MuZero_tf.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 69
    },
    "colab_type": "code",
    "id": "pLzVDKNOxRH-",
    "outputId": "1668210c-2e4c-49cd-96aa-86206fe6e0c7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.1.0\n",
      "Num GPUs Available:  0\n",
      "[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'), PhysicalDevice(name='/physical_device:XLA_CPU:0', device_type='XLA_CPU')]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-04-30 22:35:05,400\tINFO resource_spec.py:212 -- Starting Ray with 8.79 GiB memory available for workers and up to 4.42 GiB for objects. You can adjust these settings with ray.init(memory=<bytes>, object_store_memory=<bytes>).\n",
      "2020-04-30 22:35:05,849\tINFO services.py:1148 -- View the Ray dashboard at \u001b[1m\u001b[32mlocalhost:8265\u001b[39m\u001b[22m\n"
     ]
    }
   ],
   "source": [
    "import argparse\n",
    "import gym\n",
    "from gym import wrappers\n",
    "import numpy as np\n",
    "import pdb\n",
    "import os\n",
    "import random\n",
    "import time\n",
    "import gc\n",
    "import cv2\n",
    "import pdb\n",
    "from tqdm import tqdm\n",
    "\n",
    "import collections\n",
    "import math\n",
    "import typing\n",
    "from typing import Dict, List, Optional\n",
    "\n",
    "import tensorflow as tf\n",
    "print(tf.__version__)\n",
    "print(\"Num GPUs Available: \", len(tf.config.experimental.list_physical_devices('GPU')))\n",
    "print(tf.config.experimental.list_physical_devices())\n",
    "\n",
    "import ray\n",
    "ray.init()\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython import display\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "prYhI8PsyRtm"
   },
   "source": [
    "## TPU Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 355
    },
    "colab_type": "code",
    "id": "1qgpL6hax0j2",
    "outputId": "ca3ebdbf-0bf1-422b-de0c-bedfdd4c405e"
   },
   "outputs": [],
   "source": [
    "# resolver = tf.distribute.cluster_resolver.TPUClusterResolver(tpu='grpc://' + os.environ['COLAB_TPU_ADDR'])\n",
    "# tf.config.experimental_connect_to_cluster(resolver)\n",
    "# # This is the TPU initialization code that has to be at the beginning.\n",
    "# tf.tpu.experimental.initialize_tpu_system(resolver)\n",
    "# strategy = tf.distribute.experimental.TPUStrategy(resolver)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "DvznC_cvyd0F"
   },
   "source": [
    "## Gym Env"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "GkMoyNHzxRID"
   },
   "outputs": [],
   "source": [
    "ENVS = {\n",
    "    'breakout': 'Breakout-v0',\n",
    "    'pong': 'Pong-v0',\n",
    "    'cartpole': 'CartPole-v0',\n",
    "}\n",
    "STATE_PREPFN = {\n",
    "    'breakout': lambda s: s[50:, :, :],\n",
    "    'pong': lambda s: s[50:, :, :],\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "KnO36eBkxRIF"
   },
   "source": [
    "Env has:\n",
    "* observation\n",
    "* reward\n",
    "* done\n",
    "* info\n",
    "\n",
    "Actions are:\n",
    "['NOOP', 'FIRE', 'RIGHT', 'LEFT']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 379
    },
    "colab_type": "code",
    "id": "GYux-vt4xRIG",
    "outputId": "2c0eff51-b3be-49a6-cdc7-131095e85b43"
   },
   "outputs": [],
   "source": [
    "# env = gym.make(ENVS['breakout'])\n",
    "\n",
    "# # set seed\n",
    "# # env.seed(args.seed)\n",
    "# # np.random.seed(args.seed)\n",
    "# # torch.manual_seed(args.seed)\n",
    "\n",
    "# # episode loop\n",
    "# for i in range(1):\n",
    "#     o = env.reset()\n",
    "#     render = plt.imshow(env.render(mode='rgb_array'))\n",
    "#     # rollout inner loop\n",
    "#     for t in range(100):\n",
    "#         # render code\n",
    "#         render.set_data(env.render(mode='rgb_array')) # just update the data\n",
    "#         display.display(plt.gcf())\n",
    "#         display.clear_output(wait=True)\n",
    "        \n",
    "# #         print(o.shape)\n",
    "#         a = env.action_space.sample()\n",
    "# #         print(a)\n",
    "#         o, r, done, _ = env.step(a)\n",
    "# #         print(r)\n",
    "#         if done:\n",
    "#             print(\"Episode finished after {} timesteps\".format(t+1))\n",
    "#             break\n",
    "# env.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "CdFpMQ0jxRIK"
   },
   "source": [
    "## Helpers "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "l1wW58E0xRIK"
   },
   "outputs": [],
   "source": [
    "MAXIMUM_FLOAT_VALUE = float('inf')\n",
    "\n",
    "KnownBounds = collections.namedtuple('KnownBounds', ['min', 'max'])\n",
    "\n",
    "\n",
    "class MinMaxStats(object):\n",
