16:22,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3700: loss = 0.07289864867925644, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 76%|███████▌  | 3803/5000 [08:38<15:06,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3800: loss = 0.06126633659005165, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 78%|███████▊  | 3903/5000 [08:51<13:51,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3900: loss = 0.012533322907984257, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]t/s]\n",
      " 80%|████████  | 4003/5000 [09:05<12:33,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4000: loss = 0.0729455053806305, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.03it/s]t/s]\n",
      " 82%|████████▏ | 4103/5000 [09:18<11:23,  1.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4100: loss = 0.060445066541433334, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]t/s]\n",
      " 84%|████████▍ | 4203/5000 [09:31<10:01,  1.33it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4200: loss = -0.08400257676839828, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.04it/s]t/s]\n",
      " 86%|████████▌ | 4303/5000 [09:45<08:49,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4300: loss = 0.0662526786327362, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.06it/s]t/s]\n",
      " 88%|████████▊ | 4403/5000 [09:58<07:31,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4400: loss = 0.0637577548623085, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  2.87it/s]t/s]\n",
      " 90%|█████████ | 4503/5000 [10:12<06:38,  1.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4500: loss = 0.05468904227018356, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.00it/s]t/s]\n",
      " 92%|█████████▏| 4603/5000 [10:26<05:05,  1.30it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4600: loss = 0.010259442031383514, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.05it/s]t/s]\n",
      " 94%|█████████▍| 4703/5000 [10:39<03:45,  1.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4700: loss = 0.008904497139155865, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.03it/s]t/s]\n",
      " 96%|█████████▌| 4803/5000 [10:52<02:29,  1.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4800: loss = -0.07732591032981873, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:06<00:00,  3.03it/s]t/s]\n",
      " 98%|█████████▊| 4903/5000 [11:06<01:13,  1.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4900: loss = 0.06026251241564751, reward = 0.15449999272823334\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5000/5000 [11:12<00:00,  7.43it/s]\n"
     ]
    }
   ],
   "source": [
    "num_steps=5000\n",
    "_loss_test, _, _rewards_test = train(\n",
    "    agent, test_env, test_env, replay_buffer, collect_driver, steps = num_steps, eval_interval=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAnYAAAHWCAYAAAD6oMSKAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8pXeV/AAAACXBIWXMAAA9hAAAPYQGoP6dpAADE2UlEQVR4nOydeZwbdf3/XzM5986ePbbbu0C3pS1XsaXcCFgEOQQVlBv5yqmAIioIAhZE5fL3VU4FQUS/gMohcpS7tAKlLfSg97Vt995Ndjf3zO+PZD4z2U2ymcnMZCZ5Px+PfcBmJzOfpMnMe17v9/v15kRRFEEQBEEQBEHYHr7QCyAIgiAIgiD0gQI7giAIgiCIIoECO4IgCIIgiCKBAjuCIAiCIIgigQI7giAIgiCIIoECO4IgCIIgiCKBAjuCIAiCIIgigQI7giAIgiCIIsFZ6AUUA7FYDJ9++inGjBkDnqdYmSAIgiC0IggC2tvbcdBBB8HppDBFLfSO6cCnn36K+fPnF3oZBEEQBFE0/Pe//8Vhhx1W6GXYDgrsdGDMmDEAEh/CcePGFXg1BEEQBGFf9u7di/nz57NrK6EOCux0QEq/jhs3DhMmTCjwagiCIAjC/lBpkzboXSMIgiAIgigSKLAjCIIgCIIoEiiwIwiCIAiCKBIosCMIgiAIgigSKLAjCIIgCIIoEiiwIwiCIAiCKBIosCMIgiAIgigSKLAjCIIgCIIoEiiwIwiCIAiCKBIosCMIgiAIgigSbBfY9Tz9NDYfdzw2zJmLbed8A8E1a7Ju73/1VWz5ymJsmDMXW089DQPvvJPyd2FwEPt+cTs2HX0MNsydhy2nfBW9f/2rkS+BIAiCIAjCEGwV2PlfeQUdd92NhiuvxJTnn4N3//2x89LLEOvuTrv90MpP0Xb9DfB9/SxMeeF5VJ5wPHZddTVCGzeybdrvuhsD77+P8b/6Faa+/DLqzj8f+26/A4GlS816WQRBEARBELpgq8Cu+09PwHf22fCddSY806dj7G23gvd60ffc82m37/nzk6hctAj1l1wCz7RpaLr2WnhbZ6L36b+wbYKrPkXN6V9DxeHz4Z7QjNpvnAPv/vuPqgQSBEEQBEFYDdsEdmIkgtDatahYuIA9xvE8KhYsQHDVqrTPCa5anbI9AFQesShl+7J5B2Fg6VuItrdDFEUMLl+ByPbtqDziiIxrCYfD8Pv97CcQCOT12giCIAiCIPTANoFdrLcPiMfhqK9PedzRUI9YV1f653R1wVHfkHX7MTf/DJ5p07D56GOw4cA52HXZZRhzy80oP+ywjGtZsmQJampq2E9ra6v2F0bYDmFwEIE330T3Y48h1ttb6OUQBEEQBMNZ6AUUmt4/P4Xg6tWY8L//C1fzeAx99DHaf3E7XE1NqFi4MO1zbrrpJlx33XXs97a2tpIJ7oRIBMLAAIRAAPGBAQgDgxAGEoplxRFHgPd6C7xC/RFFEZHt2zH47rsYeOcdDH30McRoFAAQHxhA07XXFniFBEEQBJHANoGds9YHOByID2uUiHd1w9nQkP45DQ2Id3dl3F4IhdBx332Y8OADqDrmGACAd//9EdqwHt2P/zFjYOfxeODxeNjvfr9f46sqPME1axB46y2IQ0MQhoIQgokfMTgEYXCI/S4MDkIYGIAYiWTcV8NVV6HxqitNXL1xCKEQhj76CAPvvIuBd99FdOfOlL9zZWUQg0HE9uwt0AoJgiAIJU9+uB0PvbMVnQNhzBxXjdtOm4V5Lb60225sD+C3r23EZ239aOsL4uavtuKSRVNGbLevP4S7/r0eb2/sRDASx+T6Ctxz9hzMmeBDNC7g1699gbc3dGJnzxCqvE4smt6AG79yAMZUF07ksE1gx7nd8M6ahcEPl6PqhBMAAKIgYHD5ctSed17a55TNm4vBD5ej7oIL2GODy5ahbN68xPNjMSAaBcenZqQ53gEIgjEvxGK0ff8HiO7Zo/p5fHk5+MpK8FVVEINBRPfsQXjLZgNWaD7Rffuw7fQzEO/rkx90uVBx2KGoOOooVB51NIKfrsTen/4Msd6egq2TIAiCSPDi6j2446X1uOOM2TioxYfHP9iG8x9bgaU3HIOGSs+I7YOROCbWl2PxnHG4/aV1affZPxTFWb9fhgXT6vGni+ajvsKNbV2DqClzJfYRjWNtmx9XHz8dM8dVoz8YxW0vrsOlT3yMF69eZOjrzYZtAjsAqL/wAuz58U3wzp6NsjkHoueJJyEEg/CdeQYAYM+NN8LZNAZN1yfSpHXfOR87zj8f3Y//EZXHHA3/y68guHYtxv7iNgCAo7IS5Ycdho577gHn8SZSsf/9CP3//CfG/PjGgr1OsxBFEdH2dgBA7be/DUddLfiycvBlZeDLy8CVlSUCuLJy8BUVcFRVJoK5igpwDgfbj//f/0bbD65DrLOzUC9FV/pffBHxvj44fD5UnXgiKo8+ChVf+hL4igq2TWT7dgBAvLevMIskCIIgGI++vw3fnN+Ccw5tAQDcefqBWLqhA3/7eBeuOGb6iO3ntvgwN6nm3f3vDWn3+ft3tmC8z4tfnz2XPdZSV87+v9rrwlOXHp7ynF+cNgtf+38foK0viGZfWb4vSxO2CuyqFy9GrKcXnQ8+gHhnFzwzZ2LiIw+z1Gp0z16Ak9W38oMPQvOv70Hnffej89574Z48CS2/exDe/fZj2zT/9jfo+O292PPDHyLe3w/X+PFo/P734fvmN01/fWYjBoNAPA4AaLruB+DLy0d5Rnqk9z/emb6JxW4EXnsdAND4/WtRm+Fz4KyrBQDEe0ixIwiCKCSRmIDP2/pxxTHT2GM8z+GI6Q1YuaNP837fWN+Oo2Y04oqnP8GKrT0YU+3FdxZMwrfmT8z4nEAoBo4Dqr2FC69sFdgBQN23z0Pdt9OnXif9+ckRj1WffDKqTz454/6cjY0Yv+SXuq3PTsQlmxanE1yZ9jsLZ2MjABSFYhdta0Pos88AjkPV8cdn3M5RmwzsqCuWIAjCEAKBQEoN+/D6doneoQjigjgi5dpY6cGWzkHNx9/ZM4SnVuzApYum4IpjpmPN7n7c+q+1cDl4fP2QCSO2D0XjuOvV9Tht7nhUeV2aj5svtrE7IfRHSH5hHFVV4DhO834cDYnAThgagjCo/UtkBfyvJ9S68kMOYQFrOhx1dQCSrzkcNmVtBEEQpURra2uKtdiSJUtMPb4oipg9vho/OvkAzG6uwbmHT8S35k/E0yt2jNg2Ghdw1V9WQhSBO06fbeo6h2M7xY7QD0mx46ur8tqPo7ICXHk5xKEhxLq64FbUotkNKQ1bdeKJWbfjKysBlwuIRhHv6QE/bpwZyyMIgigZ1q1bh+bmZvZ7OrUOAGrL3XDwHLoGUm+yOwfCaEzTOJErTVVezGhKvT5Oa6rEvz9PdUOIxgVc+fRK7O4N4pnLvlRQtQ4gxa6kEZKBnaMyv8AOkOvsMplF24FoeweCK1cCAKpO/HLWbTmOg9PnA0DpWIIgCCOoqqpCdXU1+8kU2LmdPGY312DZZvn6Iwgilm3uxsGTfJqPf8ikWmztGkh5bFvnYEpThBTUbe8exNOXHo7aCrfm4+kFBXYlTNyvj2IHFEedXeCNhFpXNncuXGPHjrq9VGcX66HAjiAIopBcumgKnvloF/7vk93Y3BHAT//xOYYiMZx9SKJL9rpnV+HuV+Xu10hMwNo9/Vi7px/RuIB2fwhr9/Rje5dcTnTJoin4dGcf/t9bm7G9axD/XNWGZ/67E+cvmAwgEdR976mV+KytH/d94yDERREdgRA6AiFEYoWzTKNUbAkTD0g1dtV574spdjbujGVp2JNOyml7qc6OFDuCIIjCcurc8egZjODe1zeiMxDGzPHVeOLi+WisSqh8bX3BlFrydn8IpzzwPvv94Xe34uF3t+LwKXV49vLEjPm5LT489J1D8KtXv8D9b25CS20Zbjm1FacflEgP7+sP4Y31CcuwxQ+8l7KeZy77EhZMSx2BahYU2JUwAil2jFhPD4Y++gjA6GlYCUetDwAQJ5NigiCIgnPBwsm4YOHktH+TgjWJlrpybL/rlFH3efzMMTh+5pi0f8t1H2ZDqdgSRprxSjV2QOCNNwBBgLe1Fe4JI9vY0+GsTSh2MVLsCIIgCItAgV0JQzV2MoH/vAYg9zQsoEjFUo0dQRAEYREosCthdK2xa7SvYhfv68PgihUAck/DAspULAV2BGEEff/4B3ZffQ3iCpNaI/C/+h9sPeNMhLduM/Q4BGEGFNiVMFKNnaPEFbvA0reAWAye/faDZ8qUnJ/nZIod1dgRhBF03v8AAq+/jv4XXjD0OP3/+AfC69ej96k/G3ocgjADCuxKGGZQXKVfjV28pwdicv6sXQi8lkzDjmJKPBxmd9JHih1B6E20owOxvQkjWP+r/zH0WNK5MLD0LYiiaOixCMJoKLArYZhBsQ6BnaOuDuB5QBAQ6+7Oe39mER8YwOAHHwAAqk9SG9hRjZ0RhNavR9cfHkJk9+5CL4UoIKHPP2f/H/z0U0Tb2w07luDvBwDE9u1DaO06w45DEGZAgV0JI48Uy7/GjnM44KhPBjo2qrMbeOttiNEo3FOmwD19uqrnshq7vj6IQuHMKIsBIRhE33PPY9s3voFtZ5yJzvvuQ9f//r7QyyIKSHDNmpTfpQYnI5AayQBgYOmbhh2HIMyAArsSRvBLzRP5K3aAPevslGlYpXllLjiTqVgIAuL9/XovrSQIbdyIfbffgU1HHY29P/0pQqvli3mso6OAKyMKTWjNZwAAz4zEDZf/P8alY6WbXAAIvLnUsOMQhBlQYFeiCOEwxEgEgD6KHWA/LzthaAgD7yXcwtWmYQGAc7lYfWK8t0/PpRU1QiiE/n/+E9vPPQ/bTvsaep9+GkIgAFdLCxqvvw7j7rgdAChYLmFEQUAwmYpt/MF1AIDgypWItusf7IvRKMShIfZ7+IsvqAyAsDUU2JUoUn0dOA58ebku+3Q2SIqdPQK7gXffgxgKwTVhAjwzZ2rah6MuodoVavqEGI2i9+9/x96bb7GFUbIQDGLL4sXYc+OPEVy5EnA4UPXlL6Pl0Ucx7T+vouGyy+CeOhUABXalTGTHDgh+PzivF5VHLkLZ3LmAKCLw+uu6H0up1pUdcggAYGDpW7ofhyDMggK7EoWZE1dVgeP1+RjYLRXL0rAnqU/DSjh9UmBnblAlxuPo/+c/sWXxKdh38y3o+/vf2axbKxPdvRuxPXsBlwuN116D6UuXYsKDD6By0RHsc+ioqQGQqF0kSpPQZ4k0rLe1FZzLhaqTTwYABF59VfdjSSUpfEUFqr58QuI4SykdS9gXCuxKFCGgb30dYK9UrBAOY+DttwEA1SptTpRI0ydiJnnZiYIA/6uvYuupp2HPjT9GdNcu9rfo3j2mrCEfhGT631lXh4bvfQ+uMU0jtnH4fIlt/X7bWecQ+hBM1teVHXggALlUYuiTTxDVufZSMj/ma6pRddxxieN89BEpxoRtocCuRJHHielTXwfYS7Eb/OADCENDcI4bB++cOZr3I3nZGW15IooiAkvfwrYzz0Lb93+AyNat4Gtq0Hjddai//HIAQMyA+iO9keo6Obc74zYOxWfS6IkDhDUJfpZoovHOSQR2rvHjE99TUUzMddYR6VzoqKqGe+LERLNGPI6Bd9/V9TgEYRYU2JUowkDyZFZZqds+7TRWLJDssKv68gma07AA4KwzPhU7uGwZtn/jm9h9xRUIb9gAvqICDVdeielvvI6G714G95TJAIBY+z7D1qAXYiQKIHtgx7lc4JOfSyukY4VwuNBLKCmESAThdesBAGWKm67q5BzngM5mxcOzF5XHHZ84DnXHEjaFArsSxWjFzsru7WIkkhgjhvzSsIBi+oRBzRMD77yDnRdfgtCaNeDKylB/2aWY9sbraLz6KnYhco0dCwCGdAzqjRhJBEnZAjvAGnV2oY0bsfPiS/DFQQcjkEzbF4qhlZ9i58UXY9NRRxd9x2b4iy8gRqNw+HxwTZjAHq9KBnZDH3+s681jvF9KxSY+c1XHJ9Kxg+++y0oHCMJOUGBXohhZYycGgxAGh0bZunAMrlgBIRCAo7EBZQcdlNe+2PQJg+xO+p5PzMisPP54TH/9NTRdf73sn5fEOWYMgIRrvtVhFjujBXbJOrtC1DnFenux97bbsO30MzC4bBkgCAh+usr0dQBA8LPPsPPSy7Dj3HMxuOxDxDo6EPz004KsxSwkY2LvnANT1HT3hGZ4DzwQEARd07HxYedC7+zZcDY2QhgawtCK/+p2HIIwCwrsShRZsdMvsOPLy8FXVAAAYp3WVY8ko9OqE04A53DktS82fcKA5gkhHGY+ew3/czkLnIfjSgZ2wuAg4gMDuq9DT3KpsQNkxU4wMbATIxF0/+lP2HLiSeh75q+AIMCRLC+I9/eZtg4gMVZt1/euwPazz8Hg++8DTqdcz1nknokh1jgxsvZVaqLQc3asMOxcyPE8KpNNFAGaQkHYEArsShS5xk6/wA6QVTsrjxUbeOcdAPmnYYFEdydgTI3d4IcfQhwagnPMGHhnzcq4HV9ezlLqMQPnaepBzoGdpNjlkYr1v/IK2n91D/qeew7BNWsgDA6mX1OyMWXrqaeh4667IQQC8MyciYlPPoGGy76bWEceAaYYi2H393+A7eeehz0/+Sm6Hn4E/tdeQ2jjxhH1e+FNm7D7mmux7YwzMfDWWwDPo+b00zHt36+gKhnUWKHu0EiCSauTsmTjhBLJ9mTov//VbSa11KDjqK6Rj5NMxw4sfUvXspLQhg3Yefnl6H/5ZUuXqxD2xlnoBRCFwQjFDkjU2UV27LB0Z2y8L3GRdk+Zkve+mN2JAYHdQLJ4u+r440b1GnSNaULY70esvR2eadN0X4teCDkHdskaO40BlRAKoe3GHwPRaMrjrgkT4Jk+HZ4ZM+DZbwac9fXofvSxRMoVgKOhAU3fvxY1Z5wBzuFg6W2hX3t3bmj9eua/Fly5MvWPHAfX+PFwT54MzuVK3HSIIsBxqD7lFDRccQU8UxOfU0eND0BxGzfHAwFEtm4FgETadRjuCRPgnTULobVrEXjjTdR+45y8j8nKUhTnwvLDDwdXXo5YeztCn69F2YGz8z4OAPS/+CIG33kXg++8C/9LL2Psz29hNbIEoRcU2JUocl2Jfs0TAFjqyqqdsaIgsIv9aMFFLkjpMTEYhBAMgi8ry3ufQMKAWDJJrTz++FG3d44Zi/CmzYjus7hiF04Gdh5P1u34PJsn4r29iX9nhwMVh89HaNMmxDu7EN29G9Hdu5mHoQTncqHuwgtRf/l3UzrFWRNHHsGU9Bqc48fBd9ZZiGzfgcj27Yhs2wZhYADRtjZE29rY9lUnnYTGq66EZ8aMlP3ooWJanVByjJhrwgSmhg+n6uSTEoHdf17VJbCTzdrlcyHv8aBy0SIEXnsNgaVv6hbYiSFZoR146y1s/egjNN1wA3znnK2bUTxBUGBXogyvK9ELq3vZiQoFR4/Ajq+oAOdyQYxGEe/t1S2wC65eg3h3N/iqKlQcdtio2zuTRr9WtzyRU7GurNs5WRCjLaCSDKOddXWY+Pjjicd6exHZvBmhTZsQ3rQJkU2bEdm5E2WHHIym66+HW9GBKaFPYJdUiCdNQuOVV7LHRVFEvKcnEeRt345YZycqjz4a3gzj7UohsGPGxGnSsBLVJ52Ezt/8FoMr/otYb++IZiK1xNModkBCKQ+89hoG3lyKpmuvzesYEtLnv/q0UxHdsRPB1aux79Zb4X/5ZYy7/RdwT56sy3GI0oYCuxJleCeYXlh9XqyosC/QI7DjOA6O2lrEOjoQ6+2Fa/z4vPcJAIE3E11/lUcfndM6XWMkyxPtit3AO+9AGBqCa8IEuJqb4aitzcvjLx251tjlr9glnudQXPSdtbVwHnYYynMIlEesQwfFTkqlSnAcB2d9PZz19ShPzijNBktPF3NgJxkTp2mckHBPnAhP60yE161H4I03UHv22XkdU0qzO4ZZP1UefTTgcCC8cSMiu3enDfzVIt1YevfbD+OXLEHv039Bx733Yuijj7D1a6ej8ZqrUXfBBeCcdGkmtEOfnhJFCCS6J3m9AzurK3bKwE6nk6ejrg6xjg7dpk+ICnf9qhNGT8MCSssTbYFdcNUq7Lr8f1Ie48vL4WpuZoGea0IzPNOmoeKIIzSnjcyyO4knfQUdGdJ5ucLGmwUCEGMxTZ8Z6TVI6p/mtZRAjV0oB8UOAKpPOhmd69Yj8Op/8g7s4oH0np4Onw/lhxyCof/+FwNLl6Lu/PPzOg6QemPDORyoO/87qDzuWOy75RYMLvsQHff8Gv5X/o1xd94B7wEH5H08ojShpH6JIp3Mht+l5ovV58WywM7l0q2mhVme9OkT2EW2bEF0x05wLhcqFh2Z03NcYxOBXbRDW2AX+mIjgMTFzdmUSOsKQ0MIb9qEgbfeQu9TT6Hjrrux67LvoveZZzQdA1AYFLtyNCjWHNgl/i2kfxutKBVt6Tujei1SYOfLcy1FnoqNtrcj1tEBOBwZ09ESku3J4PLleTUuiaIIwZ85eyF1x+o1hUJS7DiXXIrgnjABLY89hnF33gm+uhqhtWux/ZxvILx5sy7HJEoPCuxKEDEahTiUMBDWPRXbZHHFLnli5V3Za7zU4JRMinXysgu8kfDOKl+4AI7KitzWkOys06rYRZPTDGpOPRUz3n0H+69ehamvvIKWRx7G2J/fgvpLL4F7eqLbNrpnj6ZjAIqu2FGaJ/INYliNXW1+ih3ndOY93kxOxeap2CWDVGFgIKVWtFiQjIk9M2aALy/Puq178mR4DjggMdP1Te1ec2IoJJ8Tqkf++0h+dkMff6xLQJ2pFIHjOPjOOhNTX3oRZQcdBDESQc8TT+Z9PKI0ocCuBFGa2PI6zooFFD52vb2WvPjkWuOlBnmsmD6KXSB5oarKoRtWQlLZ4j09msYgSV2Z0ggn3uOBZ+oUVB55JGq/9S003XADqo4/AYA871ULan3stAYx6WrstJKvWbJkbpx3YKdQ14sxHSunYTPX1ympPjkxYsz/n9c0H1PqiAXPg68YGUy6W1oS3cnxOAbefVfzcSTSKXZKXE1NaLruBwAS1ijFqs4SxkKBXQkiSDUl5eW6F+k6amsBhwMQRaaaWAlDAru65EQAHWrsovv2IfTZZwDHoerYY3Nfg8/HVLBYh/qpH5G2hGLnas7c/CF1sorDTHXVIAWFowZ2yhSoX72HnKSeSv82+cDSwhrWAShSsXmmhTmHg9WBFWNgJyl2o9XXSUizYwc//FBzAKQcrZipUahSx3RsLuefskMPhWe//SCGQmykIEGogQK7EkQ2J9a3vg5IjONx1tcDsGZnrJGKnVSwnw+Sd13ZvHmsESUXOI6TGyg0dMZG2xLp1Wydf3wycBTzGIyeq90J53TmFcRINXb5WmEA+Zsl65WKTazFl7LPYkGMx5mHXbaOWCWeKVPg2W8/IBbTHHRJwTqf5d9GUs4H33tPkxquRIiOfv7hOA613z4PAND7l79AjMfzOiZRetgusOt5+mlsPu54bJgzF9vO+Qa7y8uE/9VXseUri7FhzlxsPfU0Nk5KSXjLFuz63hX44tDDsOGgg7Ht62fnVUdkdQSDrE4k5M5Y682LZTVeetbY6Th9YiBZX5drN6wSaWZsdJ86LzshGGQj4FzNzRm3ky5GUgOEFiS1jx+lxg5QKGUagpgYa57IP7CTrVc0TsHo06crFijewC6ybRuEwUFw5eXwTM99ckqVlI59Tdvs2HiWxgkJ76xZcDY1QRgawtCKFZqOIzFaKlai5tRTwVdXJ8y038k/BUyUFrYK7PyvvIKOu+5Gw5VXYsrzz8G7//7YeellGWcGDq38FG3X3wDf18/ClBeeR+UJx2PXVVcjtHEj2yaycyd2nHse3FOnYNKTT2DqP/+Bhiu+N2pxt50xUrEDrN0Zm2sqUA0Onz6p2Ljfj8H//heAuvo6CVmxUxdQSzcxfFVV1uCDcye+E0JYD8Vu9PffkYdJMeuKzdPuBMivQ1cUBDl4yLMrNrGP/L3swlu2pNTZWgFmTNzaCs7hyPl50rznoWUfalK2WFlKFqN2judReVyiLCKQR6MGoPj8j9IVzpeVwXfWWQCA3qefzuuYROlhq8Cu+09PwHf22fCddSY806dj7G23gvd60ffc82m37/nzk6hctAj1l1wCz7RpaLr2WnhbZ6L36b+wbTrvuw8VRx+FMT/8IbytrXBPnIiq445j6cRiRBhIWp3o3DghwcaKWbAzVswhFaIWVmOXp2I38M67QCwG9/RpmhzoJcsTtdMnpI7YbGodoFTs8gjsVLz/WgMqURDk9KdPh1RstfbATvD7E7NfoY+1UL6K3dDHH2PrKV/F5uOOR+eDv7NMrR4zJs6xcULCPWkSgIQSpqkWk5kTZ1dTpRutgaVvJcYSaoQpdqOUIgBA7bnfAjgOgx98gPDWbZqPaSTxvj4MrvgvQhs2FHophALbBHZiJILQ2rWoWLiAPcbxPCoWLEBw1aq0zwmuWp2yPQBUHrGIbS8KAgbefgfuyZOx85JLsXHhEdh2zjeYOWwmwuEw/H4/+wlo9LcqFIYrdslUbLxEFDupjive15dXPYzcDXuCtnU0SalYdTV2EdYRO0pg50kGdnk0Twg5KhaA9iAm3t8PJC++zjwbFgBlgKluHWwtSDYq6TGbWDIp1pgWDm/aBCARcHb9v/+Hzccdj45779Oto1sruRoTD4dzuZjJutQJrQYhwzix4ZQffjjgciUmzOzdq/o4Euz8k8Pn393SgspjjgGQqLWzIsHPPsfOCy7Anpt+UuilEApsE9jFevuAeByOYUqao6E+Y8ov1tUFR31Dxu3j3d0QhobQ/cijqDhyESY+9iiqTjgBu6++hqXE0rFkyRLU1NSwn9bW1vxenMkYXmPXYGHFLsfifTWwFJsoau6cFMJhDCbtFLTU1wGAc6y25ono7kRg5x5FsdOleSKsRbHrU3UM6QLPV1bqE0zl0TyhlznxiLVo9ffrTjT4eOfOgWf//SEMDqL7oYew+fgT0P6rewpSPiGEwwh98QUAoOxAdYEdoGhe0mAQzm5yq7Lf5PJuN/PWE/LpCleh2AFgTRT9L7yA+MCg5uMaBTMc1/F8SuSPbQI7IxCFRIqk6rjjUH/hhfDOnImG716GymOOQd9fn834vJtuugn9/f3sZ926dWYtWRdkxc7o5gkrKnb6p2I5l0vu4NRo8TK0fDmEoSE4x4yBd9YsTftwjdU2L5Z52DVnn4UpvWdCPs0TzKDYQMVOp3FibB3Mx05Dqi+5dt6Xf+MEkH8qVvp8VixciCkvPI8Jv3sQ3tZWiEND6Hn8cWw+/gTs++Uv85o5rJbw+vVALAZHfT2cGmYts8kvGlTHuD8ZeOdwLpQaHnTpCs9BsQOAigUL4J4yBcLgIPr/8Q/NxzUKNiIwx9dDmINtAjtnrQ9wOBAf1igR7+pmCtGI5zQ0IN7dlXF7Z60PcDpHdGF5pk1FNIvc7vF4UF1dzX6qDFK+jEIqGDZOsbPu9AlW46VjVyygSMdqTGlJ0yaqjj9O86gz1jzR2akqJcxq7EZLxbr1szsZbVYsoL3GTq9xYvmuQ/kcPTpiAR0nctTVg+N5VJ1wAiY/939oeegPKJs7F2I4jN4n/4wtXz4Rg8vz6wDNFdY4ceCBGb3ksiFNF9HimymoKEthPo55GK+rzRhwPI/a85LWJ08/DTFZr2kVxBwnyRDmYpvAjnO74Z01C4MfLmePiYKAweXLUTZvXtrnlM2bm7I9AAwuW8a259xulM2ejfC21MLU8PbtcGm4c7QLbOi1UYGdNFasq8uyJ6JcAgs1SOqQlouLGI8z/7pKDd2wEs6GhoQ5dCyWsVM8HWoVO9G0rlhtaUe9xolJ8PkEdn06p2KlGjutnnrJz4WzXn5vOI5D5dFHY9Jfn8HExx9LBHiRCNrvvMMUD7XgZ4nAzquyvk5C9pHsU/1cNjN7lFQsIKtS+TUP5V5jJ1Fz+ungKyoQ2bYNg8uWaT62EQgGZECI/LFNYAcA9RdegL6//x19L/wD4S1bsO/W2yAEg/CdeQYAYM+NN6LjN79l29d953wMvP8+uh//I8Jbt6Lzwd8huHYtas87V97mkovh//er6P3b3xDZsQM9Tz2NgbfeTnQkFSls6LXBdidiOMzUQaug5cSaC/lcXIKr1yDe3Q2+qgoVhx2meQ2cwyHXN+ZoeRIfGGSB06hdsR4dLmxa7E5UK3Z9iefr4GEHpAZTam9U9DQnBnRQ7LKkqTmOQ8XChWh5+CHwNTUIb9qM/n/+S+tScyYkTZzI0Zh4OI481HL5XJhDKjbPrnAxHgeSgbKamjRHZQVqTj8dAFIcHayAEaUtRP7oO0/KYKoXL0aspxedDz6AeGcXPDNnYuIjD7OLWXTPXoCTY9Xygw9C86/vQed996Pz3nvhnjwJLb97EN799pP3+eUvQ7j15+h6+GG03/lLuKdMwYQH7kf5IYeY/vrMwmjFjvd6wVdVQQgEEg0sBgWQWjDqRCRbnqhX7AJvJrqwK48+Ou91OceOQay9PWF5cuDsUbeX1DqHzwdHZUXWbXm3jl2xKgyKBZUdoHqOE0usI/n5jcchDA6qsgmSU7E+fdaiCOxEUVSduox3j15/6KipQcN3v4uOe+5B54MPovqUxTkZSmsh3teHyI4dAICyHD6v6civxi45eSKnVGyegZ0ihav2xrL2vPPQ+/TTGHjrLUR27846IcZMjHAZyIcnP9yOh97Zis6BMGaOq8Ztp83CvBZf2m03tgfw29c24rO2frT1BXHzV1txyaIpI7bb1x/CXf9ej7c3diIYiWNyfQXuOXsO5kxI7FcURdz7+kY889Eu+INRHDq5FnecfiCmNGQ/nxqJrQI7AKj79nmoS3YKDWfSn58c8Vj1ySej+uSTs+7Td9ZZzAyyFDC6xg5IqHaRQACxjk54pk417DhqMSp1oLXGThRFZq+jtRtWiatpDELIvYEi2pabhx0gB2P5jFXSZlDcp+oYUnekHuPEgMSNCufxQAyHEe/r1xbY6dYVm9iPGIlADAbBlY8cXJ8JMR5n7+VoPp213z4PPU89hdjeveh9+i+ov/girUvOSvDztQAA16SJmt8jefKL+psqlorNJbBL1uVq/fwrA0JeZRepZ+oUVBxxBAY/+AC9f3kGY370Q01r0BvpJs8KXbEvrt6DO15ajzvOmI2DWnx4/INtOP+xFVh6wzFoqBx5YxKMxDGxvhyL54zD7S+lb4DsH4rirN8vw4Jp9fjTRfNRX+HGtq5B1JTJr/cP72zFH5dtx2/OnouWunL85rWNOP/xFXj9B0fD68rdbFtPbJWKJfTBaMUOUHTGWszLzjDFjhVwqwvsIlu2ILpjJziXCxWLjsx7Hc5kZ2wsRy87yerElYMCwOk5KzYXHztJsRsaUnVMeZyYPjV2yrWotl7RORXLV5QDyQBDfcDblzBL5rhRgyje40HjVVcBALoeekizjc9oBNesBqA9DQtoL4MQBUGePJHDuZCdMzQ2T6Q0XWho3pKaKPqeew5CMKhpDXpjpVTso+9vwzfnt+CcQ1swY0wV7jz9QJS5Hfjbx7vSbj+3xYefLJ6J0+aOh9uRPhT6/TtbMN7nxa/Pnot5LT601JXjqP0aMak+ocaJoojHP9iGq4+bjhNnjcXMcdX47Tfmot0fxmvrzOssHw4FdiWGKAgQkuOEjEyRWtXLzqjUgdY6H2l4efnCBaOmQnPBNaYJABDNcfqEKsVOes9iMU1F9WIspqrGiK+uBpKpRjV1dtJoN71q7ACl5YnKtDBT7PQJ7DiO09ylKzXUOHy+nMZ21Zz+NXhmTIfQ34/uRx5Vv9gc0GpMrISN9FP53RMGB5mRdS6Bt2z3k59ix7lcmrp/K48+Cq4JEyD096P/pZc0rUFvJJcB3l3YrthITMDnbf04YrrskMHzHI6Y3oCVO/o07/eN9e04sNmHK57+BIfc/joW3/8envnvTvb3XT1BdAbCKcet9rowr8WHlTsKZ/pNgV2JIQwMsBFH5ih2VgvsDLI7kWrsVHbFytMm8k/DAoBzTFKxy7F5gk2daB69C1zZSaxFtUtJReVQs8XxPLv5UBXY6Wx3AigUO5XKlaTw6ZWKTexLW7cwqz2sz03J5BwONP7gOgBAz5NP6u5tJ4qi3BGrwZhYQmuNndQ4wbnduX0edaqx03ru4RwO1J6baPzrffovlnAcMFqxCwQCKVOewhnqe3uHIogL4oiUa2OlB50D2muCd/YM4akVOzC5vgJPXDwf3/7SJNz6r7X4v08SN8SdAyF2HD2Pmy8U2JUYkjkx5/HobvmhxGnRebGy87sxip2a0Uyx7u5ERyDHoerYY3VZhzOp2MX25ajYSVMncknFKgM7DQ0Uygtiru+/ljo76d/AqZNBMSAbDKsd5cXsTnRKxQJ51B5qsIGpPPYYlB1yCMRwGF2/+52q441GbM+ehP2K0wnvzJma9yPVUgoDA6qCLlaSkmPmQg7sNKZidQiCfGedCc7rRXjDBgQ/+UTzfvTCaLuT1tbWlClPS5YsMeQ4mRBFEbPHV+NHJx+A2c01OPfwifjW/Il4esUOU9ehFgrsSgxhwNipExJWnRdrxEgxQO4yVKMahDduBAC4Jraw9ytf2PSJjo6c7uhlD7vRU7FwOoGkebKWdBR7Ds+Dc+bWt8WrHOclhEIQh4YAGJOKVaMcivG4bKdhgcBOGic2fCxjNjiOQ9P11wMA+p57HuEtW1QdMxtMrdtvP/Ber+b98NXVCf9GJEdP5ki8X53tEzMozjcVm0cQ5KipQc2ppwIAep56WvN+9MJoxW7dunUpU55uuummtNvVlrvh4Dl0DVPJOgfCI9Q0NTRVeTGjKfVaOa2pEnv6EjWOjZVedhw9j5svFNiVGFIqKRdDznxwWLbGzqDmiWSdjxgKQUgGFqMR3rIVAOCZOm2ULXNHmj4hBoMsqMhEvL+fFY/nVGPHcXmlo7TUN7KAKscghgXWTqeupQaOavWBnRAIsLIHXQM7zRM5pKkT6gLe8oMPShhnCwI677tP1XOzEdqwAQDgna3N5kSC43lFsJv7jZXamdn5jhTLNxUr4fvmNwAAA2+9VfB0rJrZz1qoqqpKmfLkyZAydzt5zG6uwbLNspAgCCKWbe7GwZN8mo9/yKRabO0aSHlsW+cgmn1lAICWujI0VnmwbLNsCB8IRbFqVx8OnqTfjaVaKLArMcywOgGsOy/WqMkTfEU5O7nlqtpFtibUD880/exgeI+HXeRGq4liHnb19eDLynLaP+uM1ZSKlawRVAR27IKdWxCjrK/TUqCecR0aumLZnNiKCl1rOtl7orILlCl2dbkrdhJNP/g+wPMIvP4Ghj79VPXz064n2bmdk1o8Clrq7OIqxokBilSs1q5YnW4qPdMSN4JiOMwa4QqFURkQLVy6aAqe+WgX/u+T3djcEcBP//E5hiIxnH1ICwDgumdX4e5XN7DtIzEBa/f0Y+2efkTjAtr9Iazd04/tXYNsm0sWTcGnO/vw/97ajO1dg/jnqjY889+dOH/BZACJm92Lj5iCB5duwuvr2rFhnx/X/W01xlR7cGLrGFNfvxLb+dgR+cFOZiYFdvG+PoiRiCXa4QHjFDuO4+CorU2YA/f05nSxkhQ7t46KHZCwPIn39SHW3g4ozLiHE8lxRqySfNJR8lxJNYqdD0Du6pRkN6PXODG2DpUpYeW2eqp1ibX4EvtXXWM3cpxYrnimT0fNmWeg//+eQ8dvfoNJf/5z3oFzrCPR4ONsasprPwDg9NUiArWBXfLfJ8fAji9w8wRbh9cLvqICwuAg4t3dht+kZ4PdKFtgVuypc8ejZzCCe1/fiM5AGDPHV+OJi+ejsSqxtra+YMpntt0fwikPvM9+f/jdrXj43a04fEodnr18AYCEJcpD3zkEv3r1C9z/5ia01JbhllNbcfpB8jnzf46eimAkhpue/wz+UBSHTa7FExfNL5iHHUCBXcnBFDuDa+wcNTUJr6ZoFLHubrjGjTP0eLliVFcskKizi7W355wOChug2AGJBorwhg2JwC4L0bY9AAC3CsWEd3sQR37NE7wK1321HaCyYqdvGkTLFAy9zYnZWrTW2DEbGG1Bb+NVV8H/4ksIfvwJBt5+O++Gn2hHUrEbk39gp2VWs+BXV2+cb1esno0Gjvp6CIODiPX0wD15ct7704qVfOwA4IKFk3HBwslp/yYFaxItdeXYftcpo+7z+JljcPzMzOobx3G47sT9cd2J+6taq5FQKrbEiCfrSniDa+w4nmfu9lYyKTaqKxYAnFI6KIeLS9zvRzyZpnbrPJnDlbQ8iY5iUhyVFLvm3McT5TN9QstFQK1iF+/Vd5yYvA71difMnFgnD7sRa1FbY9etXbEDEo05ded/BwDQ+dvfavIyVBLrSNTfSnWh+aDFpJhNncjxXCiZakvebWrRS7EDFNM2urtH2dJYrBbYEQkosCsxpLtUoxU7wJomxULUuBMRmz6RQzpI6i50NjXpnkphlic51tipqXFiqkXYpMBOpTrFrE50Vux4DcGUVBfIWyQVG+tR3xU7nPpLLwVfXY3wps3o/+e/NO9HUDT36JGK1VJjxzqWc1bs9Bkppoti15D4N5Rm/xYKI8+nhHYosCsx5HFixip2gDUbKNSMtFILUw1yGCsW2Zqsr9M5DQsoLE9GmT7Bpk6oGCgu1cdpUS0Eaa6kinoctepU3IBxYol1+FStQ7mtFVKxYiQiBzJ5BL2Omho0XP5dAEDngw9qTktKNx1ceTl4FbN3M6FlVrOkvubcPCGdM/IcKaaPYpfMhnQX9tzKumINOJ8S2qHArsSQu2LzP5mOhhUVO6NGigFy+i+Xi4sRVicSzqZEaivbvFhRFBFJ1tjlMnVCQqqP01Zjp8HuRG2NnQHjxADAUZO4+IuhEIRQKLe16Dwnlq1FCuz8fojJkVijwfzdHI6811N73nngq6sR27sXoY2bNO0jmmyccDU26tK9zG6qVNidSGUpufvY6TRSTJcau6RvZoEVO0rFWhMK7EqMgih2FhorZuSJyMmmT4x+so0kU7HGKHbJwC5LKjbe28uMfFWlYiW7k7xq7HJXLFgQo1qx86la22jwlZXMBFcyth11LawrVt+1MAVQEEb1KmRrUdQecnx+p33e64V74kQAQGzfXk37kEbe6VFfByjKIHJQyyUElQ4B+U+e0O+mkil2KkcY6o3cFUuBnZWgwK7EUFtXkg/yWDErpmIN6IqtlaZP9I26bXirgYpdMhUb7+/PqC5J9XXOpiZVnn5MtcijK1aLQbEYDOZ0zBgz4dU3FctxnGJubV9OzzEqFcu73eDKy1OOMRpSkb1eNjCucbk16GRcT4fegZ32VKxkPj0ahZ4Vq8TJFDtqniBGQoFdiSErdmYEdlZW7IywO5Fq7LLfRQuhEOtI1dvqBEioS9KFP5NqxzpiVdTXAUqDYi2KXSIwUxNI8lVVslKWg9WIFFTrnYoFFJYnuaqHBqViAS0paqlxQp/Azjk2YV+kWbFLWp04m/QZpedUNE/kOo1Ba/NE/iPF8j/3OCym2FFgZy0osCsxzJo8AVi0xi5516z35Akg9wLuyPbtgCiCr65mo9f0hOM4uJJKSCZFRUtHLCCnXPJLxebePKFGKRMFQQ6mdG6eADQ0cjDFzojAzpc4hsrATjfFTmrQ2Zu9QScT0WQq1qWzYidGIqzEIBtiLMZG/+XePGGNkWIA4GRdsaTYESOhwK6EEEWxIDV28c6ugs80BBKv38gTkWSSGu/vz+rxJVmdeKZN03XslRIpxSUpI8PRMnUCUKaj1KditRq05hrECH4/kHzf9a6xAwCeTZ9QWWOncyoWAJxqbWC687c6USKnYrUFdnpOnQAArqyMqcmxHEohpPMgoGJWrEVGigHyv2O8r0/zevRAT9NlQj8osCshxKEh+cJnQo2dpEaJ0WjO6StDicXYUHZDAjsp5SaKWS+4kS3GWZ1IyIpd+guvlqkTgKy2mWVQDOSulEmF83xlpSGKrBrFTozH5VSfAalYtb56sTzGiaWDpWL3am2ekFKx+ih2HMfJN1Y5NC9J/zZ8eTk4Z24DmKwyUgxIfqaSTTC5+GYagSgIiXMqKLCzGhTYlRDsLtXlAuf1Gn483uNhFyArTJ9QnpCNOBFxTqd88c9ysjWycUKCKXbJlNdwNNfY5WVQrK0rMFfFTrK6MKK+DpCL7HMJpuJ+P7uJyNVOQ9VaVKdi9fX3Y4pdR4fqCRSiKDLFTo9xYhJqTIrZzGwV/zZ5N0/oqG5xPC8HsgWqs0s9nxZ+ViwhQ4FdCcG6wCorDUsBDsdKdXZKlcmIrlhADiqyFTVHthgzI1aJk1mejFTsRFFEdI/kYadSscunxi7Z1arWGiHXpgVmdaLzOLHh68ilK1ZaK19ZaUwHttrALs9xYsNxNjYmFKNYDLEudXVeyvShVK6hy5p8uXfGxv3JNLmZgZ2Oih1Q+LFiyveBN6AZjdAOBXYlhDAwACD3odd6IHfGWkGxS9ai8HzO6Re1yOmgvvRriMUSzRMA3NOMU+xYKjaNYhfr7EwEWTzPiuBzRVbszLE7ARQdoKOmYpMNAj6jA7scFDsDO2IBDTV2UldsnT41dpzTKX+3VXbGSmlYR12drso5u6nKIbCTmsjUnAulgEzQOitW8nzT6TWzBgoLKHYw6EaZ0AYFdiUEU+xMaJyQYIpdR+EVO3bHbGA9iOynlf5kG929G2I0Cs7rhWt87hMf1OIckwjYYmlq7JiH3dgxqtUDnhkUawjsNM6VzDkVK1md6OxhJ68jd7sT2ZzYmMBOfk9y7NCVgl4d1UytnbF6N05IyKnJXBQ79efC/A2K9W00YJYnKhVTvVC+HrMyQERuUGBXQjCrk5JV7Izv4HKOMlZMqq9zT5mS9wSAbLDpE11dEJMFzhLR3YnAzt2srr4OkGdC5tU8oXKuJGsUGCWIYV5tRtXY5bgOwNiOWED5nvSNuq0QDkMYHEysR6euWABwjks0UERVKnZRqXFCx/o6QF2NnRaj9ry7YvVOxUomxT2FD+wIa0GBXQnBhl5XmhjYWajGTqtipAaHT6qxyxDYSfV1U42rrwOS6oXLBQjCiKBaq4cdkJ9BsSANDPeoK7TONe1o1DgxCRbY5TDGi6ViDfCwA9SlYqWAl3O5EqPRdEJS7GIaFTuXTh2xEk4V82Ll5onc/32s1DwBKBS7As2LZd9nCuwsBwV2JYQQKECNXZMVFTvj6kHkGrv0FxczrE6ARNeci9VApV54o23aOmKBPJsnNL7/uVp7GDVOTO06AFnV4w1PxfaNui3zsKur0zVlptXLjs2J1TsVq6LGLh6QUrHqFTvEYgmrD5UYptgVuHmCAjvrQYFdCSGfzApQY2cFxc7AObESLB2UoaDZDKsTCWeGBop8FDu+IM0TPgAqauwMTsUKgcCI9PaItRicipX2KwwNjRpkS6k6vcaJSchjxdQGdgalYllXbN+o2wp+Lc0T8uc2vxsbnRW7QjVPSBkQlV3uhPFQYFdCSCezUq+xM8K8VoJZEKRJB4miaIrVCVtLBsuTyG4psFPfvMFSsXlc2NS+/44aH4DRlTLDa+wU1hjKyQVp12JwVyxfVQUk1becu4V1HrOmVbGLdhrVPJHbrGZA0TyhKhUr3xBq+vzrrdgVeKyYGedTQhsU2JUQbJxYAWrshP5+CBpUHj2RT6wmdMWmqbGLdXQkitgdDrgnTTJsDRJSDZNUrA4kJiJEk9MC3FpSsW4dmic0KnZiOAwhFMq4nZT+dhoU2HFOJ6tRG1U9ZF2xPmPW4nDIM3RHWwsbJ6a3YpessevsHFXBVBLTeU4sW4/03RtlpB+gsXlCEZBpaaAQdK7xlWvsugsyslFrMxRhPBTYlRCF6Irla2rYCTFeYNXOjJoQydk/3ts74mQrqXXulhZT6lLYhXefHNjFOjqAaBRwOlmqVg2Sw7wWxUJIWqSobZ7gK8qBpO9gpiBGCIfZUHej7E4AFWbJLBVrjGKX2LcvcaxRAjs2TkwnDzsJZ0OD3KDTkX7CyXDEaFQ2S9Y5sGNpb0EYtcFFy8xsjuPYuUxTA4XuBsWJQFYMhyEMDumyTzVQjZ11ocCuhNByMssXjuPgaLRGnZ0pdifJGjsxHE7M5lUQZo0TxtfXAfK4pphCsWP1dePGgXM4VO9TSkflVWOn8g6f47hRzYFZs4rTmUhTGkSuJsVyKtZn3FqSgUxsNMVOGiemc8DL8TxcyXRqrulYdg5wuXSvP+RcLvZvP1qdHUvF1qg7F+bTGSvofP7hy8vBlZcDKIzliZSBocDOelBgV0JoST/ogbPBGnV2ep9Y08GVlzNFKjbs4hLeao7ViYSk2ClTsRFpRqyGxglAYVCsKbDTbhDN1KkMF2yl1YmRZqm5TsEwU7EbTT1kip3OqVgAcEp1dntz87KTPouuxkZDfBxZnV0Gg3AJdi5UeROQT2DHPv86pi4LOVYsn+8zYSwU2JUQhVDsAEUDhVUUOwO7YjmOU1iepF5czLI6kXA2Sc0T7SwtzBS7CdoCOz2aJzQFdqMoZUaPE5OQLU8yp/rEeFwOHAzqik3sOzeTYqMUOwBwqeyMlSbQ6N04IZHLvFghHJYL/1XMigUUY8XyaZ7Q0W7JWV+4sWLs+6yytIIwHgrsSghWY1eln0lpLsiWJ4WusTPnDjOT5QmzOjErFZv0EBQjEXbxl6dOaAzs8mmeSKp8vAZ7hNHqyYweJ8bWUT26Yqes71IzZF71WnK1gZFq2owI7MapGysmW53oW18nkYuXHfu343nwFRWq9p+fYqd/s4E0SaQQY8XM8AUltGG7wK7n6aex+bjjsWHOXGw75xsIrlmTdXv/q69iy1cWY8Ocudh66mkYeOedjNvu/fmtWH/ATPQ88YTeyy44+dyl5otlFDsTZsUCsq2E8uIS7+9nzSPuKeYodpzbDYcUVCcVFVmxU98Rm9inNRU7o61Ocl0HIAdafGUluGTThyFrybHGTlIz9RwnJsHS/TmOFYtJVic6e9hJyLOa+zJuI7DMRZXqdHA+Y8WMUewKN1aM7E6si60CO/8rr6DjrrvRcOWVmPL8c/Duvz92XnpZxvqCoZWfou36G+D7+lmY8sLzqDzheOy66mqENm4cue/XX0dw9WrDUgSFRkoNgefBJwtuzYIpdpbpijX2DjOd5YnUOOEcOxaOSnUqQT6w4vakUhLNs8aOvXfxuCqLCyDPwE5Sp/r70v5dGiNl1Dgxtg4W2KVfByDXvBmZhlXuP5tiJwwNQUxaxBii2KkcK8Zq7Aw6z8qBXRbFTmN9HaBU7DQEdkYodgUcK2bGiEZCG7YK7Lr/9AR8Z58N31lnwjN9Osbedit4rxd9zz2fdvuePz+JykWLUH/JJfBMm4ama6+Ft3Umep/+S8p20fZ2tN9xJ5rv+ZWhd9iFJJ7HXWq+sLFihVbsTGrPlwu45YtLxOTGCQml5YkYi8kX1mZtih2vqKdRo9qJopiXYuoYZei9USa8I9aRQ/NEzGBzYrYWyXqlL8tapDmxXq8hN3SyYpdrKlZS7IxJxTpzMCmWm8jUZy60pmLFeBxIeusVi2LHumLJx85y2CawEyMRhNauRcXCBewxjudRsWABgqtWpX1OcNXqlO0BoPKIRSnbi4KAPT+6EfWXXAzPjBlGLN0SsPo6HYeA54rVFDujUwdOVucjX1zMtjpha5EsTzraExffeByc2w1n0oJGLcqgTI3htPJCqE2xG83upC+xnUmp2GzBFFPsjA7sRlExAWPr64CEbY