    "    \"\"\"A class that holds the min-max values of the tree.\"\"\"\n",
    "\n",
    "    def __init__(self, known_bounds: Optional[KnownBounds]):\n",
    "        self.maximum = known_bounds.max if known_bounds else -MAXIMUM_FLOAT_VALUE\n",
    "        self.minimum = known_bounds.min if known_bounds else MAXIMUM_FLOAT_VALUE\n",
    "\n",
    "    def update(self, value: float):\n",
    "        self.maximum = max(self.maximum, value)\n",
    "        self.minimum = min(self.minimum, value)\n",
    "\n",
    "    def normalize(self, value: float) -> float:\n",
    "        if self.maximum > self.minimum:\n",
    "            # We normalize only when we have set the maximum and minimum values.\n",
    "            return (value - self.minimum) / (self.maximum - self.minimum)\n",
    "        return value"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LFaqZtJBxRIN"
   },
   "source": [
    "### MuZero Config Params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "excWi-RwxRIN"
   },
   "outputs": [],
   "source": [
    "class MuZeroConfig(object):\n",
    "\n",
    "    def __init__(self,\n",
    "                action_space_size: int,\n",
    "                max_moves: int,\n",
    "                discount: float,\n",
    "                dirichlet_alpha: float,\n",
    "                num_simulations: int,\n",
    "                batch_size: int,\n",
    "                td_steps: int,\n",
    "                num_actors: int,\n",
    "                lr_init: float,\n",
    "                lr_decay_steps: float,\n",
    "                visit_softmax_temperature_fn,\n",
    "                known_bounds: Optional[KnownBounds] = None):\n",
    "        ### Self-Play\n",
    "        self.action_space_size = action_space_size\n",
    "        self.num_actors = num_actors\n",
    "\n",
    "        self.visit_softmax_temperature_fn = visit_softmax_temperature_fn\n",
    "        self.max_moves = max_moves\n",
    "        self.num_simulations = num_simulations\n",
    "        self.discount = discount\n",
    "\n",
    "        # Root prior exploration noise.\n",
    "        self.root_dirichlet_alpha = dirichlet_alpha\n",
    "        self.root_exploration_fraction = 0.25\n",
    "\n",
    "        # UCB formula\n",
    "        self.pb_c_base = 19652\n",
    "        self.pb_c_init = 1.25\n",
    "\n",
    "        # If we already have some information about which values occur in the\n",
    "        # environment, we can use them to initialize the rescaling.\n",
    "        # This is not strictly necessary, but establishes identical behaviour to\n",
    "        # AlphaZero in board games.\n",
    "        self.known_bounds = known_bounds\n",
    "\n",
    "        ### Training\n",
    "        self.selfplay_iterations = int(1) ##\n",
    "        self.training_steps = int(1000e3)\n",
    "        self.checkpoint_interval = int(1e3)\n",
    "        self.window_size = int(1e6)\n",
    "        self.batch_size = batch_size\n",
    "        self.num_unroll_steps = 5\n",
    "        self.td_steps = td_steps\n",
    "\n",
    "        self.weight_decay = 1e-4\n",
    "        self.momentum = 0.9\n",
    "\n",
    "        # Exponential learning rate schedule\n",
    "        self.lr_init = lr_init\n",
    "        self.lr_decay_rate = 0.1\n",
    "        self.lr_decay_steps = lr_decay_steps\n",
    "\n",
    "    def new_game(self):\n",
    "        return Game(self.action_space_size, self.discount)\n",
    "    \n",
    "def make_atari_config() -> MuZeroConfig:\n",
    "\n",
    "    def visit_softmax_temperature(num_moves, training_steps):\n",
    "        if training_steps < 500e3:\n",
    "            return 1.0\n",
    "        elif training_steps < 750e3:\n",
    "            return 0.5\n",
    "        else:\n",
    "            return 0.25\n",
    "\n",
    "    return MuZeroConfig(\n",
    "        action_space_size=4,\n",
    "        max_moves=27000,  # Half an hour at action repeat 4.\n",
    "        discount=0.997,\n",
    "        dirichlet_alpha=0.25,\n",
    "        num_simulations=50,\n",
    "        batch_size=16,#1024,\n",
    "        td_steps=10,\n",
    "        num_actors=4,#350,\n",
    "        lr_init=0.05,\n",
    "        lr_decay_steps=350e3,\n",
    "        visit_softmax_temperature_fn=visit_softmax_temperature)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "iSbIRX3NxRIQ"
   },
   "outputs": [],
   "source": [
    "# class Action(object):\n",
    "\n",
    "#     def __init__(self, index: int):\n",
    "#         self.index = index\n",
    "\n",
    "#     def __hash__(self):\n",
    "#         return self.index\n",
    "\n",
    "#     def __eq__(self, other):\n",