50nFzqLiW/O2ejsYpduskvEnE2TkxLYKfNx06Zui0axY587CyLbQK7WG8fEI+PqBNxNNRnDBhiXV1w1Ddk3b77kUfBORyo/c53cl5LOByG3+9nP4FRxgtZgXxOZvmiHCumZXi2Xpim2DGT4j72GLM6MakjVsI1RlJU2lnjhGv8eM2qLed0Akn/O1WKnTKw09BFN2rzhFRHVmeBGjuTU7HZauxYfZ1BgZ2jtla291HY6mRcj0FzYpXrAbLX2MkzszWkYiWD4qj2wI7XU7GTxooVpMYuqcBTV6zlsE1gZwTBz9ei589/xrglS1R5Xy1ZsgQ1NTXsp7W11cBV6oOQx8ksX6SWfMRio/p/GQmrCTHQ7gRQ1tjJd9HM6sTsVOwY2fKENU5orK+T0JKOSgnsNLz/o018kGrsjBonJsHsTrJMNmDmxAZ62CX270serz/jSCkW8BrgYQck7H2kmcSjednFBwbZdBDDauxysTtJ/tvxGvw8NadildvreP6RAnbqiiWU2Cawc9b6AIdjxMDjeFc3S/WNeE5DA+LdXRm3D37yMeLd3dh83HFYP2s21s+ajeiePWi/+1fYfNzxGddy0003ob+/n/2sW7cuvxdnArJiZ35gx7ndsrrQUbg6O9O6YofV2AmhEAuqzLI6kXBJF932fYi2JRsnNHbESkipbDUmxUoPQS0GwqOpUzGz7E4Uil3GYCqZpuVNqrFDNDpiyomElKLTe5yYkly97GIdCbWOr6xUbTOSK8O/e+mQzoWSdY0aeI1dsfl+/jPBfOz6+lQ3M+ULpWKti20CO87thnfWLAx+uJw9JgoCBpcvR9m8eWmfUzZvbsr2ADC4bBnbvvq00zDln//AlBeeZz/OpibUX3IxWh59NONaPB4Pqqur2U9VAVQwtbD0Q2Vh1irVdEkn90Jg1tDq4V5akW3bAFGEo6bG8MBjOLJi16GfYqfBpDjfi4Cytm14QCUKgmLyhDmpWMTjEAYH024jqdJOg1OxXHm5PIe5gClqqTN2NC87oz3sAPnfXxgYyFjzx7IXWhQ7l0bFTuc5sRIOnw/gOEAUR/Uz1BuyO7EutgnsAKD+wgvQ9/e/o++FfyC8ZQv23XobhGAQvjPPAADsufFGdPzmt2z7uu+cj4H330f3439EeOtWdD74OwTXrkXteecCSKRtvPvtl/LDOZ1wNjTAM3VKQV6jUQiBAQCFUewAwD1tOgAgtG59QY4PmDNSDJBVI6G/H2IsltI4YeS4q3RI0yeEQAChjZsS69A4dUKCmRSrUOxYB53GehxJsROjUYjBYOq+AwHWcWh0YMd7vew1xDM0UDAfO4MVO47jRlcypXFiBip2bKzYKF52rHEi2SVvBHxVFasBzTh+zq99Ak++qVi9zz2cwyHfSJrcQCFGaFasVbFVYFe9eDGafvQjdD74ALadfgZCGzZg4iMPs9RqdM/eFEuN8oMPQvOv70Hf3/6GbV87HYHX/oOW3z0I7377FeolFAy5YNj85gkAKJs7FwBGNZQ2EtMmT9TUJO6ikbjIRwrUOAEAjsoKNhg9/MUXAPRU7HJPR+X73nNlZRnVKalBgK+oMEU9GM3LLm5SVyyQS1NJUsk0qMYOUKRiR1HsokmrE1eTcYodx/OK9yR9Ojbul/59NAR2GkeKGaXYAYWzPDHrRplQj+1M2+q+fR7qvn1e2r9N+vOTIx6rPvlkVJ98cs77n770Tc1rszICqyspjGJXNncOACC4ejVEUTRduQLMqwnhHA44amoQ7+tDvLdXVuymmltfJ+Ec04RIIAAkO5LzrbGTVQsNNXZaA7ukOhXr7ES8vx+u8ePZ38waJybhqKlBrKMjcyOHSV2xymNkWous2BkY2I3LzcuOKXYGpmKBhEl1vLs7Y52d4Jc9PdViNcUOQML5YdNm0xso5NdEXbFWw1aKHaEdSbHjC1Rj521tBZxOxLu6ENuzpyBrUBYwGw1Lj/T0FlSxA2TLEyChfOUbAPF5dMXm00GXaeh9vNeccWJsHaONN2MGxT7j15J8TzKlYuPdUo2dgalYptiNkoqVauwMnu4jmVRnMimWzNrzMihW2zxhpGJXVxjFzqwMCKEe2yl2hDYKXWPHe73w7r8/QmvXIrhmTd7pQC0YMasxE466OmDbNsS6OhHevgNAIRU7WSFxNY/PWy3lNHXFJrbl87i75zMEVHLjhE/zvlWtI4tZshiLyWbgBtudJI7hS6wlTWAniiILbpxGpmKTil28vx9CMAi+rCztdtEOYz3sJIY3Lw2HGUjnNXlCY1esIYpdYUyKrWh38uSH2/HQO1vRORDGzHHVuO20WZjX4ku77cb2AH772kZ81taPtr4gbv5qKy5ZlFpbf+/rG3H/m5tSHpvaWIGl1x/Dfu8IhLDklQ14b1MXBsMxTG2swFXHTsdXDhyn98vLGQrsSoRC19gBiXRsaO1aBFetRvVXvmL68c1sz5eCjNCaNUA0Cq6sDK7xhfmiS5YnQP71dYBcYyeEzeuKBTIHMbEeycPOpFRstRTYjfSyiyvMyrUEDqrXkiWwEwYG2M2MkWlqvqoKfHk5hKEhRPfuy9h4JlkdGeVhJ5HNpFgURcV4RfMnTxhx7pGC9pjpip21auxeXL0Hd7y0HnecMRsHtfjw+AfbcP5jK7D0hmPQUDnyhjIYiWNifTkWzxmH21/KbFm235hKPHXp4ex35zBz9+v/thr+YBSPXnAo6srd+OeqNlz5l5X411WLMLvZ+Ju7dFAqtkQodI0dAHjnJOvsCtRAYWZ7vhRkDH30MQDAPWWy6TN62VoUxepujTNilRTC7gRQzovNpNgVPhUrBRN8VZUpc6ezBXaS5ydfUZEy41dvOI6Dc5zkZZc+HSsKAmtsM6PGDkjvZScMDrFaU012Jxpr7AQDy0CkoD1utmIXllR4awR2j76/Dd+c34JzDm3BjDFVuPP0A1HmduBvH+9Ku/3cFh9+sngmTps7Hm5H5nOzg+fRVOVlP3UVqa/3kx29uGDhZMxr8WFifTmuPn4Gqstc+LytcGb8FNiVCOwutQAjxSSkztjQunWqT4x6YKpilzzZhjZsAAB4CpSGBcAmAwA6KXaSaqHG7kRPxW54KtakcWJsHVm6YqXHzOiITV3LyIuIpGSa0VTiGiNNn0jfQBHv6QFiMYDjMhrK64WTKXZpArtkRyzncoHzelXvm40Us1DzBJvFXcKKXSQm4PO2fhwxXf5s8TyHI6Y3YOWOvrz2vb1rEPPvfANH/moprv3rp2jrS7VbOmRSLV5asxd9QxEIgoh/rd6DcFTAl6YaV9c6GpSKLQGU3l+OysqCrcM9eTL4mhoI/f0IfbERZQfONvX4hUjFSuqAZ3rhAjvXGH0DO6lOTs28TH0UOx+ANKlYk8aJsXVkqbEzsyNWeZy0ip0JHbESzMuuPX1gF002Tjga6g1XMtmYrd6RCpbyBldLranmrlgzmidM7ooVpBGNBqrBgUAAfsX4Po/HA0+a4/UORRAXxBEp18ZKD7Z0pjcSz4V5E3349dlzMbWxAh2BMO5/YyPO+cOH+M8PjkKlJ/E5/t25B+Oqv6zEvF+8DifPoczlwEPfOQSTG4yZrpILpNiVAPGBAfb/fAEDO47jUHbggQCA4JrVph9fMGlWLDDyYmr2jFglzrFyV2y+VieANoNiMSxdBPJR7DI1T/Ql/m6i3QmQmIIxHLkj1iTFLktgJxXTS8X1RjKal13MBA87CXlebN+Iv0lzYrXOzM57pJiRzRMZuoCNwoyu2NbW1pS57EuWLDHsWOk4dv8mnDJnHGaOq8bR+zXijxfNhz8YxctrZGeH3772BfyhGJ6+9HD866pFuOTIKbjyLyuxYV/medJGQ4pdCcCGXldUmFL3k42yuXMx+P77iaaC89L7ERqFme35w+u9zJ4Rq8Th88E5ZgyEQADuyZPy3l8+NXb51ONkUuxYKtZX+Bo7wURzYkAZ2KWr9zMvRT2al51ZHnaAsnliZCo2Lp0LNZakWFmxE4NBCEND4MvLdT9GOszIgKxbtw7NiixDOrUOAGrL3XDwHLoGUm82OwfCaEzTOKGVmjIXpjRWYHt3Yjbzju5BPPHhDrz2g6Ow35jEzULr+Gp8tL0HT364A78840Ddjq0GUuxKADZCp4D1dRLMqHiV+YqdkZ1pw3EoOzSdTrgnTjT8mJngOA6Tn/0rpjz/nC6peNnuREVgFzWwxi55AXeaVGOXyXZF+ZjZqVjB74eYHKsmISl2Ro4Tk2BedhmaJ6QZ0UaOE2NrUTRPDJ8rHPdr97AD8jEoNu7cw5WXs3pBM1U7M2ZvV1VVpcxlzxTYuZ08ZjfXYNnmLvaYIIhYtrkbB0/y6baewXAMO7qH0FSVWEcwmvjO8cOy+jzHjfjsmQkFdiWAMJA8mRUwDSvhTaZiIzt2FGxotRmBnVPhqeaeONGU9G82XGPHwj15si77ktKpmrpi87gIpDMoFiIRCIOJGhrTu2L9aexOpFSsCR52gCJAEcUR6zHDw06CKXYZUrHRpGLnMkOxkxSsSATi0FDK3wTJ9kmjOwArQ1BRXwoYq9hxHCfX2XWbU2cnxmJsPrNVfOwuXTQFz3y0C//3yW5s7gjgp//4HEORGM4+pAUAcN2zq3D3qxvY9pGYgLV7+rF2Tz+icQHt/hDW7unH9i65Ju/Ol9dh+dZu7OoZwic7enD5nz+Bg+dw2tzE5JtpjZWYXF+Onzz/OVbt6sOO7kE88u5WvL+5Cye2jkWhoFRsCWAlxc5ZWwv3pEmI7NiB4GefofLII005rigIia48mNsVCxRu4oRRyJMnVHTFJuvx8im0VqZApbF0LN3mcJj2+ZbWIYZCEEIh8IruSiklapZix7nd4CsqIAwOIt7Xl9JAInVJmtIVm6zjFAYGEB8YGHETKdXYGT11AgD4sjJwXi/EUAix3l64K+QidnYu1OjnKXfFaq2xMyYIcjQ0ILpnD2JmBXaKmzqr2J2cOnc8egYjuPf1jegMhDFzfDWeuHg+GpPqWltfMKVhpt0fwikPvM9+f/jdrXj43a04fEodnr18AQBgb38I1zzzKfqGoqircOPQybV44YqFqE+md10OHn+8aD7u/vcGXPrERxgMxzGpvhy/OXsujj3A+M96JiiwKwHYXarGgmG98c6dkwjsVq02L7BTnIiMTB1IKC8uhZo4YRTSbEhVzRM6pKJYsBSLQRgcgqOyQq6vq601bf4wX1kJOBxAPI54vz81sDO5xg5IvC9SYKdEHidmfGDHV1SAr66G4PcjtncvHDNmpPyd1diZ0DwBJD4Psb17E4G/omEo7tc+dQKQzx1qOsIT2xtbBiIpdoUI7IzsilXLBQsn44KFk9P+TQrWJFrqyrH9rlOy7u935x486jGnNFTgD985JOc1mgGlYksAWbGzRmBXNifhZ2emUXHKicik1IFUtF5sip3cPJG7aqGHYsF7vezYUhDD6utMGicGJFJfUmAw3MtOWhdvcmCXWEtqzZ9k9+E0oSsWkFW7dA0UbE6swePEJDKZFAt5ngvzHilmUEmGo95ck2LJlxI8X/CGPGIkFNiVAGycWKVFAru5cmCnpsA0smsXtn3jG+h/6WXVx0xV7MwJ7CoOmw++vBzlhx1myvHMQotBsV71jXIQ0wdAYcJr0jgxtg7J8mR4I0fyd6dJqVjlWpSKnSgIiJv83jAvu72pDRRCOMzWZvQ4MbYWX/rOWMnHTutoRe3NE8bW90oNMmaZFJvpMECohwK7EkAIJHzsrKLYefffD5zbDaG/H5Ht23N+XvejjyG0eg36X3hB9TGVxctmpezG3bUEM5Z9ANe4wg2DNgK+QHYnwMggxuxxYiPWkSGwK4hipwjs4v39rLjdLDWTedkNU+ykUWKcx2Pa+8JMinuGKXYsVa41sMtzVmyRKHZWmjpBjIQCuxJArrErfPMEkDgZeFtbASDhZ5cDQigE/yuvJP4/HFJ9zEKciDiOS6m/KhZkuxM1il3+zROAwt4jeYE206tNCZ/GLFmMxSBIipCZil26wC4Z8PLV1aZ95jN1xrI0bFOTaTdVmbzs2OSJvJsntCl2RjUaOOvNHSvGvs8U2FkSCuxKAKvV2AGKdOzq3AK7wOtvsIumGMo9oJCgO0z9kIIzNZYPgk6eV5JSFksGMTFWY2eyYlctBXayxYjSbkRrcb6mtaQL7LrNGycm4WQ1dqmpWNmc2LwuwUw1dvE87U54CxoUA7KljVljxYzu8iXygwK7EiBusa5YQGFUvDo3o+L+F55n/y9qUOz0GEJPJGCdgWoMinWrsUutbWPjxApUY6dU7CSrE766GpzDYd5a0jRPsNpDkxonALl5YvhYMWlOrBnjxCSkQD/eNzwVKwV2edbYRaOq6oONvrE0e6wYUyBNcBgg1EOBXQnAauwsFdglFLvQF19ACGUP1KJtbRj8cDn7XchHsSuwUXAxwAyKTbY7AUaO0FLanZiJHNj1scfMnhPL1pLGuDneIyl25r0vrCu2vT0l6DHTw05CCvRjCsVOjMWYmXW+I8UgiswXMxfMsjuJ9/aOmEBiBOx8aiGrE0KGArsSQE4/WKPGDgCc48fD0dAAxGIIrVufddu+f/wDEMXE9kgYw6qFurj0Q1PzBDMoNqZ5wswAJmUdSsUuGeSZWV+nPJ5yXmyMediZp9hJqVgxGEzpFjZzTqwEq7FTNE8IAwPy3zXe5CrPH9omrxjUPCHd2AhC2lF3ekMZEGtDgZ2NCa1fz+5AsyF5N1kpFctxHMrmjJ6OFQUB/S/8AwBQe87ZANQZ47L9mDgntthhY5VUTJ7QrSt2WNoxVqiu2GEpYeWazFfsfInjpyh25o0Tk+C9XvbvoPSyk5snjJ8TK5Guxk6qgeTKyzUHWMrzh6AlsDPo/MM5nfLnwASTYqpZtjYU2NmU4KpV2HbGmdjz4x9n3U6Mx9mdqpVSsYDSzy5zYDf00ceI7t4NvrIS1V89FYBWxY5ORHqRn0Fxfu8/r1DsRFGU7U5MbBIAlMqhssauL+Vvpq0lbY2dlKI2931J52UX7TRvTixbh6TY9fez1GRchxtczuEA+MRlU9Xn3+DmCQBwNCTr7EywPKEMiLWhwM6mhLdtBwAElr6VUkcyHKWiZ73ALqHYhbJ0xvY/n2iaqP7KV5hKIkYiidmvKqAuLv3QZneiT2DnVAQxgt/PvNqs4GPHFDuzU7HS7NqhIaYisa5YExU7YKSXnSiKco2dmalY6d9AEJhSJ+TZESshN1Co6AqPGn9jyUyKu7sMO4YEK62g86klocDOprDO0HgcgTfeyLiddJfKeb2WGdYs4Z09G+C4xPDqpImpkvjAIPyvvQYAqDnzDFbbBagLKgD5JGy198COKN33c+0M1OvCplTsmFdbRYXp/6589cjATihQKpavqmIqktQlHOs1v8YOUDRQJDtjBb+fKezORvNSsZzLxRokpM+JZE2j1cOO7VvLWDEzFDsTTYpFEwJVQjsU2NkUZSdp4NX/ZN7OglYnEo7KSnimTweQfm5s4NV/QwwG4Z46FWXz5oFTmP2O1kk7HFboTF2xecOCKEHIuTNQskbh3PoYFMf7+xXpRnPVusQ6kjV2AwMQk+8BS8X6zA3sOJ4f2VTSXRjjZpaKTXrZSY0Tjpoa0826h9fZ6dVEpkmxM6EUxMyxYnLNLHXFWhEK7GyK0qR3cPnyjOlY2ZzYOh2xSrzMz25kYNf3XCIN6zvzDHAcl6hvkZzfNQZ2pNjlj9LiINfOQN187CQ1TBAQ3bUr8VghAjvF90maZiDV25mdilUeM97fBzEeZwGe00QfO0CRik0qdtECWJ1IDJ8XK+hk1K5lrJgpNXZmKnZUs2xpKLCzKSljteJxDCxdmn67gWTBcGWlGctSTabO2PDWbQh++ingcKD6tNPY41I6VrVix06sdCLKFy2dgXrVOPIeD7iyMgBAeNs2AOarUkCiC5FPfqeYSlagVKzymPG+vsR6RBHgONODTDZWLFljVwirEwkp4I8NV+zyTcVqGCtmRrOBrNgZH9iR3Ym1ocDOpkiKnfTF8mdIx1pdsSubOw8AEPrssxRjzf4XXgAAVC5aBJfibl9Kx6qusaMTkW5wDgfgdALI7d9BjMUSaVvoo5hKwUok2UAkKTNmIwVTbApGgbpigVTLk1iyccLh85k6AQMAnIrmCVEQEOtIWp2YOE5MwsFMe/sAJOr9AMBRk9+5UMtYMTNuLJ3Jrth4lwnNE3Q+tTQU2NkUqXmi8vjjAACDH36Y4mMlYeUaOwDwTJ8GvrwcwtAQwlu2AEhYtPT/858AgJozz0zZnpnjqlTs6A5TX9Rc3JTBnx5O9VLgFGGKnbmdn8PXISl1heqKVR4z3tfPTHkdJnfEAoBrTBPAcRCjUcR7etg4sUKkYlmNXVLBYje5eSt28lixXDGjK99homIn18zS+dSKUGBnU6SxWt4DZsIzYwYQiyGw9K0R28V1qisxCs7hSHTHQk7HDn7wAWIdHXD4fKg69pjU7ZOKndqxYnSHqS9qLE+U6Vo93n+m2O3Ykfi9ADV2iXXIgZ0Yjcp+kQVW7Ng4MZM97IBEmtKZnBAT3bsPsY5Et7uZHnYSzMtuRPOETnYnmiZPGKjYsRo7Mw2KqRnNilBgZ1MkxYr3elB18kkAAP9/Xh2xnRCw3tSJ4bC5scnOWKlpovrUU0cEAjxLxaptniBDTT2RlLdcauyYLYTDoUtqkPm2JYNKs8eJSTDLk75+5pUGjivI6D5l8wQbJ2Zy44SEc1wyHdu+TzF1ohCKXbLGri/ZPCHZnejVFZtrfWk8zvwWDVXskv/ewtAQhGDQsOMAimY0mhVrSSiwsykCm73pRfXJJwMABpd9KF9gkkgde/mmH4ykTNEZG+vtZY0gvjPPGLFt3ood2Z3ogqzY5RDY6ex5NTzVWTDFTkrF+v0sDctXV5te1wYo1MO+fsSTHnbOAqWoJXUuodhJXbGFa56Qauzi7CZXn8Au58YhRcrWSMWOr6hga4sbnI6lDIi1ocDOpkiKHef1wDNtGjwzpgPRKAJvpnbH6uW2biTeZGdseNMm9D37N4jRKDytM+GdOXPEtqzGTq1ix2bFUmCnB5xHfY2dXlYzw5sTzB6bNXwd8f5+2eqkAGlYYHjzhORhV5j3hXnZtbUhlizkdxWieaJ2uN2JTqlYlV2xKYGdgecfjuPksWJGB3ZR41PLhHYosLMpkt2JlJqsOimh2gVeTU3HygXD1g3sXE1NcI4fB4giuh5+GADgO+PMtNvKip02Hzu6w9QHOR2VQ1eszu/9SMXOl3Y7o5EDu76CdsQCGWrsCtA8AchedqHPP090QzscBQkyWY0da56QUrH5/RupnTyhDACNzhgwyxODO2OpGc3a2C6w63n6aWw+7nhsmDMX2875RtqJBUr8r76KLV9ZjA1z5mLrqadh4J132N/EaBQdv/41tp56GjYcdDA2HXkU9tx4IzPVtDKsKympYFWfdCIAYGDZMpZyAJTeTdYN7ACgbE6izk4cGgLncqH6q6ek3Y73Sl2x2lKxZFCsD5LjvJCL3Ynegd2w4KlQKUdl80QhO2KVx01M5Eh2xRZIyZS87IJr1wJIjBIrSHo6GdgJg4OIDwww5Vi/WbE5BnbSdi4XOI7L69ijwUyKjVbsqCvW0tgqsPO/8go67robDVdeiSnPPwfv/vtj56WXMd+m4Qyt/BRt198A39fPwpQXnkflCcdj11VXI7RxI4CE6hNatw4NV3wPU557DhMefADhbdux+4orzHxZmpCbJxIKlmfGDLinTQOi0RSzYiGQ7NSzfGA3h/1/5XHHsbvt4XCepGKnunmCTkR6It1Q5JSK1V2xUwR2DkfBPtvMx66vH/H+vpTHCrWWeF8f64oslGLnTM6LFZMF/IVonACS57xkQBnduTPxIMcxY2mtqJ08wW4qTajvZYqdwdMn6HxqbWwV2HX/6Qn4zj4bvrPOhGf6dIy97VbwXi/rohxOz5+fROWiRai/5BJ4pk1D07XXwts6E71P/wVAQsWa+PjjqP7KV+CZOgVl8+Zh7M0/Q2jtWkT37DHzpalG2TwhUX1SsjtWYVYs15VYt3kCAMrmzWX/n65pQoLLU7GjE5E+qGme0Dtto1TFHD4fOL4wp7HUGru+lMdMX4v0nsRiiLa1JR4rUFesK9kVy34vQH0dkJyhO8wah6+szPvzororNmpeR75ZlifsfOqh86kVsU1gJ0YiCK1di4qFC9hjHM+jYsECBFetSvuc4KrVKdsDQOURizJuDyTtQTjOspMaJJR2JxKS7cng++8jHghAFEVFV6y1FTvvrFlwT54M76xZqDjiiIzb8R6NdidR6orVE3XNE6llA/miDJ4KZXUCKOxOLJCK5cvKZBU1GUgUqlvY2djIlDKgMB2x7NjJzwfzPNThvK66ecLEjnyzTIqptMXa2Cawi/X2AfH4iLtQR0N9xkLRWFcXHPUNOW8vhMPo+PVvUH3KKVlnq4bDYfj9fvYTUNS0mQVT7LyyYueZMQPuqVMhRqMYePttCINDbJST1WvseI8HU//9Cib/7VlwyXFV6ZAUOyFIkycKCa+peUKfC1uqYle4wI7V2Pn9BVfsEuvxKX5xFGwtnMORkn4tVCoWkD8fkR2JVKweN+xqR4qZmS1gY8W6jW2eoAyItbFNYGc0YjSKtu//ACJEjL3151m3XbJkCWpqathPa2urSauUYYqdQgXhOA5VySYK/6v/gTCQDDhdrpQA0KpwHDdqkbVUU6i6xs7EdEgpwLlV1NhF9b27V6ouhbL0ABRBXDyOaFuidCOl/s/s9SgD3rragqWoAcCVrLMDCjMnVkJSLZlip8MNrtw8oTIVa6ZiRzV2JY1tAjtnrQ9wOEbUDsS7utkImxHPaWgYceeSbnsxGsXuH/wA0T17MPGxx7KqdQBw0003ob+/n/2sW7dO/QvKA1EUWYfX8ICNmRW/9x6ie/YCSJzMjO7GMgupplB9jZ3xQ7hLCWbSqqYrVqf3nnO7wVdUACic1QmQuMmQ0p9RKXAoUCoWGJaiLlBHrITUGQsk7IwKBQvsks0TjhodUrFqu2LNVOySNXaxHmNr7CgDYm1sE9hxbje8s2Zh8MPl7DFREDC4fDnK5s1L+5yyeXNTtgeAwWXLUrZnQd2OHZj4x8czdmMq8Xg8qK6uZj9VJqc5sw1V9+y3H9yTJ0OMRND/4r8AWD8NqwZmd0JdsQVF7ood/eImN/roN35ICmIKZXUyfB2sxs4iqVhHgTpiJZxj5QYKZwHmxEo4kjV28WT5jR4TeKQbFNXNEyYqdvGeXojJMhwjoPOptbFNYAcA9RdegL6//x19L/wD4S1bsO/W2yAEg6yLcs+NN6LjN79l29d953wMvP8+uh//I8Jbt6Lzwd8huHYtas87F0AyqLv2+wh9vhbj77kHiMcR6+xErLNT1YBnsxEV5rzDZ/VxHCfPjn3xpcQ2RRTYMbsTzV2x1DyhB8zyoQA+dgDAJ1OehayxA9JMwbBIYCfZXhSK1FRsAZsnht2o65mKzXWkmJnqFmsmisfZzYYRUGBnbTJXqVuQ6sWLEevpReeDDyDe2QXPzJmY+MjDLLUa3bMX4ORYtfzgg9D863vQed/96Lz3XrgnT0LL7x6Ed7/9Etu3dzDPt22np1psTHziCVQcPt+kV6YOlv5yONLeBVaffDK6//AQhIGEh52Vx4mpRTYo1qjYUSpWF9hot5yaJ/Svb3SNH4/wuvVwTWjWbZ9aGBHYFTIVm1JjV2DFLpmK5crLWdq8EAzvDOb1SMWyrtjcUrEwsb6Xc7ngqKlJdGr39OSUgdICdcVaG1sFdgBQ9+3zUPft89L+bdKfnxzxWPXJJ7O6s+G4JzRj5ob1uq7PDNI1Tijx7L8/XJMmIip1gumQfrAKnLcMQG61XUpIsdMXqXkiF9XCiPd+zI9vQtWxx6HyqKN026cWeGVgx3EFVcdTauwKnIr17rcfwHHwzJhe0Pre4YquQ49UrMquWMFEuxMg4V8Y7+9HrKsbnmnTdN+/KIqk2FkcW6ViiQRSGjJTpyvHcag+SQ5mSbGT61zoDlMf1BgUG3ERcE9ohu+sM7Na45iBMpjiq6sLMjqLrUWp2BW4ecI9aRIm//3vaPnd7wq6juGKnR7nQq0jxcwKgqS607hRDRSxGCCKACiwsyoU2NkQqXGA82YuRq9O1tkBxabY0UgxK6DOoDhxI5JJYbYzysCukFYnQGqHcKEVOwAomz0rYVZcQIYbWOvhY6d1pJiZih1gnOWJoLiZo/OpNaHAzoYILBWb2ZvOM3MmXBMnAgAcVfnNRrQSrBtTRfMEpQ70R1bscqixixZvfWNKYFfjK9xCMLzGrrDNE1ZhpGJnfirWiBrTbLCxYgYpdkr/PjqfWhMK7GwIG9GUxXSY4zg0fPcy8JWVWUd02Q3JoFhVKlaRMqETkT7IzRPmz4q1EkqVrpAdscOPX8hRa1aCLytLOU/qUQOpeqSYiXYngPEmxex1OxwFLT0gMmO75glCTsWOltryff3r8H3962YsyTSY3YmK5glB0b1Gs2L1QbZ8KIzdiVVITcX6CrcQpKpThe6KtRKO2lrE9ibN2gs6Usycc480Vixm0Fgx9nosWFrx5Ifb8dA7W9E5EMbMcdW47bRZmNfiS7vtxvYAfvvaRnzW1o+2viBu/morLlk0JWWbe1/fiPvf3JTy2NTGCiy9/piUxz7Z0Ytf/+cLrNrVBwfPoXVcNZ68ZD68rsIEvhTY2ZDRmieKGWXzhCiKOXXcUepAf9hIsVyaJySF2VN8731qKrbQNXa1qPna1wCXU5cAplhw6hzYqU7FRs2deiMF9XGDFTveYjfJL67egzteWo87zpiNg1p8ePyDbTj/sRVYesMxaKgcGYQGI3FMrC/H4jnjcPtLmadH7TemEk9dejj73TlsVN8nO3px4eP/xfeOnYbbvjYLDp7D+r1+FHLYEwV2NiRXxa4YUQazYjicU3DL/KYodaAbai5uxazY8dXWCew4jsP4u+8q6BqsCFMynU5wZWV570/7SDGTFDupecKoGjuLfp8ffX8bvjm/Becc2gIAuPP0A7F0Qwf+9vEuXHHM9BHbz23xYW5Szbv73xsy7tfB82iqynyduf2ldbjwiMkpx5jWWNi6dgrsbIjUPFGSip0imBVDISCXwC5qzRORneE9KponitjMNKXGrsCpWCI9UmCn18xszYqdSZ9/wxU7aUSgCa8nEAjA7/ez3z0eDzxpBI1ITMDnbf244hjZt4/nORwxvQErd/TltYbtXYOYf+cb8Lh4HDyxFj86+QA0+xI3CF0DYaza1YfT543Hmf/7AXb2DGFqYyV+eNL+OGxy4cohqHnChkgdoXwWu5NihXO5gKTqlmudnVXvMO0Mp6J5opjffyvZnRDpYYGdTulpVl+qVrEzKXUpKXbCwIBqI/dcMLMZqrW1FTU1NexnyZIlabfrHYogLogjUq6NlR50Dmh/D+ZN9OHXZ8/FExfPxx2nH4hdPUM45w8fYiAcAwDs7BkCANz35iZ8c/5E/Omi+Zg9vgbnPbIC27oGNR83X0ixsyGShxuXxe6kmOE9HghDQzl3xtLUCf1RMy/TysXW+cJXViZuNOLxgqdiifRI/n56eNgB2rtizVKs+aoqcC4XxGgU8Z4e8OPG6bp/M+1b1q1bh+ZmeWxgOrXOSI7dv4n9/8xxwLwWHxbdtRQvr9mDbxw2EWLSqPnc+RNZCnh2cw2WbenC3z7ehRtPPsDU9UqQYmdDJMWuGC+UucBMinP0sjP7jrkUYM0TauxOitDHjuM4FtBRKtaaSJMYHDqNe2MBTSwGURBG3d7s8w/HcbJJcZf+dXZm3qhVVVWhurqa/WQK7GrL3XDwHLqGqXOdA2E0pmmc0EpNmQtTGiuwvTuh1Em1dzPGpNbUTWuqxJ6+oG7HVQsFdjaENU+UYCoWkCduiDlOn5C7uIovsCgUqgyKizgVCwD1l16KqhNPhLe1tdBLIdJQeeyxqFi4AL5vfVOX/Sk/x1YtRTByrJgVMyBuJ59QyjbLFi+CIGLZ5m4cPMmn23EGwzHs6B5CU1XiGjShtgxjqj3Y2pmadt3WOcjq8AoBpWJtCLM7KdlUrKTY5RjYmVy8XArwipFio9nOWPFCoCf1F19U6CUQWXCNGYOJjz+u2/5SArtodNQGLrMNigFjx4pJzWhWa4a6dNEUXP/31Thwgg/zWmrw2PvbMRSJ4exDEinS655dhTE1XpYejcQEbOoIAACicQHt/hDW7ulHhduJyQ0VAIA7X16H42eOQbOvDB2BEO59fRMcPIfT5o4HkFBHv3vUNNz3+kbMHFeN1nHVeG7lbmzpHMDvv31wAd6FBBTY2RCptqxkFbuy5PQJap4oGOy9FMXEZI8s7y29/0QxoQzQSlKxk7piLZYBOXXuePQMRnDv6xvRGQhj5vhqPHHxfDQm1bW2vmDKDWi7P4RTHnif/f7wu1vx8LtbcfiUOjx7+QIAwN7+EK555lP0DUVRV+HGoZNr8cIVC1GvSO9esmgKwrE4bn9pHfqGopg5rgpPXXo4JtVXmPTKR0KBnQ2ROp1IsctNsSvmkVaFQllfI0QicGQL7JKf11L0XSSKD47jAJcLiEZzC+yKTLGz8vn0goWTccHCyWn/JgVrEi115dh+1ylZ9/e7c3NT3a44Znpar7xCQYGdDRGZj11pXihZjZ3a5gkLnojsipo6I3r/iWKDd7kg5BrYFUKxSwZ2/pdeAl9ejupTToFn6pRRnpUb9H22PtQ8YUME1jxR2opd7s0T5t8xFzsczydUC4yeEheoxpEoMlRNXinA579iwZfAV1Qg1tmJrv/3/7B18WJsO+vr6P7jnxBtb89r3+x8Sgq8ZaHAzobIszdL84ul2e6EAgtdyWUYuigIiRo80PtPFA9qxoqxyTcm3lh6W1sx4913MP6ee1Bx9FGAw4HQ2rXouPtubD7mWOy44EL0/v3viPf3q953sTdDFQMU2NkQuXmiVBU7lXYnFFgYAjMpzqLYKS98kvcdQdgddbOSC3P+4SsqUHPqVzHxoYcw4713Mfbnt6DskEMAUcTQihXYd/Mt2HTMsQitX69qv3SjbH0osLMhpd48ISt2NHmikMhjxTKrFso0LU/vP1EkqJm8IhRAsRuOs64Otd/6FiY//RSmv/kGGq+7Do7GBojBIIKr16jaVzHPfi4WKLCzIaVud8Kz5gm1gR2diPREVi2yKHbKCx/VOBJFgjxWLIdUbIEUu0y4mpvR8N3LULEg0SUqBNVNSJC+71Z5PXbn7S868NF2uXv5yQ+34yv3v4drnvkU/UO5zSMeDgV2NoQpdiWaiuU82mrs6A5TX5hJcbZUrCKozmZiTBB2QlPzhMVubPiycgCAMKRuWL2V7U7syJJXNmAgFAMAbNjnxx0vr8ex+zdiV+8Qbn95naZ9kt2JDWGKXck2T6hU7FgqhE5EeiK9n1mbJ+giQBQh6mrsrPkd4MsSI69E1Yqd9HpK8/qjN7t6hzC9KTFr9t+f7cPxBzThRycfgM/b+nHhHz/StE9S7GxIqSt2zKBY5axYqrHTF6nGLlvzhFDiHdxEcSKdS3LrirWoYlcuKXZDqp5ntdSy3XE5eISicQDAB5u7cOSMRgBATZkLA2FKxZYEYjwu20eU6MVStUExO7HSiUhPcmqesKhaQRD5kKtiJwoCEIulPMcq8OUJxU4Y0qrYWStQtSuHTa7F7S+vxwNvbsLq3X047oAmAMC2rkGMqynTtE8K7GyGMv1YqqlYyeYlV8WOakKMgakWWe1O6CJAFB/MwzE6SmCntPux2I0lJyl2mlOx1no9duW2r82Gk+fwymd7ccfpszG2JnF9e/uLThy9X6OmfVKNnc1Qpr1KNRUrvW4aKVZYmJ9glosbNa4QxYjcFZvbOD3Aejc3cvOEylSsNPuZvtO60Owrw+MXHjbi8VtObdW8TwrsbAabE+tyJcY6lSA8C+xUjhSjE5GusOaJbIqdVA9KhdZEEZFL4xAwXLGzWGDHFDu1NXZ0o5wvgVDutXNVXvWfGwrsbIZk8VGqah2gsDsZZUapBNWEGANrnshycaM0OFGM5DpSjAV+Lpfl7H7kGjt1gR0zXC7RUiA9mHPba8j107B1ySmq90+Bnc2QxmhxJWpODORhUGyxGhe7wy5uYbI7IUqLXCdPsFIEi6l1gMLuRGvzBJ1PNfPMZV9i/7+7N4i7X92Arx8yAQdPrAUArNzZi+c+2Y0fnXyApv1TYGczBOZhR4odKXaFhcvJoJjS4ETxkXNXrEWtTgCyOykkX5paz/7/3EeW42enzMTX5jWzx77cOgYHjK3CX1bsxNcPmaB6/6VZpGVjJHWEFDs1BsV0IjIC1jyRi0Gxh957onjIdaSYlRVrLqnYUVdsYVm5sxdzJvhGPH5gcw1W7+7TtE8K7GyGlIotacXOq02xoy4ufZFSMUK2WbHUQUcUIaoVOwt+/vnyCgCJwE4UxZyfJzdEWU+FtCPja8rw1//uHPH4sx/twniNPnaUirUZ1DyhMMZNnpBGK0qmO0xj4HJR7GicG1GE5BzYsXo06wVBUvMEYjGI0WjO50e6UdaXm7/aiv956hO8/UUn5rX4AACrd/dhW9cg/vDtQzTtkwI7myErdqWcik0GtaKY0wmJAjtjoOYJolTJdaSYpRW7MlkNEoeGAJWBnRVfkx059oAmvP3DY/DU8h3Y3DEAADh+ZhPOO3wSxvtMnDzR98I/EHj7bfZ7+z334IvD5mP7N7+FaFubpoXkSs/TT2Pzccdjw5y52HbONxBcsybr9v5XX8WWryzGhjlzsfXU0zDwzjspfxdFEZ0PPICNRx6JDXPnYcdFFyGyfbuBryA/pOaJklbsFK89lzo71p5vwbtmO8OaJ8juhCgxikGx45xOubs3xwYKMXkzDZDdiR5E4wLOfWQ5QlEBPzzpADz0nUPx0HcOxQ9POkBzUAdoDOy6H3qIqSZDn36K3r88g6YbboCjthbtd92leTGj4X/lFXTcdTcarrwSU55/Dt7998fOSy9DrLs77fZDKz9F2/U3wPf1szDlhedRecLx2HXV1Qht3Ci/lkcfRc+fn8K4W2/F5L89C76sHDsvvSzn+i2zkaYt8CXcPMG5XEAy/SrkENhRF5cxsLFKWQ2KyfOKKD74IqixA2TVLtcGihTDZYu+JjvhcvDYsC+g+341BXbRffvgnjgRADDw5puoPvHLqP3GOWi67gcY+vgTXReopPtPT8B39tnwnXUmPNOnY+xtt4L3etH33PNpt+/585OoXLQI9ZdcAs+0aWi69lp4W2ei9+m/AEjcffQ8+SQa/ud/UHX88fDuvz/G330XYh0dCLzxhmGvIx+k+ahcKTdPcJw8ViyHANzqJ1e7IhsUZ7M7IasZovhQO1LMioodAHAV6ixPUkek0flUD06f14xnP9ql6z411djx5eWI9/XBNX48Bj5YhvoLLwCQONEbpXSJkQhCa9ei4buXscc4nkfFggUIrlqV9jnBVavZ2iQqj1iEwJtvAgCiu3cj3tmFioUL2N8dVVUomzMHwVWrUXNKesfncDiMsOJ1BgL6R9yZkBS7UreP4D0exIPBnFKxVBNiDPJYpcx1RvTeE8VIrqlYq5ciyPNic1TslLPKLRqs2o24IODp5bvwweYuzG6uQbnbkfL3m7+qfmaspsCuYuFC7P3ZzfC0zkRk+3ZUHHUUACC8eTPczeO17HJUYr19QDwOR319yuOOhnqEt21L/5yuLjjqG0ZsH+vqSvy9M/HfkftsQKyrM+NalixZgttuu03tS9AFsjtJwCxPQjkoduSUbgi5GRRTBx1RfLDatKh9DYoB9fNiU0akleiscr35oj2AWc3VAIBtXQMpf+NyHjyWiqbAbuwtN6PzvvsR3bcPEx64H87axBiM0OdrUZ1B5SombrrpJlx33XXs97a2NrS2qo+qtUB2Jwl4lorNrtiJ8TgQjwOgdKDeqDIodlONHVE8qJ0Va13FThorpi6ws+KINLvy1+8uGH0jlWgK7BzV1Rh7y80jHm+85uq8F5QJZ60PcDgQH9YoEe/qhrOhIf1zGhoQ7+7KuL2zMfHfeHc3XE1Nim264Jk5M+NaPB4PPIpicL/fr+q15ANT7Eq4eQLIXbFLqQkhxU5XuByaJ6T6O6te2AhCC3IZgs2bJ5hil2MqNkLNUHZAU2A38N574MvLUX5Iwjyv5+mn0ff3/4Nn2jSMveVmOGpqdF0kkPhieGfNwuCHy1F1wgkAAFEQMLh8OWrPOy/tc8rmzcXgh8tRd4FcZze4bBnK5s0DALgmTICjsQGDHy6HNxnIxQcGEFyzBr5vfVP315AroigiGI2n/dtgKIqQw42Qy4uhSMzklVmHsLcCIYcbg0NB8Fneh/hAECFH4qQaAg+uhN8zvQk6XAg53IjGhYyfxaFIPPF5dbhL+vNKFBdB3oGQw41YTMz6uR4KxRByuOF2eiz5+Q+XJ8+jgSG4c1hfcDCEkMMNh7dixOspczlGNYsn0rNmdx9eXrMXbX1BRONCyt8e+s6hqvfHiWpmiSTZeuppaLrhelQefTRCX2zE9rPPRt2FF2JoxQq4p07F+CW/VL2QXPC/8gr2/PgmjL3tNpTNORA9TzwJ/6uvYtorL8PZ0IA9N94IZ9MYNF2fSJMOrfwUO84/H03XXYfKY46G/+VX0PXww5jy3P/Bu99+AICuRx5B9yOPYvxdS+BqnoDOBx5A+IsvMPXll3I2Ad69ezdaWlqwa9cuTJigfmDvcIYiMbTe8p+890MQBEEQZrDuFyeh3K3PzAO9r6lW5l+r9+D6v63CUTMa8d6mLhw5owHbugbRORDGSbPG4tdnz1W9T03/CpG2NrinTQcABF57DZXHHIOm636A4Nq12HX5/2jZZU5UL16MWE8vOh98APHORLp04iMPs9RqdM9egJMLOssPPgjNv74Hnffdj85774V78iS0/O5BFtQBQP2ll0IMBrH3lp9D8PtRdsjBaHnk4ZKe7EAQBEEQhPH871ubcfNXW3H+gsmYdcur+Pmps9BSV4afvPAZGqu01dJrCuw4lwtiKJGTH/zwQ9R87WsAAEeND8LAQLan5k3dt89D3bfTp14n/fnJEY9Vn3wyqk8+OeP+OI5D4zXXoPGaa3RbY76UuRxY94uT0v5t53cvR/CjjzF2yS9Rc3L6bUqBtht+hIE330TTTTei9pxzMm4X2bET204/A3xFBWa8/66JKyx+onv2YOspp4Jzu7DfiuVpt9l+7rcRXr8ezQ/cj8ojF5m8QoIwhvD2Hdh+xpngKysx4713Mm7Xce996H3yz6i94Dto+v73zVtgjnQ++Dv0PP5H+M79Fsb88IZRtx/4YBnarroa7v33x5S//iXlb2UuR4ZnEdnY0T2EY/dP1Pi7nDyGojFwHIdLFk3Btx5Zgeu+vN8oexiJpsCu/OCD0X7X3Sg7+CAEP/sMzff+FgAQ2b4drjFjtOySUMBxXEZJ2xschBiPoLLCq5vsbUfKvU7E4hF4I+Gs7wMvxuCNR+BwVpb0+2UEscpyeOMRIBjJWF/jDQ+Bi0dQUe6h958oGlwVXnjjEXChwayfa280DG88gnK305Kf/4qKMgzFI/CO8jok4kIU3ngEXlfmaxShjpoyFwaT9Ypjq734Yl8AB4ytRn8whlAkfa39aGgyohl788/AORwI/Oc1jPv5LSyYG3zvXVQceaSmhRC5IRlAl/LkCUD28RNC2bu52Dgxas/XHWVnXCbbB6vbPRCEFpQGxdnK1K3++efLkyPFBlXanZB9kW7Mn1KH9zcl3DsWHzgOv3hxHX783Bpc88ynWDi9fpRnp0dTyO0aPx4tD/1hxONjbrpJ0yKI3JEmLZDdSdLHLke7E6ueWO2M8j0Vw2EgzXssGbjS+08UE+xGURSBWAzIcONoH4NilXYn9H3WjV98bRbCsUQn7FXHTofTwWHljl58ZfZYXH3cDE371KylivE4Am+8icjWLQAA9/TpqDruOHAOyrMbCSl2CaTAdlSDYov7SNmZlMAug58XU0zpDp8oIoZ/9jMFblafvMIlDYpznRVr9RFpdsRXLr+XPM/himOm571PbV2xO3Zg13cvR7SjA+4pkxOPPfwIXGPHouWhP8A9cWLeCyPSQ4pdAs6jzqCYTkT6w3FcopEqGs1oUiw9TlM/iGJCeT4RIhHwFRVpt7P6jSWbFZurYhem86neXPfsKnxpWj0On1KHSfXpP0dq0RTY7bvzTrgmTsTkZ/8Kh88HAIj19mLPj27EvjvvxMSHHtJlccRImGJX8iPFkopdaDTFTjoRUWBhBJzHkwjsMip21lYsCEILnNMJ8DwgCFnHislzqq15/mGp2KHBnLanG2X9cTl4/P7tLbjxuTUYW+3F4VPq8KWp9Th8aj2mNGgL9DQFdkMffYzJf5WDOgBw1tai6frrsP3c9FYkRP6IosgCmVIf6cIUuyzjrADlbEM6ERkBG4YeHhnYiaJIFwKiaOHcboihECs3SIflFbtyaVas2ho7awaqT364HQ+9sxWdA2HMHFeN206bhXktvrTbbmwP4LevbcRnbf1o6wvi5q+24pJFU1K2uff1jbj/zU0pj01trMDS648ZsT9RFHHhHz/COxs78dB3DsFJs8bmtOa7vz4HALCvP4QV27qxYlsPHnlvK37ywmdoqvJi+U+Oz2k/SrT52LndEAZHRvjC0JBl70yKgmgUEBJFlqVuoMyXSc0Toyh2Fj8R2R3pBkOMpAmwY7FEcTmse2EjCK0k/FxDWefF2kaxU9k8YcXrz4ur9+COl9bjjjNm46AWHx7/YBvOf2wFlt5wDBoqR643GIljYn05Fs8Zh9tfWpdxv/uNqcRTlx7Ofnfy6c1EHnt/G/KZqFZT5kJtuRs1ZS5Ul7ng5HnUVWg7b2oK7KqOORr7fn4Lxt1xB7xzEtFmaPVq7Pv5rag69lhNCyFGR6lOlXoqVlbscgzsSLEzBClgTndxU6p4pa4wE8UHszyJ5hDYWfTGhlfZPMFKWyx4Pn30/W345vwWnHNoCwDgztMPxNINHfjbx7vSNiTMbfFhblLNu/vfGzLu18HzaBplAsTaPf149L1t+NfVR2D+nW+qWvevXt2A5Vu7sXaPH9ObKnH4lHp87+hpOHxKPWrKtd0QaArsxvz0p9jz45uw/ZvfStQaABBjMVQefxzG/IQsT4yCqVMcZ9kThVlwrMZulFSsxVMhdkfys0rXPKG84NH7TxQbSi+7TFjd7oRLKnZiOAwxHh/V1cLsQDUQCMDv97PfPR4PPGluEiMxAZ+39eOKY6axx3iewxHTG7ByR19ea9jeNYj5d74Bj4vHwRNr8aOTD0Czr4z9PRiJ49q/rsIvvjZr1AAwHb9/ZwvqK9y49oQZOHnWWExtrMxrvYDGwM5RXY2W//1/iOzYgfCWrQAAz7SpcE+alPeCiMzIVieetC7/pQTvVanYUWBhCJISJ6S5uLELntMJLkP6giDsSja1WkK0uI+jlIoFEulYR2X2oIJdg0x6Pa2trSm///znP8ett946YrveoQjigjgi5dpY6cGWztwaQ9Ixb6IPvz57LqY2VqAjEMb9b2zEOX/4EP/5wVGo9CTCp1+8tA6HTKzFiTnW1A3n5auPxIpt3Vi+tRuPvrcNLgeHw6fU40tT6/GlqXWaAr2cA7v2JXdl/fvQihXs/8fc9GPVCyFGh1mdUFpLru0aRbEj3yVjYapFuuYJeu+JIoZnqdhsXbHWzhhwbjfr7hWGhkYN7Mx+PevWrUNzczP7PZ1aZyTSDFcAmDkOmNfiw6K7luLlNXvwjcMm4vV17fhwSxdevkb7xK3W8dVoHV+Ni45ING6s2+PHY+9vwy3//ByCKGLrklNU7zPnwC60fn1uG5a4kmQkkmdbqdfXAbJil3vzhDVPrHaH82ROR5HVCVHMSHVm2RQ7NnnFqqlYjgNfXg5hYABiDnV2Zp9Pq6qqUF1dPep2teVuOHgOXQOpN/qdA2E0pmmc0EpNmQtTGiuwvTvxXi3b0oUdPUOYc9trKdt976lPcNjkOjx7+YJR9ymKItbu8WP51oRq99H2XgyEYzhgbBUOn2LwSLFJTz6h6QCEfkidh1yJmxMDauxOrH3HbHfkOqM0NXYmp20IwkykYC2nGjsLfwf4sjIIAwM5dcaywM5jrdfjdvKY3VyDZZu7mM2IIIhYtrkb5y/Ur0RsMBzDju4hnHFQ4hr8vWOm4ZuHpQ5kOOm+d3HzV1txwswxOe1z7m2vYSgSx8xx1Th8Sh2+edhEHDalDjVl2m8GNI8UI8xHTsWSYscMikc5GVndbsDusOaJLIqdlS9qBKGVnJonItZungDUWZ5YWYW/dNEUXP/31Thwgg/zWmrw2PvbMRSJ4exDEl2y1z27CmNqvLjx5AMAJBouNnUEAADRuIB2fwhr9/Sjwu3E5KQx8J0vr8PxM8eg2VeGjkAI976+CQ6ew2lzxwMAmqq8aRsmxvvK0FJXPuLxdNz3zXk4bHIdqrz6fUYosLMRlIqV4bw5Knbsjtm6J1Y7IxsUj/x3oPpGophhn32bK3ZSZ6wwaL1UrBpOnTsePYMR3Pv6RnQGwpg5vhpPXDwfjVWJm8+2vmBK02G7P4RTHnif/f7wu1vx8LtbcfgUOYW6tz+Ea575FH1DUdRVuHHo5Fq8cMVC1OuY3j3ugISyt71rEDt6hnD4lDp4XQ6Ioqi5SZICOxshDbyn5gnFexCPQ4xGRx3CbcUTUTHAmljSNk9EU7YhiGKCy6l5wvoZA+ZlF7R3YAcAFyycjAsWTk77t+H1bi115dh+V/bGhN+de7DqNYy2z+H0DkZw5V9W4sOt3eAAvH3DsZhYX44f/d8a1JS58LOvto66j+GQB4GNIMVORvkeZFPtrJw6KAayN09INXbWvagRhFZGS8WKgpCYvgLrBkKAcl7s6IEdqfD6c/tL6+B08Fj24+NQ5pJ9BL86dzze2dipaZ8U2NkIpthR80SKCpStM9bqd5h2h8/WPEFzeokiRm6eSK/YKZU8Oyh2o9UrAzTJxwje3dSFH598AMbVlKU8PqW+Am19uY16Gw4FdjZCSAYwHDVPgOM42Rw3i5ed1Q1C7Q5HzRNEiTKqYqd43MrfATWKHX2n9ScYiaHMPXLiR18wArdTW4hGgZ2NEENkd6JESseKWaZPCDaocbEz2Zon6CJAFDOjTZ6wi2LHlUvzYlV0xVrM7sTOHDalDs+v3M1+57iEVctD72zFgqkG+9gRhUcanyVZTJQ6vNcLob+fKZnpoODCWFjzRJp0lHIEHkEUGzkrdi6XpUdAarE7ofOpfvxk8Uyc+8hyrNndj2hcxJJ/r8fG9gH0DUXx3PdGNzhOBwV2NkKk5okUJOUy3QB6CXkIN52IjICpFmkVO+tbPRCEVkYbKSY9zltYrQMAvkxKxY4+U5UCO32JxgXc+q+1ePSCw/D+pk5UepwYjMRw8qyxOH/BJDRVa7vWU2BnI6h5IhXJqDl78wQFF0bCM8UuWyrW2hc2gtBCroqdldOwgLrmCeqK1ReXg8eGfQHUlLlw1XEzdNsv1djZCGZ3Qs0TABQmxdmaJ+hEZCjZTFrpvSeKGdYVG81eY2f1zz9fQc0TheT0ec149qNduu6TFDsbwUaKkWIHQKEWZWmeINXIWFhXbFqDYslDkD6vRPFRbIrdaM0TdvHlsxtxQcDTy3fhg81dmN1cg/JhHbI3azAopsDORsjF6KTYASoVO4ufXO1KbgbFdBEgig+pbjfTSDG7qFtcjnYnqfYtdLOmF1+0BzCruRoAsK1rIOVvHGikWNEjKXZkd5KA96pR7Kx9crUrTLWgWbFEiTHaSDG5ccvaN5WseWKUGjtlYMdTBkQ3/vpdbZ2v2aAaOxshXTx56ooFICuXQjBLYCd1plFwYQhy8wTV2BGlRc6pWIt//mW7k9wVO1g8WC11KLCzEeQLlgpHil3BkZsnyO6EKC1yHSlm9c8/zwyKRwnswnJphZV9+QgK7GyF3DxBih0g252QQXHhkEeKjby4sQsBudQTRUixNU+IozRPUGmFfaDAzkZQ80QqTLHL0DwhiqJt7prtSrYaO7krlt57ovjIdaSY1c89ylmxoihm3I4UePtAgZ2NILuTVJhilyEVa5dZjXZGmhmZLbCjCwFRjIym2NllTjWXbJ6AKGaf4iO9HioFsjwU2NkIpthRKhaA/D5kVOxS2vMpuDACjponiBIl15FiVv/882Xy9SRbnZ1kxEyeoNbHNnYn8b4+7LvjTgy89RbA86g68csY+5OfgK+oyPgcIRxGx913w//yKxCiUVQecQTG/vwWOBsaAAChDRvQ/fAjGFq5EvHeXriam1H7zW+g7vzzzXpZqmCKHd0xARjd7iRFsbP4ydWuKC0fREEAx8v3igK7ENDnlSg+iqXGjnM4wHm9EEOhhElxXfrtqLTCPthGsWv74Y8Q3rwZEx9/DC1/+D2GPv4Ye2/5edbntC9ZgsBbb6P5/vsw6cknEevowO6rr2F/D61dC0d9Pcb/6m5MfelFNPzP5ej47b3oeeppo1+OakSFTE5SeILRDIrZCdfpTAk4CP1QfhaHKxfSNAoKqoliRO6KzRTY2UOxA+Q6OzGL5Qm7/ris/3pKHVsoduEtWzD43nuY/Pe/o+zA2QCAsT/7GXZ993I0/ehHcI1pGvGceCCAvueeR/M996DiS18CAIxb8ktsXXwKgqtWoWzePPjOOivlOe6WFgRXrULg9ddR9+3zjH9hKlDWPlAqNgHPUrEZFDtKBRqO8r0Vw2FAGejRODeiiBlVsbNR6pIvK0Mc2VOx1BVrH2whYwRXrQJfXc2COgCoWLAA4HkE16xO+5zQ2rVANIqKhbKrs2fqVDjHj8PQqlUZjxUPDMBRU5N1PeFwGH6/n/0EAgF1L0gDyuCFUrEJmEFxhoJfljqweCrEzijTTMMvcJS6IYoZ5uGYqcZOUuxsoHAxL7ss0yfoRtk+2CKwi3V2wVmXmvjnnE44amoQ7+rK+BzO5YKjujrlcWd9Q8bnDK38FP5//xu+c87Jup4lS5agpqaG/bS2qh/SqxYWvDgclq/ZMAtWYzeKYgcb3DHbFY7jMlqeUBcdUcyMrthJqVjrn39ymRfLAlX6PluegqZiO37zG3Q/8mjWbaa+8rIpawlt3IjdV16JxiuvQOWiI7Jue9NNN+G6665jv7e1tRke3FHjxEi40exOmGJHd5hGwnk8ECMRCOFhip3CqZ4gig32uY7FRjQOAfZpngAU82KzmBSTYmcfChrY1V10EWrOOCPrNu4JE+BsbECspyflcTEWQ7y/H45kh+twnI0NEKNRxP3+FNUu1t014jnhzZux86KL4TvnHDR873ujrtvj8cCjCLD8fv+oz8kXqUGA6utk+NEMim1iN2B35M7Y1MBOoPefKGJS6kuj0RFKlp3OP7nMi6WaWftQ0MDOWVc3IsWajrJ58yD4/Qh+vhZls2cBAAaXrwAEAWVz5qZ9jnfWLMDlwuCHy1F90okAgPDWbYjt2YvyefPYduFNm7DjwotQc/rX0PSD7+f9moxCsvTgyJyYIXfFUvNEIeEymBTT+08UMyPqS4cHdrZS7KSxYtkCu8T3m2pmrY8tauw806ah4sgjsfeWmxFcswZDK1ei/fbbUb14MeuIjba3Y8tXFiO4Zg0AwFFVBd9ZZ6L97rswuHwFgp+vxd6f/ARl8+ahLBnYhTZuxI4LLkTFEQtRf+GFiHV2Jn6GqYNWQLpo8jROjMHMcTMEdtTFZQ68e6RJsSgIQCwGgN5/ojjJ1jgEyIqdHQIhap4oLmxhdwIAzff8CvtuvwM7L7woaVB8Isb+9Cfs72I0hsi2bRCC8kV+zE03geN57L72WoiRCCoXHYGxt9zC/h74z2uI9/TA/68X4f/Xi+xx1/jxmL70TXNeWI5QKnYkkt3JaF2xdCIyFtYdqPh3UF7o7HBhIwi1cDwPuFxANJp18grsoNhJqdjBHOxOqGbZ8tgmsHP4fGj+za8z/t09oRkzN6xPeYz3eDD2lltSgjkljVdfhcarr9J1nUYhpWKpeUKGBbnRKMR4HJzDkfJ32W7A+idWO5NurFiK7yIFdkSRwrtcEKLRtGPF7GT3w5WpUOzoGmR5bJGKJUixS4cyyE2XjiXFzhxkuxNFYCcFeRwHOG1z/0gQqshmeWKv5onEaM6c7E5s8HpKHQrsbAIpdiNRBrnp0rF2OrHaGd4zsitWGVRzHFeQdRGE0WQbK2bH5gnqii0OKLCzCVLnJyl2MhzPyyfWrIodnYiMRKq5UaZfBUrbECVA8Sh2yVmx2XzsyJfSNlBgZxMkrzae7E5SYLUhabzs7FTjYmek4C1d8wRdBIhihjUO2V2xk7pis6Zi6XxqFyiwswnSdAXOTYGdEik1LaaZPiEP4aYTkZHIzRNyATmppUQpICt2aZonbKTY5dQ8QedT20CBnU0QqXkiLdlMikVqzzcFKXgT0yh2NM6NKGYyTV0B7KbYjT4rlnxB7QMFdjaBNU9QKjYFNlYsXfOEjU6sdoZPZ3dCFwGiBCgWxY7Nis3JoJiuQVaHAjubwOxOaPJECtL7kU6xoztMc2DNExFF84RUaE3NE0QRk60rVoja58aSrxhdsSO7E/tAgZ1NkLo+SbFLRZqdK6ZrnrDRHbOdYc0TpNgRJQYrQ0jbPGGf809Os2JZV6z1A9VSh5xDbYKsgJBip0SanZu2eYKCC1NIb1Bsn4saQWgla42djW4smd1JcopGOpXRDufTJz/cjofe2YrOgTBmjqvGbafNwrwWX9ptN7YH8NvXNuKztn609QVx81dbccmiKSnb3Pv6Rtz/5qaUx6Y2VmDp9ccAAPqGIrj39Y14b1MX2vqCqK9w48RZY3Hdifuh2lu4AJgCO5sgMh87UuyUsOaJYLrAzj4nVjvDSQbFaRU7ursnihc+m4+djWp8JcUOSNTZObIEdla1O3lx9R7c8dJ63HHGbBzU4sPjH2zD+Y+twNIbjkFD5cjrZjASx8T6ciyeMw63v7Qu4373G1OJpy49nP3u5OVEZ7s/jHZ/GD9ZPBMzxlSirTeIn/7jc7T7Q/j9tw/R9wWqgAI7myDVL/HUFZtCVrsTCi5MgV3c0nXFUo0dUcSw+tJhs2JFQQBiscQ2Fg2ElHBud2L0XyyWCOyqq0dsY3XF7tH3t+Gb81twzqEtAIA7Tz8QSzd04G8f78IVx0wfsf3cFh/mJtW8u/+9IeN+HTyPpqr01939x1bhD9+RA7hJ9RW44cT98YNnVyEWF+B0FKbajQI7myA3T9CFUolsd0JdsYVCrrFTBnbJzyvZnRBFTKYaO2WgZ5fzD19eDsHvhzCYvs6ONYOYeA0KBALw+/3sd4/HA0+a40diAj5v68cVx0xjj/E8hyOmN2Dljr681rC9axDz73wDHhePgyfW4kcnH4BmX1nG7QOhKCq9zoIFdQA1T9gGuXmCFDslst0J1dgVCsn+gOxOiFJDunEZPnkiJbCzyXdgtHmxhShtaW1tRU1NDftZsmRJ2u16hyKIC+KIlGtjpQedAyNv+nNl3kQffn32XDxx8XzccfqB2NUzhHP+8CEGwrG02/cMRvDg0s341vwWzcfUA1LsbAI1T6RHtjvJ3BVr1ZqQYiFd8wRZzRClQKZZscrfOac9LrPyvNhMgZ353+l169ahubmZ/Z5OrTOSY/dvYv8/cxwwr8WHRXctxctr9uAbh01M2TYQiuKiP32E6U2V+P4J+5m6zuHY4xNHkN1JBmS7E1LsCkXW5gkqHSCKmEwGxUyxc7nA8fZIjPGjjBVjdicmlldUVVWhOk2933Bqy91w8By6hqlznQNhNKZpnNBKTZkLUxorsL07NfgdCMdwweP/RaXHgYe+cwhcBUzDApSKtQ1MsaNUbAqS3YlAqdiCkbZ5IkyNK0Txk8nuRB6pZ5/PvzxWbGRgJ8ZigCAAsOZ32u3kMbu5Bss2d7HHBEHEss3dOHiST7fjDIZj2NE9hKYqOVgMhKL4zmMr4HLwePT8w+B1OXQ7nlZIsbMJTLEjBSQFvizpY5c2FUuBnRlIqpzy4kZBNVEKZFTsbNi4xZUnFbs0qVilGm/V0pZLF03B9X9fjQMn+DCvpQaPvb8dQ5EYzj4kUe923bOrMKbGixtPPgBAouFiU0cAABCNC2j3h7B2Tz8q3E5MbqgAANz58jocP3MMmn1l6AiEcO/rm+DgOZw2dzwAKaj7L0LROO77zjwEwlEEwonPQn2FBw6eM/ttAECBnW2gEU3p4bIodoINT652RLq4CeGRgZ1VLwIEoQeZRorZyZxYQp4Xmz2ws+o16NS549EzmDAM7gyEMXN8NZ64eD4ak+paW18QHCcHWu3+EE554H32+8PvbsXD727F4VPq8OzlCwAAe/tDuOaZT9E3FEVdhRuHTq7FC1csRH0yvft5mx+rdvUBAI6+5+2U9bz3o2PRUldu4CvODAV2NkCMxwHpREGp2BSyjhQjg2JTYF2xaXzs6L0nipmMdic2vKmUU7EjAzvW9cvzlm4GuWDhZFywcHLav0nBmkRLXTm233VK1v397tyDs/59wbT6UfdRCKjGzgYoGwMoFZuKZP+StnnChnfNdoRP1zzB0uD0eSWKl4xdsTY897B5sWmaJ+gm2V5QYGcDBIUSQopdKswcN5zNoJhORkbCpWmeEMKk2BHFT6aRYrZU7CqSil0ag2JmOE7fZ1tAgZ0NYHNibdQ6bxZZFTtKB5oCC64Vpqz03hOlgNwVm97uxE6ffy6L3Ql9n+0FRQk2gI0TI7VuBLJBMc2KLRTsZB+NJmZkgt57ojTI2DxhR8VOqrHLGtjZ5/WUMhTY2QBpXBZH5sQj4DMYFKf4Ltno5GpHlHV00gWAdcVSTShRxLCO8OE+djZU7FhXbBa7E55qZm0BBXY2QKpd4mmc2AgkFXN4jZ0dfJeKBV5xFy99VplLPb33RBEz2kgxO33++XKpeSJzYGen11PKUGBnA+RULN0tDUdShEYodjYcwm1bXC4g6Q/FFDsyhyZKgEwGxXb00GSp2DTNEzT72V5QYGcDpFQsKXYjkRQ7MRJh9V3S74kNOMDCvkvFAMdxI0yK6UJAlAJSx30xNE9kmxUrUpe7raDAzgZQ80RmOEWwm8kgV+k2ThgDGyuWtEVgvldkNUMUMaOmYu2o2GWpsaPAzh5QYGcDZMWOUrHD4RXpaWVnLClG5sINMylmFwIPvf9E8VJMI8W4MuqKLRYosLMBUsBCit1IOKeTpVpTFTtJMaITkRnwUkpqWPMENa4QxUxRKXYVWQK7KHW52wkK7GyANAeVp+aJtKQzKabUgbnIqdhhih29/0QRo5wVK4oie9yOih0bKTY0lFKvDNAUH7tBgZ0NECQfO/IQSks6yxM7nljtjLJ5QhRFCuyIkoAp0qIIxGLscTtmDKTADkjjMkDfZ1tBgZ0NEMnuJCvpLE+oJsRcUhQ7pdUMpW6IIkYZ6Cg7Y+UbS/ucfzivl9kWDW+gEMiX0lbYJrCL9/Wh7YYf4otDDsUXh83Hnp/+FMLgYNbnCOEw9v3iF9h4+Jew4eBDsPvqaxDr6kq7bay3F5uOPgbrD5iJuN9vxEvQDNmdZIcpdiGlYkd3mGYip6TCrHEl8Ti9/0TxolTklHV2dlS4OJ7POC+WKZA2ej2ljG0Cu7Yf/gjhzZsx8fHH0PKH32Po44+x95afZ31O+5IlCLz1Nprvvw+TnnwSsY4O7L76mrTb7v3ZzfDsv58RS88bsjvJDlPswiMVO55qQkxBGjUkhsMpFzg7paIIQjVOp6xypQvsbPb5z2R5YsdAtZSxRWAX3rIFg++9h3G3346yuXNRfsghGPuzn8H/yiuItnekfU48EEDfc89jzI03ouJLX0LZ7FkYt+SXCH76KYKrVqVs2/vMMxD8ftRffLEJr0Y9TLGjVGxaZMUuTSrWZidWuyKlXIVIRA7sXC5wvC1OMQShCaU5t3L6hJSKtVtXODMpzhjY0fnUDtjirBtctQp8dTXKDpzNHqtYsADgeQTXrE77nNDatUA0ioqFC9hjnqlT4Rw/DkOKwC68eTM6//d/Mf7uuwDOmm8HU+woFZsWKeClrtjCwS5u4YhCLaWLAFH8pLM8Ud7c2AnWGTsiFUt2J3bCFrOWYp1dcNbVpTzGOZ1w1NQgnqlmrrMLnMsFR3V1yuPO+gb2HCESQdv1N2DMD38I1/jxiOzandN6wuEwwooOzEAgoOblqEZkPnb0pUqHFPCmKHbUFWsqSoNi2ZyYPq9E8cMCu6gisLOrYkep2KKgoIFdx29+g+5HHs26zdRXXjbs+J2/+S0806ai5rTTVD1vyZIluO222wxa1UiE5Jgmap5ID8cUOznYpskT5iKrFmHqoCNKirSpWJuWgvDl6ZsnpGsQ+djZg4IGdnUXXYSaM87Iuo17wgQ4GxsQ6+lJeVyMxRDv74ejoSHt85yNDRCjUcT9/hTVLtbdxZ4zuGIFwhs3wv+fZIo3aTC5ccFCNFx+ORqvuTrtvm+66SZcd9117Pe2tja0trZmf7F5QHYn2ZECXiFN8wQFF+bAmieUih2990QJwMaKpVHs7PYd4CTFbnC4YmfP11OqFDSwc9bVjUixpqNs3jwIfj+Cn69F2exZAIDB5SsAQUDZnLlpn+OdNQtwuTD44XJUn3QiACC8dRtie/aifN48AMCEB+5PSd+FPvsce3/6U0x66s9wT5yYcT0ejwceRZrJb7A9ipSK5akrNi3pFDs6EZmLbFAcpveeKCmU0yckbKvYZZgXSzdr9sIWNXaeadNQceSR2HvLzRh3660QYzG03347qhcvhmtMEwAg2t6OnRdehPF334WyOXPgqKqC76wz0X73XXDU1ICvrET7HXegbN48lCUDu+HBW7y3jx1veG1eIWGpLUrFpkVS7NLZndjtxGpXZIPiqKLGji4CRPGTrXnCboGQXGOX6hFr19dTqtgisAOA5nt+hX2334GdF14E8DyqTjwRY3/6E/Z3MRpDZNs2CEH54j7mppvA8Tx2X3stxEgElYuOwNhbbinE8vNCVuwoFZuOtAbFdCIyFaZahMMQpZpQqschSgDpc54S2Nk0FTtaVyzdrNkD2wR2Dp8Pzb/5dca/uyc0Y+aG9SmP8R4Pxt5yS87BXMXh80fswwqQYpcdviyLYke+S6Ygm0SHKagmSgq5K7aImieGMtid0HfaFljTuI1IgRS77Mh2J8qRYva8Y7YrnNQ8EaXmCaK0YM0TxaDYZbQ7oU53O0GBnQ2QFTsK7NKRzaCY7jDNQW6eiMhWM/R5JUoA9tlXBHZC1J6KXaZZsWQfZS8osLM4oijKBsWUik0LR3YnBUdunohADNN7T5QO6ZonYNPOcL68AkA6xc6er6dUocDO6kSjgCAAoFRsJvh0difsjplORGaQ2jxB9Y1E6ZDOoFiQUrE2U+zYrNggTZ6wMxTYWRxBMbqMIx+7tLCuWIViJ1BwYSq8hwyKidIkrd2JXWvsKmikWDFAgZ3FYXVjHEdfqgywNGAwnY8dvWdmINcZham+kSgpmFqdDOZEQUhkWmC/QIjZnVBXrK2hwM7iKBsnOI4r8GqsCZ9GsbPrHbNdYV2x4YicBndT6QBR/HDDfOyUtid2S8Wma54QRREizX+2FRTYWRxmdUIdhhmRmidopFjhkIxLxUhEvhmh954oAYaPFEsJ7Gz2HUhrd2Lj11OqUGBncdhFkurrMpLN7oRq7MyB1RmRQTFRYgyvsVPW2nFO28wAAKAI7IJBiKKY+P8IBXZ2gwI7i8MkcOqIzYjcPEEjxQpFavMEqaVE6SDVnUklCEyxc7nA8fa6xEqBHWIxuWYwqghU6TttC+z1qStB5FQsKXaZYEFFKMTuMu060seuKE1aWaE1zZUkSoBMip0dzz1S8wQAiMl0LFMgnU5wDkchlkWohAI7iyONyaJUbGaU783wkyt1cZlDqkEx1dgRpQNrnmAKV+K/vA0DO87plG/ShgV29H22D/YqAChBpMH21DyRGeV7I4ZCgMdDXbEmw97nWAyCNCmF3nuiBJDqeIXhNXY2re/ly8oQj0RYZyy7SbZBoPrkh9vx0Dtb0TkQxsxx1bjttFmY1+JLu+3G9gB++9pGfNbWj7a+IG7+aisuWTQlZZt7X9+I+9/clPLY1MYKLL3+GPZ7KBrHnS+vx4tr9iASE3DUjEbcfvpsNFYV7ppNgZ3FIcUuB1wuwOEA4nEIoTAcNXSXaTZKZVQYGABA7z1RGmSyO+Ft6qHJlZcD/f0Qkl52dlHgX1y9B3e8tB53nDEbB7X48PgH23D+Yyuw9IZj0FA5MsgKRuKYWF+OxXPG4faX1mXc735jKvHUpYez353D6iZvf2kd3trQgf8992BUeV245V+f43+e+gTPfW+hfi9OJZSKtThMsaPmiYxwHCfX2SXfLwrszIVLCewCicdsemEjCDUMHylm93PPcMsTu7yeR9/fhm/Ob8E5h7Zgxpgq3Hn6gShzO/C3j3el3X5uiw8/WTwTp80dD7cjcyjk4Hk0VXnZT12F/D74Q1H87eNd+NlXW7FwegMOnFCDe74+F5/s6MXKnb26v8ZcIcXO4shpLQrsssF5vcDQEHu/BErFmovTCfA8IAiIB5KKHTVPECVAMTVPACPnxQoFDOwCgQD8fj/73ePxwJOmLCkSE/B5Wz+uOGYae4znORwxvQErd/TltYbtXYOYf+cb8Lh4HDyxFj86+QA0+xLv0ee7+xGNizhiegPbfnpTJZp9ZVi5oxcHT6zN69haIcXO4oghsjvJBen9EcPh1JE+Nj252g2O41gDhRBIKnYUVBMlwIiRYja/qZQUO7krNvl6ClDn3draipqaGvazZMmStNv1DkUQF8QRKdfGSg86B8Jpn5ML8yb68Ouz5+KJi+fjjtMPxK6eIZzzhw8xEI4BADoHwnA7eNSUpV5nGirdeR03X0ixszhCmOxOcoFn0ydCtnZ+tzOc2w0xGIQwOAiAOpKJ0mBEjV3RKHapzROFOJeuW7cOzc3N7Pd0ap2RHLt/E/v/meOAeS0+LLprKV5eswffOGyiqWtRAwV2Fkek5omcYCbFoTAFdgWCd7shKH6n954oBTKNFLPr55+vkGrshgd25geqVVVVqK6uHnW72nI3HDyHrmEqWedAGI1pGie0UlPmwpTGCmzvTqiZjZUeROIC+oPRFNWuayCi63HVQqlYi0PNE7khNU8IoWDqSB+b3jXbkeEXskKkbgjCbPgiq7HjJMWOpWITwZKVFXi3k8fs5hos29zFHhMEEcs2d+PgST7djjMYjmFH9xCaklYmsyfUwOXgUo67pXMAbX1BHDypMPV1ACl2lofZnVAqNiuSYieG5FmldhzpY2eGB3J2VSwIQg0jmifsrtiVSfNih3XFWrzL/dJFU3D931fjwAk+zGupwWPvb8dQJIazD2kBAFz37CqMqfHixpMPAJBouNjUkagHjsYFtPtDWLunHxVuJyY3VAAA7nx5HY6fOQbNvjJ0BEK49/VNcPAcTps7HgBQ7XXhnENbcMfL61FT7kKVx4Wf/+tzHDzRV7DGCYACO8vDRoqRYpcVpd2JnQw1i4kRip1NL2wEoYbhgZ1gc8VuuN1JIbti1XDq3PHoGYzg3tc3ojMQxszx1Xji4vnMKLitLwiO49j27f4QTnngffb7w+9uxcPvbsXhU+rw7OULAAB7+0O45plP0TcURV2FG4dOrsULVyxEvSLNevNXW8Fz6/G9p1YmDIr3a8Dtp8826VWnhwI7iyMNtifFLjspNXY2P7HaleH2Jla/wycIPWCBnVTba3fFrjyRihWHN0/YoLTigoWTccHCyWn/JgVrEi115dh+1ylZ9/e7cw8e9ZhelwO3nz674MGcEspTWRzm+k2KXVZ470jFzq4nVrvCD/Na5MnHjigBpBtIyTvTLgpXJphiNzjM7sSmr6cUocDO4sipWFLssiEpmoLC7oROROZCqViiFGGf82gUoiDYPmPAZbQ7sefrKUUosLM4ciqWFLtsMINiZSqWAgtToeYJohRRfs7FaNT2N5Zy84S9ZsUSMhTYWRxS7HJDMigWwiHbp0LsSsr7zfPgnFTCSxQ/KYFdJCKnLm2q2GWaFWtluxMiFQrsLA41T+QGVzbS7oQCO3NR1tTRe0+UCsoALlWxs2tgJzVPJAO7KJ1P7QYFdhaH7E5yg40UC4dsf8dsV5QnfiodIEoFjueBpDqdUOzsXWM3vHmCMiD2gwI7i0M1drkh1dgJpNgVDE7RFWtXtYIgtKD0srN/jV2m5gm6BtkFCuwsjqTYUSo2OzybPBGyfSrErigvZDx52BElhGSGrlTs7FqTNrLGzt6BailCgZ2FEUWRdSRRKjY7zO4kTIpdoeCoxo4oUdIpdrBpKpZLBnZiOAwxHld0xdrz9ZQiFNhZGOkLBciTFYj0MIPiYND2d8x2hVeUC1DpAFFKpAR2Nj//SKlYABCCZPhuRyiwszBSGhZIvWgSI0lR7KQuLkoHmkpK8wRdBIgSQjlWzO7NE5zHA/CJ0EAYGrR9oFqK2Cawi/f1oe2GH+KLQw7FF4fNx56f/hTC4GDW5wjhMPb94hfYePiXsOHgQ7D76msQ6+oasV3f8y9g62lfw4Y5c7Fx4RHY94tfGPUyVCE1TsDhsO1JwiyYYhcKkVN6gUhtnqCLAFE6sLFiRdA8wXEcU+2UGRC7vp5SxDaBXdsPf4Tw5s2Y+PhjaPnD7zH08cfYe8vPsz6nfckSBN56G83334dJTz6JWEcHdl99Tco23X/8Ezrvuw/1l12GqS+9iIl/fBwVixYZ+VJyhlmdkFo3KlKqWgiH5fZ8UuxMJVWxo6CaKB3SpWLtfDOubKAQpAwIXYdsgy2s4cNbtmDwvfcw+e9/R9mBswEAY3/2M+z67uVo+tGP4BrTNOI58UAAfc89j+Z77kHFl74EABi35JfYuvgUBFetQtm8eYj396Pz/vvR8vv/RcWCBey53v33N+eFjYIQShatUn3dqEgnnVTFjgI7M6HmCaJUkQM7+48UAwCuXLY8EelG2XbYQrELrloFvrqaBXUAEoEYzyO4ZnXa54TWrgWiUVQslAM2z9SpcI4fh6FVqwAAg8uWAYKAaHs7tiw+BZuOPga7v/8DRPfuNfT15IoYkQI7ulMaDd6rrLGz/4nVjihrcHjyvCJKiOJT7CoAJEyKxTDdKNsNWwR2sc4uOOvqUh7jnE44amoQT1MzJz2Hc7ngqK5OedxZ38CeE9m1G6IoovuhhzHmppsw4f77EO/vw86LL2FfznSEw2H4/X72EwgE8nyF6ZFTsaTYjYbUPEGKXeFQpmrovSdKCan0oBgMigGlSfEQnU9tSEFTsR2/+Q26H3k06zZTX3nZuAUIAhCNYsxPf4rKRUcAAJp/8xtsWnQkBlf8F5VHpq+1W7JkCW677Tbj1iUtj1KxOcN8/gSBGWvSichcqHmCKFWkNGVKV6yNvwPpmyfsq0CWGgUN7Oouugg1Z5yRdRv3hAlwNjYg1tOT8rgYiyHe3w9HQ0Pa5zkbGyBGo4j7/SmqXay7iz3H2dgIAPBMnyY/r64OjtpaRPfuybimm266Cddddx37va2tDa2trVlfhxbEMDVP5Aqn9F7yJxRUOhGZC9mdEKVKWsXO1qlYuXmC7E7sR0EDO2dd3YgUazrK5s2D4Pcj+PlalM2eBQAYXL4CEASUzZmb9jneWbMAlwuDHy5H9UknAgDCW7chtmcvyufNS+z34IMAAJFt2+AaOxZAwlYl3tsL1/jmjOvxeDzwKIItv98/+ovVACl2ucO5XADHAaKIeDI1bucTqx3hlc0THroIEKUDz3zsIsWh2EnNE0MKxY4EBttgixo7z7RpqDjySOy95WYE16zB0MqVaL/9dlQvXsw6YqPt7djylcUIrlkDAHBUVcF31plov/suDC5fgeDna7H3Jz9B2bx5KEsGdp4pU1B5/PFo/+UvMbTyU4Q2bsSeH98E99QpqDh8fqFeLoMUu9zhOE62PEkG2nY+sdqRlFmx9N4TJYSyeUIoAsVOGismDA4WRc1gqWELuxMAaL7nV9h3+x3YeeFFAM+j6sQTMfanP2F/F6MxRLZtgxCUpzWMuekmcDyP3ddeCzESQeWiIzD2lltS9jv+7rvQvmQJdv3P/4DjOJTPPwwTH3nEEl9KIdk8QXdKucF7PIgHg0yxo+DCXKh5gihVpBq7YjAoBgC+LBHYxfv72WN2fj2lhm0CO4fPh+bf/Drj390TmjFzw/qUx3iPB2NvuWVEMJey38pKjL/zTuDOO3Vbq16IIbI7UQNT7KRULJ2ITIVq7IhShSl24QhQBIqdVGNHgZ09sUUqtlQRwmR3ogYpZS2NmqMTkbmkdsXSzQhROrCRYkPymEs7n3+krtiUwM7GgWqpQYGdhRGpeUIVw98nO59Y7YiyC5nee6KUkD7vwuCQ/JiNAyGpeSLe15d4wOUCx1O4YBfoX8rCsOYJSsXmxPCUtZ1PrHaET6mxo/eeKB1YYDcwID9m4/PP8FQs1SvbCwrsLAyzO6FUbE4MT1mTamQuVGNHlCqyYpdMxdpc4ZJ8QSXFjr7P9sK+n7wSgI0UI8UuJ0YqdnQyMhPO6QQcDgB0h0+UFpJCLSl2dlbrAMWsWLKOsiUU2FkYIUyKnRpIsSs8kuUJWfQQpQRrnkgqdrztA7uylN/pXGovKLCzMGKY7E7UMLJ5wt4nVzsiXdDoQkCUEtLnPS51xdr83CPV2EnQ99leUGBnYeTmCVLscmF4yprSgebDFDtKgxMlBM+aJyTFzt6ff76MFDs7Q4GdhZGbJ0ixy4URKWubp0PsCF9VlfxvZYFXQhDmwQyKg8HE7zY/93DDAju6SbYXtpk8UYrIzROk2OUCX5b6PtHJyHzG/vQnCK75DN7W1kIvhSBMY7iiZXeFS2qekLD76yk1KLCzMNQ8oY7h7xOdjMynYuFCVCxcWOhlEISpFF1gV0bnUjtDqVgLQ3Yn6hj+PtHJiCAIMxieerV9KtbhSGlGo3OpvaDAzsLIih0FdrmQotjxfMJXjSAIwmCKTbEDUhsoiuH1lBIU2FkYSbGjwC43lLYwdCIiCMIsRgR2NlfsAArs7AxJGhZGUuyoeSI3lO9TMZxYCYKwB8PtfYohEOIrZC87u3iCPvnhdjz0zlZ0DoQxc1w1bjttFua1+NJuu7E9gN++thGftfWjrS+Im7/aiksWTcm47/99ezN+9eoXuOiIyfj5qbPY4x2BEJa8sgHvberCYDiGqY0VuOrY6fjKgeP0fnk5Q4qdRRHjcSAaBTDSeJdID+chxY4gCPMpRsWOK5MDO94GWaMXV+/BHS+tx7UnzMDLVy9C67gqnP/YCnQNhNNuH4zEMbG+HDd+5QA0VmV/fat39eEvK3bigLFVI/52/d9WY2vnAB694FD85/tH4eRZY3HlX1bi87Z+XV6XFiiwsyhSGhawx5fKCvBU7EsQRAEY0TxRBOcf5fQJOxiOP/r+NnxzfgvOObQFM8ZU4c7TD0SZ24G/fbwr7fZzW3z4yeKZOG3ueLgdmUOhwXAM3392Fe46cw5qykYG7J/s6MUFCydjXosPE+vLcfXxM1Bd5qLAjhiJlIYFSLHLFWXzhF1SBwRB2J/h55tiUOzsVGMXiQn4vK0fR0xvYI/xPIcjpjdg5Y6+vPZ98z8/x7H7N2HRjIa0fz9kUi1eWrMXfUMRCIKIf63eg3BUwJem1ud13HygGjuLwhonXC5wPMXfuaC0OyFzYoIgzGL4+cbqgVAupCh2BXo9gUAAfr+f/e7xeOBJk8HqHYogLohoqEz9W2OlB1s6BzUf/1+r92Btmx//vOqIjNv87tyDcdVfVmLeL16Hk+dQ5nLgoe8cgskNFRmfYzQUMVgUNk6M1LqcSfFdskHqgCCI4qAo7U7KC6/Ytba2oqamhv0sWbLEtGPv6QviFy+uxX3fnAevy5Fxu9++9gX8oRievvRw/OuqRbjkyCm48i8rsWGfP+NzjIYUO4siRqTAjurrckVZi1gMqRCCIGyC0wlwHCCKAIrj/MNZIBW7bt06NDc3s9/TqXUAUFvuhoPnRjRKdA6E0Vip7Rr6WVs/ugYi+OqD77PH4oKI/27vwZMf7sDGO76C3b1DeOLDHXjtB0dhvzGJxorW8dX4KLnNL884UNOx84UCO4vCpk7QOLGcIad0giAKAcdx4NxuiJKpfBHU+KakYj2FOZ9WVVWhurp61O3cTh6zm2uwbHMXTpo1FgAgCCKW/f/27j6qqjLfA/j3HA7ncBAPR+RNUV58GRVFVExCtLoLrmjexqzVeB1Wqb3MZUYnXXbJtEkqK8zGbuUqp5dRm9Vd0WRqM6XeWAqKs9TSeBFpSA0HM140gwPIy4Hzu38YW46iBQJ7n+33s9ZZS/Z+zt7P/q2zNl+fzfOck9/jgakR3Tp34ohA/N+y29y2pW8txPAgP6TdPhxeRgManW0AAKPB/b1GgwHyY8hXA4OdRl1+FMsRu5+Ls2KJSC0Gb+/LwU4HI3bGjsudeMD99OFpUXjsw0LEDLFjwlB//PnAaVxsacV9cUMBAMs/KECIvw9WzBwN4NKEixPVdQAAZ5sLVY4mHP+uFv3MJkQG9oOfxYRRVyxvYvX2gt3XW9k+PMgPkQN9sWpbMVbNHoMBvt747HgVDpw8j00LbunDq3fHYKdR0swRu67iiB0RqaXjPUcP9x8tTJ7oirtiB+NCQwv+J/trnKtrxpjBNrz74BRljbqzNY0wGC4PrVU5mjD7tcuPWd/a/w3e2v8N4qMC8MF/Jfysc3p7GbF50RS8uOufePjdL9DQ3IaIgb5Yf18s/m10cM9eYBcw2GkUJ090HRcoJiK1uAU7PYzYaWDyRFctmBqJBVMjO913ZVgbGuCL02tnd+n4nQW+qMB++NP9cV06Tm/jrFiNujxix0exP5fBYFDCnafciIhIHzhiR1rBYKdRrvZ17BjsuqR9hFMP/2MmIs/RccKEHu4/brNiuXyUR2Gw0yjh5IluMSojdp5/YyUiz9FxVMsTJhv8lI6TJ9SaFUvdw2CnUS5OnugWZcROBzdWIvIcHUfp9HD/6fgoVg9B9WbCYKdRHLHrnvYRO96IiKgvGb05eYK0gcFOo7jcSfdwxI6I1MDJE6QVDHYaxeVOusfIWbFEpALdLXeiga8Uo+5hsNMoZcSOj2K75PKsWN6IiKjvcMSOtIILFGuUMmLHR7FdYrRyuRMi6ntukyd0cP8xmM2wjByJttpaeAUEqN0d6gIGO41SvnOQI3ZdYpv9H2j59iz6TZumdleI6CaitxE7AIj8aCvgdHIymofxmGDXVlODyueeR31ODmA0ov+Mf0foqlUw9ut3zfe4mptR/eKLcHy6Ey6nE36JiQjNWA1TYKDSpvHYMVSvfxlNx48DBgOsMTEITv9v+Iwe3ReXdU1c7qR7bDNTYJuZonY3iOgmo8dgZzSbAZ1cy83EY/7G7mz642g+eRLhm/6MoX/aiItHjqBidcZ131OVmYm6nFyEvfoKIv7yF7RWV+Pb3z+q7Hc1NODMw4/Ae9AgRH7wASL/9z0Y+/VD+cOPQJzO3r6k6+JyJ0REnkNvkyfIc3lEsGs+dQoNeXkYtGYNrLGx8I2LQ+gf/gDHzp1wVlV3+p62ujrUfLQNIStWoN+tt8I6biwGZb6Axvx8NBYUXDruN2Voq61F0KO/h2VYFCwjRyJw8WK0nT8P53ff9eEVXk2a2idPcMSOiEjr3L5SjKNcpCKPCHaNBQUw2mywxoxTtvVLSACMRjQWFXb6nqbjxwGnE/2mJijbLMOGwTR4EC7+GOzMUVHwsttRs/UjSEsLXE1NqPloK8zDh8M7LOya/WlubobD4VBedXV1PXOhHbiaOXmCiMhTcMSOtMIjgl3rufMwXTErx2AywcvfH23nz1/zPQZvb3jZbG7bTQMDlfd4+fVD+F/eRe3f/45/TpiI0klxaMg7gPC33oTBdO0/P8zMzIS/v7/yio6OvsErvNrlETs+iiUi0jq9faUYeS5VJ09Ur1+P799+57pthu38tNfO72pqQsUfnoLvxIkYsP6PQFsbvt+0GWfS0hD54YfXfAy6cuVKLF++XPn57NmzPR7uLo/YMdgREWmdkSN2pBGqBruARYvgP3fudduYhwyBKSgQrRcuuG2X1tZL6+t0mOHakSkoEOJ0os3hcBu1a/3+vPIexyefwHn2LCKz3ofBeGnwMuyPL6E0/lbU7dkD/9mzOz22xWKBpUPgcjgcP32xXdQ+YsdgR0SkfXwUS1qharAzBQRc9Yi1M9YJE+ByONBYfBzWcWMBAA2HDgMuF6zjYzt9j8/YsYC3NxoOHoItZQaAS5MlWr+rgO+ECQAAV2MTYDQABsPlNxqNl352yY1d3A1qH7Hj5AkiIu1Tgp3JpAwUEKnBIz59luHD0W/6dFSsfgqNRUW4+OWXqFqzBrY774R3SDAAwFlVhVOz7kRjUREAwKt/f9jvvQdVL65Fw6HDaCw+jopVq2CdMAHWH4Ndv8SpcNU6UPnss2g+dQrNJ07gu1WrYPDygm/8FLUuFyLSYcSOwY6ISOvav8aQf19HavOYBYrDXlqHyjXPoXzhoh8XKJ6B0CdXKfvF2YqWsrJLo3A/Clm5EgajEd8uXQppaYHftESErl6t7LcMG4YhG9/A+dffwOn/nA8YjfAZMwbhb78F7+DgPr0+N04n4HIB4OQJIiJP0L7ciZGPYUllHhPsvOx2hK3/4zX3m4eEYcw/v3LbZrRYELp6tVuYu5JfYiL8EhN7rJ89of0xLHD5S+2JiEi72kfsYGawI3V5TLC7mRi8vRG8YgWkuYnD+kREHqD9Xm305j2b1MVgp0FGHx8MXLRQ7W4QEdHP1B7sOCOW1OYRkyeIiIi0zDx0CADAOzxc5Z7QzY4jdkRERDfIHBGB4bt3wRQUpHZX6CbHYEdERNQDzJGRaneBiI9iiYiIiPSCwY6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsiIiIiHTCpHYH9MDlcgEAKioqVO4JERGRZ2v/Xdr+u5W6hsGuB1RVVQEApkyZonJPiIiI9KGqqgrh4eFqd8PjGERE1O6Ep2ttbUV+fj5CQkJgNPbM0+26ujpER0ejpKQE/fv375Fj0iWsbe9hbXsPa9t7WNve053aulwuVFVVYeLEiTCZOP7UVQx2GuVwOODv74/a2lrYbDa1u6MrrG3vYW17D2vbe1jb3sPa9j1OniAiIiLSCQY7IiIiIp1gsNMoi8WCjIwMWCwWtbuiO6xt72Ftew9r23tY297D2vY9/o0dERERkU5wxI6IiIhIJxjsiIiIiHSCwY6IiIhIJxjsNOj1119HZGQkfHx8EB8fj88//1ztLmnO/v37cdddd2Hw4MEwGAzYsWOH234RwerVqzFo0CBYrVYkJyfjxIkTbm0uXLiA1NRU2Gw22O12PPTQQ6ivr3drU1RUhOnTp8PHxwdDhw7FunXrevvSVJeZmYlbbrkF/fv3R3BwMO6++26Ulpa6tWlqasLixYsxcOBA+Pn54d5771W+gaVdeXk5Zs+eDV9fXwQHByM9PR2tra1ubXJzczFp0iRYLBaMGDECW7Zs6e3LU9XGjRsxfvx42Gw22Gw2JCQkYNeuXcp+1rXnrF27FgaDAcuWLVO2sb7d8/TTT8NgMLi9Ro8erexnXTVGSFOysrLEbDbLpk2b5Pjx4/LII4+I3W6XqqoqtbumKTt37pQnn3xStm3bJgBk+/btbvvXrl0r/v7+smPHDiksLJRf/vKXEhUVJY2NjUqbmTNnSmxsrBw6dEjy8vJkxIgRMn/+fGV/bW2thISESGpqqhQXF8v7778vVqtV3nzzzb66TFWkpKTI5s2bpbi4WAoKCuTOO++U8PBwqa+vV9qkpaXJ0KFDZc+ePXLkyBG59dZbZerUqcr+1tZWGTdunCQnJ0t+fr7s3LlTAgMDZeXKlUqbb775Rnx9fWX58uVSUlIiGzZsEC8vL9m9e3efXm9f+tvf/iaffvqpfP3111JaWiqrVq0Sb29vKS4uFhHWtad8/vnnEhkZKePHj5elS5cq21nf7snIyJCxY8dKRUWF8jp37pyyn3XVFgY7jZkyZYosXrxY+bmtrU0GDx4smZmZKvZK264Mdi6XS0JDQ+Wll15SttXU1IjFYpH3339fRERKSkoEgHzxxRdKm127donBYJCzZ8+KiMgbb7whAwYMkObmZqXNihUrZNSoUb18RdpSXV0tAGTfvn0icqmW3t7e8uGHHyptvvrqKwEgBw8eFJFLwdtoNEplZaXSZuPGjWKz2ZR6Pv744zJ27Fi3c82bN09SUlJ6+5I0ZcCAAfLOO++wrj2krq5ORo4cKdnZ2XL77bcrwY717b6MjAyJjY3tdB/rqj18FKshLS0tOHr0KJKTk5VtRqMRycnJOHjwoIo98yxlZWWorKx0q6O/vz/i4+OVOh48eBB2ux2TJ09W2iQnJ8NoNOLw4cNKm9tuuw1ms1lpk5KSgtLSUvzwww99dDXqq62tBQAEBAQAAI4ePQqn0+lW39GjRyM8PNytvjExMQgJCVHapKSkwOFw4Pjx40qbjsdob3OzfNbb2tqQlZWFhoYGJCQksK49ZPHixZg9e/ZVNWB9b8yJEycwePBgDBs2DKmpqSgvLwfAumoRg52GnD9/Hm1tbW4ffgAICQlBZWWlSr3yPO21ul4dKysrERwc7LbfZDIhICDArU1nx+h4Dr1zuVxYtmwZEhMTMW7cOACXrt1sNsNut7u1vbK+P1W7a7VxOBxobGzsjcvRhGPHjsHPzw8WiwVpaWnYvn07oqOjWdcekJWVhS+//BKZmZlX7WN9uy8+Ph5btmzB7t27sXHjRpSVlWH69Omoq6tjXTXIpHYHiEi7Fi9ejOLiYhw4cEDtrujGqFGjUFBQgNraWmzduhULFizAvn371O6Wxztz5gyWLl2K7Oxs+Pj4qN0dXZk1a5by7/HjxyM+Ph4RERH461//CqvVqmLPqDMcsdOQwMBAeHl5XTWbqKqqCqGhoSr1yvO01+p6dQwNDUV1dbXb/tbWVly4cMGtTWfH6HgOPVuyZAk++eQT5OTkYMiQIcr20NBQtLS0oKamxq39lfX9qdpdq43NZtP1Lwuz2YwRI0YgLi4OmZmZiI2Nxauvvsq63qCjR4+iuroakyZNgslkgslkwr59+/Daa6/BZDIhJCSE9e0hdrsdv/jFL3Dy5El+bjWIwU5DzGYz4uLisGfPHmWby+XCnj17kJCQoGLPPEtUVBRCQ0Pd6uhwOHD48GGljgkJCaipqcHRo0eVNnv37oXL5UJ8fLzSZv/+/XA6nUqb7OxsjBo1CgMGDOijq+l7IoIlS5Zg+/bt2Lt3L6Kiotz2x8XFwdvb262+paWlKC8vd6vvsWPH3MJzdnY2bDYboqOjlTYdj9He5mb7rLtcLjQ3N7OuNygpKQnHjh1DQUGB8po8eTJSU1OVf7O+PaO+vh6nTp3CoEGD+LnVIrVnb5C7rKwssVgssmXLFikpKZHf/OY3Yrfb3WYT0aWZb/n5+ZKfny8A5OWXX5b8/Hz517/+JSKXljux2+3y8ccfS1FRkcyZM6fT5U4mTpwohw8flgMHDsjIkSPdljupqamRkJAQuf/++6W4uFiysrLE19dX98ud/Pa3vxV/f3/Jzc11W97g4sWLSpu0tDQJDw+XvXv3ypEjRyQhIUESEhKU/e3LG8yYMUMKCgpk9+7dEhQU1OnyBunp6fLVV1/J66+/rvvlDZ544gnZt2+flJWVSVFRkTzxxBNiMBjks88+ExHWtad1nBUrwvp212OPPSa5ublSVlYm//jHPyQ5OVkCAwOlurpaRFhXrWGw06ANGzZIeHi4mM1mmTJlihw6dEjtLmlOTk6OALjqtWDBAhG5tOTJU089JSEhIWKxWCQpKUlKS0vdjvH999/L/Pnzxc/PT2w2myxatEjq6urc2hQWFsq0adPEYrFIWFiYrF27tq8uUTWd1RWAbN68WWnT2Ngov/vd72TAgAHi6+src+fOlYqKCrfjnD59WmbNmiVWq1UCAwPlscceE6fT6dYmJydHJkyYIGazWYYNG+Z2Dj168MEHJSIiQsxmswQFBUlSUpIS6kRY1552ZbBjfbtn3rx5MmjQIDGbzRIWFibz5s2TkydPKvtZV20xiIioM1ZIRERERD2Jf2NHREREpBMMdkREREQ6wWBHREREpBMMdkREREQ6wWBHREREpBMMdkREREQ6wWBHREREpBMMdkREREQ6wWBHREREpBMMdkTk8RYuXIi7775b7W4QEamOwY6IiIhIJxjsiMhjbN26FTExMbBarRg4cCCSk5ORnp6Od999Fx9//DEMBgMMBgNyc3MBAGfOnMGvfvUr2O12BAQEYM6cOTh9+rRyvPaRvmeeeQZBQUGw2WxIS0tDS0vLdc/Z0NDQx1dORPTzmNTuABHRz1FRUYH58+dj3bp1mDt3Lurq6pCXl4cHHngA5eXlcDgc2Lx5MwAgICAATqcTKSkpSEhIQF5eHkwmE5577jnMnDkTRUVFMJvNAIA9e/bAx8cHubm5OH36NBYtWoSBAwfi+eefv+Y5RUTNUhARXRODHRF5hIqKCrS2tuKee+5BREQEACAmJgYAYLVa0dzcjNDQUKX9e++9B5fLhXfeeQcGgwEAsHnzZtjtduTm5mLGjBkAALPZjE2bNsHX1xdjx47Fs88+i/T0dKxZs+a65yQi0iI+iiUijxAbG4ukpCTExMTgvvvuw9tvv40ffvjhmu0LCwtx8uRJ9O/fH35+fvDz80NAQACamppw6tQpt+P6+voqPyckJKC+vh5nzpzp8jmJiNTGYEdEHsHLywvZ2dnYtWsXoqOjsWHDBowaNQplZWWdtq+vr0dcXBwKCgrcXl9//TV+/etf98o5iYjUxmBHRB7DYDAgMTERzzzzDPLz82E2m7F9+3aYzWa0tbW5tZ00aRJOnDiB4OBgjBgxwu3l7++vtCssLERjY6Py86FDh+Dn54ehQ4de95xERFrEYEdEHuHw4cN44YUXcOTIEZSXl2Pbtm04d+4cxowZg8jISBQVFaG0tBTnz5+H0+lEamoqAgMDMWfOHOTl5aGsrAy5ubl49NFH8e233yrHbWlpwUMPPYSSkhLs3LkTGRkZWLJkCYxG43XPSUSkRZw8QUQewWazYf/+/XjllVfgcDgQERGB9evXY9asWZg8eTJyc3MxefJk1NfXIycnB3fccQf279+PFStW4J577kFdXR3CwsKQlJQEm82mHDcpKQkjR47EbbfdhubmZsyfPx9PP/30T56TiEiLDMJ5+0R0k1q4cCFqamqwY8cOtbtCRNQj+CiWiIiISCcY7IiIiIh0go9iiYiIiHSCI3ZEREREOsFgR0RERKQTDHZEREREOsFgR0RERKQTDHZEREREOsFgR0RERKQTDHZEREREOsFgR0RERKQTDHZEREREOvH/Gt5Fujz+dqgAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_training(_loss_test, _rewards_test, num_steps, 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "compute_avg_return_episodic(test_env, agent.policy, num_episodes=50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Save to checkpoint."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train_checkpointer.save(global_step)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# WandB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ERROR:wandb.jupyter:Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Create sweep with ID: srh8mikx\n",
      "Sweep URL: https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/sweeps/srh8mikx\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[34m\u001b[1mwandb\u001b[0m: Agent Starting Run: 86wuplng with config:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tepsilon_greedy: 0.7\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tgamma: 0.1\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tlearning_rate: 0.009081865474555018\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \ttarget_update_period: 100\n",
      "ERROR:wandb.jupyter:Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: Currently logged in as: \u001b[33mvaleriia-sakovskaia\u001b[0m (\u001b[33mcold-machines\u001b[0m). Use \u001b[1m`wandb login --relogin`\u001b[0m to force relogin\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "Tracking run with wandb version 0.15.12"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Run data is saved locally in <code>/home/vsakovskaya/gc-ml/wandb/run-20231031_074257-86wuplng</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Syncing run <strong><a href='https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/runs/86wuplng' target=\"_blank\">rare-sweep-1</a></strong> to <a href='https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>Sweep page: <a href='https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/sweeps/srh8mikx' target=\"_blank\">https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/sweeps/srh8mikx</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View project at <a href='https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC' target=\"_blank\">https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View sweep at <a href='https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/sweeps/srh8mikx' target=\"_blank\">https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/sweeps/srh8mikx</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View run at <a href='https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/runs/86wuplng' target=\"_blank\">https://wandb.ai/cold-machines/Reinforcement%20Learning%20%28DQN%29%20-%20JVM-GC/runs/86wuplng</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.90it/s]\n",
      "  0%|          | 0/5000 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/vsakovskaya/my-env/lib/python3.9/site-packages/tensorflow/python/util/dispatch.py:1176: calling foldr_v2 (from tensorflow.python.ops.functional_ops) with back_prop=False is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "back_prop=False is deprecated. Consider using tf.stop_gradient instead.\n",
      "Instead of:\n",
      "results = tf.foldr(fn, elems, back_prop=False)\n",
      "Use:\n",
      "results = tf.nest.map_structure(tf.stop_gradient, tf.foldr(fn, elems))\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/vsakovskaya/my-env/lib/python3.9/site-packages/tensorflow/python/util/dispatch.py:1176: calling foldr_v2 (from tensorflow.python.ops.functional_ops) with back_prop=False is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "back_prop=False is deprecated. Consider using tf.stop_gradient instead.\n",
      "Instead of:\n",
      "results = tf.foldr(fn, elems, back_prop=False)\n",
      "Use:\n",
      "results = tf.nest.map_structure(tf.stop_gradient, tf.foldr(fn, elems))\n",
      "100%|██████████| 50/50 [00:00<00:00, 138.65it/s]] \n",
      "  2%|▏         | 101/5000 [00:16<18:44,  4.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 100: loss = 0.051988594233989716, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 142.71it/s]s]\n",
      "  4%|▍         | 201/5000 [00:32<17:56,  4.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 200: loss = 0.04959993064403534, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 150.07it/s]s]\n",
      "  6%|▌         | 301/5000 [00:48<17:17,  4.53it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 300: loss = 0.05310416966676712, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.06it/s]s]\n",
      "  8%|▊         | 401/5000 [01:04<21:05,  3.64it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 400: loss = 0.07190939784049988, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.21it/s]s]\n",
      " 10%|█         | 501/5000 [01:20<16:40,  4.50it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 500: loss = 0.06140190362930298, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.06it/s]s]\n",
      " 12%|█▏        | 601/5000 [01:36<16:26,  4.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 600: loss = 0.07814410328865051, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 142.03it/s]s]\n",
      " 14%|█▍        | 701/5000 [01:53<16:30,  4.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 700: loss = 0.05395403504371643, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 148.61it/s]s]\n",
      " 16%|█▌        | 801/5000 [02:08<15:39,  4.47it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 800: loss = 0.05778862535953522, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.60it/s]s]\n",
      " 18%|█▊        | 901/5000 [02:25<14:59,  4.56it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 900: loss = 0.05692460760474205, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 148.17it/s]s]\n",
      " 20%|██        | 1001/5000 [02:41<15:17,  4.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1000: loss = 0.059485163539648056, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 144.85it/s]/s]\n",
      " 22%|██▏       | 1101/5000 [02:57<14:36,  4.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1100: loss = 0.07164233177900314, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 139.96it/s]/s]\n",
      " 24%|██▍       | 1201/5000 [03:14<14:20,  4.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1200: loss = 0.04963871091604233, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 151.36it/s]/s]\n",
      " 26%|██▌       | 1301/5000 [03:30<14:23,  4.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1300: loss = 0.05595877766609192, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 146.66it/s]/s]\n",
      " 28%|██▊       | 1401/5000 [03:46<13:25,  4.47it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1400: loss = 0.05879542976617813, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.67it/s]/s]\n",
      " 30%|███       | 1501/5000 [04:02<12:52,  4.53it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1500: loss = 0.06716286391019821, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 131.84it/s]/s]\n",
      " 32%|███▏      | 1601/5000 [04:18<13:14,  4.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1600: loss = 0.05800986662507057, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.37it/s]/s]\n",
      " 34%|███▍      | 1701/5000 [04:35<12:29,  4.40it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1700: loss = 0.06368903815746307, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 139.72it/s]/s]\n",
      " 36%|███▌      | 1801/5000 [04:51<12:38,  4.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1800: loss = 0.05840213596820831, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 144.50it/s]/s]\n",
      " 38%|███▊      | 1901/5000 [05:07<11:47,  4.38it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 1900: loss = 0.06790091097354889, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 148.98it/s]/s]\n",
      " 40%|████      | 2001/5000 [05:24<11:03,  4.52it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2000: loss = 0.05456307530403137, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 147.69it/s]/s]\n",
      " 42%|████▏     | 2101/5000 [05:39<10:43,  4.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2100: loss = 0.05199477821588516, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 146.01it/s]/s]\n",
      " 44%|████▍     | 2201/5000 [05:56<10:51,  4.30it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2200: loss = 0.046316057443618774, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 151.46it/s]/s]\n",
      " 46%|████▌     | 2301/5000 [06:12<09:51,  4.56it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2300: loss = 0.0451720654964447, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 139.11it/s]/s]\n",
      " 48%|████▊     | 2401/5000 [06:28<09:59,  4.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2400: loss = 0.05063679814338684, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.10it/s]/s]\n",
      " 50%|█████     | 2501/5000 [06:44<09:13,  4.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2500: loss = 0.05682305246591568, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 151.48it/s]/s]\n",
      " 52%|█████▏    | 2601/5000 [07:01<08:57,  4.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2600: loss = 0.057316817343235016, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 148.48it/s]/s]\n",
      " 54%|█████▍    | 2701/5000 [07:17<08:40,  4.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2700: loss = 0.06233186647295952, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 147.44it/s]/s]\n",
      " 56%|█████▌    | 2801/5000 [07:34<08:08,  4.50it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2800: loss = 0.030263567343354225, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 150.17it/s]/s]\n",
      " 58%|█████▊    | 2901/5000 [07:49<07:45,  4.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 2900: loss = 0.12275780737400055, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.94it/s]/s]\n",
      " 60%|██████    | 3001/5000 [08:06<07:35,  4.39it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3000: loss = 0.0640171468257904, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.29it/s]/s]\n",
      " 62%|██████▏   | 3101/5000 [08:22<07:06,  4.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3100: loss = 0.06743437051773071, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 142.10it/s]/s]\n",
      " 64%|██████▍   | 3201/5000 [08:39<07:00,  4.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3200: loss = 0.07279805094003677, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.21it/s]/s]\n",
      " 66%|██████▌   | 3301/5000 [08:55<06:22,  4.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3300: loss = 0.03942070156335831, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 148.63it/s]/s]\n",
      " 68%|██████▊   | 3401/5000 [09:12<06:25,  4.14it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3400: loss = 0.06413926929235458, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 138.84it/s]/s]\n",
      " 70%|███████   | 3501/5000 [09:27<05:42,  4.38it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3500: loss = 0.04303019493818283, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 151.39it/s]/s]\n",
      " 72%|███████▏  | 3601/5000 [09:43<05:13,  4.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3600: loss = 0.0326034352183342, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.81it/s]/s]\n",
      " 74%|███████▍  | 3701/5000 [10:00<04:56,  4.38it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3700: loss = 0.06070145219564438, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 147.19it/s]/s]\n",
      " 76%|███████▌  | 3801/5000 [10:16<04:34,  4.37it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3800: loss = 0.05568353831768036, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.26it/s]/s]\n",
      " 78%|███████▊  | 3901/5000 [10:33<04:09,  4.40it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 3900: loss = 0.0671439915895462, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 146.18it/s]/s]\n",
      " 80%|████████  | 4001/5000 [10:49<03:42,  4.49it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4000: loss = 0.07973361760377884, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 147.49it/s]/s]\n",
      " 82%|████████▏ | 4101/5000 [11:06<03:38,  4.11it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4100: loss = 0.05744113028049469, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 145.54it/s]/s]\n",
      " 84%|████████▍ | 4201/5000 [11:22<03:00,  4.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4200: loss = 0.04437218979001045, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.76it/s]/s]\n",
      " 86%|████████▌ | 4301/5000 [11:37<02:34,  4.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4300: loss = 0.06655722111463547, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 149.96it/s]/s]\n",
      " 88%|████████▊ | 4401/5000 [11:55<02:13,  4.49it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4400: loss = 0.054657574743032455, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 150.15it/s]/s]\n",
      " 90%|█████████ | 4501/5000 [12:10<01:51,  4.49it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4500: loss = 0.05559392273426056, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [00:00<00:00, 151.40it/s]/s]\n",
      " 92%|█████████▏| 4601/5000 [12:27<01:32,  4.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step = 4600: loss = 0.062225304543972015, reward = -0.05968308076262474\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▎| 4678/5000 [12:39<00:49,  6.51it/s]"
     ]
    }
   ],
   "source": [
    "# Set this to True when you want to run hyperparameter tuning\n",
    "hyperparameter_tuning = True\n",
    "\n",
    "def main():\n",
    "    run = wandb.init(config = config)\n",
    "    \n",
    "    agent_args = {\n",
    "        \"q_network\":            get_q_net(action_spec, fc_layer_params),\n",
    "        \"optimizer\":            tf.compat.v1.train.AdamOptimizer(\n",
    "                                    learning_rate=run.config.learning_rate),\n",
    "        \"n_step_update\":        n_step_update, \n",
    "        \"td_errors_loss_fn\":    tf.keras.losses.MeanSquaredError(),\n",
    "        \"epsilon_greedy\":       run.config.epsilon_greedy,\n",
    "        \"target_update_period\": run.config.target_update_period,\n",
    "        \"gamma\":                run.config.gamma,\n",
    "        \"gradient_clipping\":    1,\n",
    "        \"train_step_counter\":   tf.compat.v1.train.get_or_create_global_step(),\n",
    "    }\n",
    "        \n",
    "    _agent = dqn_agent.DqnAgent(\n",
    "        time_step_spec,\n",
    "        action_spec,\n",
    "        **agent_args\n",
    "    )\n",
    "\n",
    "    loss, _, rewards = train(_agent, train_env, train_env_copy, steps = 5000, use_wandb = True)\n",
    "    \n",
    "    wandb.log({'avg_loss': np.mean(loss), 'avg_reward': np.mean(rewards)})\n",
    "\n",
    "    del _agent\n",
    "    return\n",
    "\n",
    "# Define sweep config\n",
    "sweep_configuration = {\n",
    "    'method': 'random',\n",
    "    'name': 'rl_greed',\n",
    "    \"metric\": {\n",
    "        \"name\" : \"avg_reward\",\n",
    "        \"goal\": \"maximize\"\n",
    "    },\n",
    "    \"parameters\" : {\n",
    "        \"learning_rate\": {\n",
    "            \"min\": 0.0001,\n",
    "            \"max\": 0.01\n",
    "        },\n",
    "        \"epsilon_greedy\": {'values': [0.1, 0.3, 0.5, 0.7, 0.9]},\n",
    "        \"target_update_period\": {'values': [1, 10, 100, 1000, 10000]},\n",
    "        \"gamma\": {'values': [0.1, 0.3, 0.5, 0.7, 0.9]},\n",
    "    },\n",
    "    'run_cap' : 32\n",
    "}\n",
    "\n",
    "if hyperparameter_tuning:\n",
    "    ! export WANDB_NOTEBOOK_NAME=\"main.ipynb\"\n",
    "    # Initialize sweep by passing in config. (Optional) Provide a name of the project.\n",
    "    sweep_id = wandb.sweep(sweep=sweep_configuration, project=PROJECT_NAME,)\n",
    "\n",
    "    # Start sweep job.\n",
    "    wandb.agent(sweep_id, function=main)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Restore checkpoint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test(policy, test_env, num_episodes: int=10000):\n",
    "    rewards = []\n",
    "    log_interval = 100\n",
    "    time_step = test_env.reset()\n",
    "    total_return = 0.0\n",
    "    for i in tqdm(range(10000)):\n",
    "        policy_state = policy.action(time_step)\n",
    "        time_step = test_env.step(policy_state.action)\n",
    "        reward = time_step.reward.numpy()[0]\n",
    "        total_return += reward # Calculate a sum of rewards\n",
    "\n",
    "        if i % log_interval == 0:\n",
    "            # avg_return = compute_avg_return(train_env_copy, agent.policy, 10)\n",
    "            avg_return = total_return / eval_interval\n",
    "            print('step = {0}: Average reward = {1:.5f}'.format(step, avg_return))\n",
    "            rewards.append(avg_return)\n",
    "            total_return = 0.0 # reset\n",
    "    return rewards\n",
    "\n",
    "# test(agent.policy, test_env)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Episodic Memory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "from util import episodic_memory\n",
    "from sklearn.mixture import GaussianMixture\n",
    "\n",
    "def embedding_similarity(x1, x2):\n",
    "    assert x1.shape[0] == x2.shape[0]\n",
    "    epsilon = 1e-6\n",
    "\n",
    "    # Inner product between the embeddings in x1\n",
    "    # and the embeddings in x2.\n",
    "    s = np.sum(x1 * x2, axis=-1)\n",
    "\n",
    "    s = s / (np.linalg.norm(x1, axis=-1) * np.linalg.norm(x2, axis=-1) + epsilon)\n",
    "    return 0.5 * (s + 1.0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "observations = replay_buffer.gather_all()[1].numpy().squeeze()\n",
    "actions = replay_buffer.gather_all()[2].numpy().squeeze()\n",
    "size = 15\n",
    "\n",
    "memory_buffer = np.array([[*observations[i], actions[i]] for i in range(size)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(15, 3)"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "memory_buffer.squeeze().shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "gm = GaussianMixture(n_components=memory_buffer.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-1 {color: black;}#sk-container-id-1 pre{padding: 0;}#sk-container-id-1 div.sk-toggleable {background-color: white;}#sk-container-id-1 label.sk-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-container-id-1 label.sk-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-container-id-1 label.sk-toggleable__label-arrow:hover:before {color: black;}#sk-container-id-1 div.sk-estimator:hover label.sk-toggleable__label-arrow:before {color: black;}#sk-container-id-1 div.sk-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-container-id-1 div.sk-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-container-id-1 input.sk-toggleable__control:checked~div.sk-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-container-id-1 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {content: \"▾\";}#sk-container-id-1 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 input.sk-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-container-id-1 div.sk-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-container-id-1 div.sk-estimator:hover {background-color: #d4ebff;}#sk-container-id-1 div.sk-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-container-id-1 div.sk-label:hover label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 div.sk-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: 0;}#sk-container-id-1 div.sk-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;position: relative;}#sk-container-id-1 div.sk-item {position: relative;z-index: 1;}#sk-container-id-1 div.sk-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;position: relative;}#sk-container-id-1 div.sk-item::before, #sk-container-id-1 div.sk-parallel-item::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: -1;}#sk-container-id-1 div.sk-parallel-item {display: flex;flex-direction: column;z-index: 1;position: relative;background-color: white;}#sk-container-id-1 div.sk-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-container-id-1 div.sk-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-container-id-1 div.sk-parallel-item:only-child::after {width: 0;}#sk-container-id-1 div.sk-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;}#sk-container-id-1 div.sk-label label {font-family: monospace;font-weight: bold;display: inline-block;line-height: 1.2em;}#sk-container-id-1 div.sk-label-container {text-align: center;}#sk-container-id-1 div.sk-container {/* jupyter's `normalize.less` sets `[hidden] { display: none; }` but bootstrap.min.css set `[hidden] { display: none !important; }` so we also need the `!important` here to be able to override the default hidden behavior on the sphinx rendered scikit-learn.org. See: https://github.com/scikit-learn/scikit-learn/issues/21755 */display: inline-block !important;position: relative;}#sk-container-id-1 div.sk-text-repr-fallback {display: none;}</style><div id=\"sk-container-id-1\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>GaussianMixture(n_components=3)</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-1\" type=\"checkbox\" checked><label for=\"sk-estimator-id-1\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">GaussianMixture</label><div class=\"sk-toggleable__content\"><pre>GaussianMixture(n_components=3)</pre></div></div></div></div></div>"
      ],
      "text/plain": [
       "GaussianMixture(n_components=3)"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gm.fit(memory_buffer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.00000000e+00, 9.99999917e-01, 8.31505676e-08]])"
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gm.predict_proba([[7, 12, 0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df size: 36\n",
      "[[ 4 24  3]\n",
      " [ 4 24  1]\n",
      " [ 7 24  2]\n",
      " [ 7 20  2]\n",
      " [ 7 16  3]\n",
      " [ 7 20  0]\n",
      " [ 4 20  3]\n",
      " [ 4 24  2]\n",
      " [ 4 20  2]\n",
      " [ 4 16  2]\n",
      " [ 4 12  1]\n",
      " [ 7 12  0]\n",
      " [ 4 12  1]\n",
      " [ 7 12  3]\n",
      " [ 7 16  2]]\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv(f\"avrora_synthetic_saved_states.csv\")\n",
    "print(\"df size:\", len(df))\n",
    "states  = {}\n",
    "size = 15 \n",
    "# replay_buffer.capacity\n",
    "for i in range(size):\n",
    "    states [i] = {\"args\": [df[\"MaxTenuringThreshold\"].values[i], df[\"ParallelGCThreads\"].values[i]], \"goal\": df[\"Average GC Pause\"].values[i]}\n",
    "\n",
    "# memory_buffer = np.array([states[i][\"args\"] for i in states.keys()])\n",
    "memory_buffer = np.array([[*observations[i], actions[i]] for i in range(size)])\n",
    "print(memory_buffer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{0: {'args': [7, 12], 'goal': 0.47, 'count': 2}}\n"
     ]
    }
   ],
   "source": [
    "perf_states = {}\n",
    "perf_states[0] = {\"args\": [7, 12], \"goal\": 0.47, \"count\": 1} \n",
    "saved_states = [perf_states[i][\"args\"] for i in perf_states.keys()]\n",
    "flags = [7, 12]\n",
    "if flags in saved_states:\n",
    "    for i in perf_states.keys():\n",
    "        \"\"\" \n",
    "        If current state is stored in a cache,\n",
    "        update the state goal value.\n",
    "        \"\"\"\n",
    "        if flags == perf_states[i][\"args\"]:\n",
    "            goal = perf_states[i][\"goal\"]\n",
    "            perf_states[i][\"count\"] += 1\n",
    "\n",
    "print(perf_states)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    }
   ],
   "source": [
    "for i in perf_states.keys():\n",
    "    if [7, 12] == perf_states[i][\"args\"]:\n",
    "        print(perf_states[i][\"count\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9975155087566253"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(1 + 0.01)**(-1/4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.96638056, 0.96084794, 0.98113075, 0.98760397, 0.99632883,\n",
       "       0.97980966, 0.97269349, 0.9640097 , 0.97034901, 0.97857303,\n",
       "       0.98519848, 0.98873447, 0.98519848, 1.        , 0.99432438])"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "observation = np.array([[7, 12, 3]] * size)\n",
    "similarity = embedding_similarity(observation, memory_buffer[:size])\n",
    "similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Trajectory(\n",
       "{'action': <tf.Tensor: shape=(1,), dtype=int32, numpy=array([0], dtype=int32)>,\n",
       " 'discount': <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.5], dtype=float32)>,\n",
       " 'next_step_type': <tf.Tensor: shape=(1,), dtype=int32, numpy=array([1], dtype=int32)>,\n",
       " 'observation': <tf.Tensor: shape=(1, 2), dtype=int64, numpy=array([[ 7, 12]])>,\n",
       " 'policy_info': (),\n",
       " 'reward': <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-1.], dtype=float32)>,\n",
       " 'step_type': <tf.Tensor: shape=(1,), dtype=int32, numpy=array([0], dtype=int32)>})"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "traj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "-11.512925464970229"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import math\n",
    "\n",
    "math.log(0.00001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /var/folders/dn/p1_3j9fs78g4wn8r0hcrqks40000gn/T/ipykernel_2341/2210688016.py:1: ReplayBuffer.gather_all (from tf_agents.replay_buffers.replay_buffer) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use `as_dataset(..., single_deterministic_pass=True)` instead.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /var/folders/dn/p1_3j9fs78g4wn8r0hcrqks40000gn/T/ipykernel_2341/2210688016.py:1: ReplayBuffer.gather_all (from tf_agents.replay_buffers.replay_buffer) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use `as_dataset(..., single_deterministic_pass=True)` instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000\n",
      "[0] [ 7 12]\n",
      "[10] [ 7 16]\n",
      "[15] [4 8]\n",
      "[2] [10  8]\n",
      "[4] [ 4 12]\n"
     ]
    }
   ],
   "source": [
    "observations = replay_buffer.gather_all()[1].numpy().squeeze()\n",
    "print(len(observations))\n",
    "# for i in range(100):\n",
    "#     print(f\"[{i}] {observations[i]}\")\n",
    "\n",
    "print(f\"[{0}] {observations[0]}\")\n",
    "print(f\"[{10}] {observations[10]}\")\n",
    "print(f\"[{15}] {observations[15]}\")\n",
    "print(f\"[{2}] {observations[2]}\")\n",
    "print(f\"[{4}] {observations[4]}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "observation_shape = [2]\n",
    "memory = episodic_memory.EpisodicMemory(\n",
    "    observation_shape=observation_shape,\n",
    "    observation_compare_fn=embedding_similarity,\n",
    "    capacity=150)\n",
    "\n",
    "memory.add(observation=observations[0], info=dict())\n",
    "memory.add(observation=observations[10], info=dict())\n",
    "memory.add(observation=observations[15], info=dict())\n",
    "memory.add(observation=observations[2], info=dict())\n",
    "memory.add(observation=observations[4], info=dict())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.         0.99556546 0.99941687 1.         1.        ]\n"
     ]
    }
   ],
   "source": [
    "similarities = memory.similarity(observations[0])\n",
    "print(similarities)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.4711529108389231"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.4711529108389231"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_df = pd.read_csv(f\"avrora_synthetic_saved_states.csv\")\n",
    "perf_states  = {}\n",
    "flags = [7, 12]\n",
    "def synthethic_run(flags):\n",
    "    assert len(flags) == 2, \"Amount of flags is not 2\"\n",
    "    row = new_df[((new_df[\"MaxTenuringThreshold\"] == flags[0]) & (new_df[\"ParallelGCThreads\"] == flags[1]))].values.squeeze()\n",
    "    goal = row[2]\n",
    "    return goal\n",
    "\n",
    "synthethic_run(flags)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "my-env",
   "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": "3.8.17"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}