    "#         return self.index == other.index\n",
    "\n",
    "#     def __gt__(self, other):\n",
    "#         return self.index > other.index\n",
    "    \n",
    "#     def __str__(self):\n",
    "#         return f\"Action({self.index})\"\n",
    "    \n",
    "#     def __repr__(self):\n",
    "#         return \"<%s(%s) at %s>\" % (self.__class__.__name__, self.index, id(self))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "nesCzurBxRIT"
   },
   "outputs": [],
   "source": [
    "class Node(object):\n",
    "\n",
    "    def __init__(self, prior: float):\n",
    "        self.visit_count = 0\n",
    "        self.to_play = -1\n",
    "        self.prior = prior\n",
    "        self.value_sum = 0.0\n",
    "        self.children = {}\n",
    "        self.hidden_state = None\n",
    "        self.reward = 0.0\n",
    "\n",
    "    def expanded(self) -> bool:\n",
    "        return len(self.children) > 0\n",
    "\n",
    "    def value(self) -> float:\n",
    "        if self.visit_count == 0:\n",
    "            return 0.0\n",
    "        return self.value_sum / self.visit_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "OHHtQR2NxRIV"
   },
   "outputs": [],
   "source": [
    "class ActionHistory(object):\n",
    "    \"\"\"Simple history container used inside the search.\n",
    "\n",
    "    Only used to keep track of the actions executed.\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self, history: List[int], action_space_size: int):\n",
    "        self.history = list(history)\n",
    "        self.action_space_size = action_space_size\n",
    "\n",
    "    def clone(self):\n",
    "        return ActionHistory(self.history, self.action_space_size)\n",
    "\n",
    "    def add_action(self, action: int):\n",
    "        self.history.append(action)\n",
    "\n",
    "    def last_action(self) -> int:\n",
    "        return self.history[-1]\n",
    "\n",
    "    def action_space(self) -> List[int]:\n",
    "        return [i for i in range(self.action_space_size)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "dNRzywV0xRIX"
   },
   "outputs": [],
   "source": [
    "class Environment(object):\n",
    "    \"\"\"The environment MuZero is interacting with.\"\"\"\n",
    "    def __init__(self):\n",
    "        self.env = gym.make(ENVS['breakout'])\n",
    "        self.obs_history = [self.prepro(self.env.reset())]\n",
    "        self.done = False\n",
    "        \n",
    "    def step(self, action: int):\n",
    "        obs, reward, self.done, info = self.env.step(action)\n",
    "        self.obs_history.append(self.prepro(obs))\n",
    "        return float(reward)\n",
    "    \n",
    "    def terminal(self):\n",
    "        return self.done\n",
    "    \n",
    "    # @tf.function\n",
    "    def legal_actions(self):\n",
    "        \"\"\"Env specific rules for legality of moves\n",
    "        TODO: if at wall don't allow movement into the wall\"\"\"\n",
    "        return [a for a in range(self.env.action_space.n)]\n",
    "\n",
    "    def prepro(self, obs, size=(80,80)):\n",
    "        \"\"\"Crop, resize, B&W\"\"\"\n",
    "        p_obs = obs[25:195,:,0] / 255 # crop and normalise to [0,1]\n",
    "        return cv2.resize(p_obs, size, interpolation=cv2.INTER_NEAREST) # resize\n",
    "\n",
    "    # def prepro(self, obs, size=(80,80)):\n",
    "    #     return obs\n",
    "    def get_obs(self, start:int, end:int=None):\n",
    "        return self.obs_history[max(start,0):end]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Zq_M9MvhxRIa"
   },
   "source": [
    "![title](images/batch_example.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "-stmlcz9xRIa"
   },
   "source": [
    "Refresher on TD Learning, which is what is used to estimate the value targets.\n",
    "\n",
    "If we were to calculate value function without estimation we would have to wait until the final reward updates could be propagated:\n",
    "$$V(s_t) = V(s_t) + \\alpha[R_t-V(s_t)]$$\n",
    "where $s_t$ is state visited at time $t$, $R_t$ is total reward after time $t$ and $\\alpha$ is the LR.\n",
    "\n",
    "However for TD methods, an estimate of the final reward is calculated at each state and (s,a) value updated for every step of the way. We are essentially doing a finite step lookahead and updating based on that estimate. TD method is called a \"bootstrapping\" method, becuase the value is updated partly using an existing estimate and not a final reward.\n",
    "$$V(s_t) = V(s_t) + \\alpha[r_{t+1}+\\gamma V(s_{t+1})-V(s_t)]$$\n",
    "and in our case we formulate this as an N step bootstrap:\n",
    "$$z_t=u_{t+1} + \\gamma u_{t+2} + ... + \\gamma^{n-1} u_{t+n} + \\gamma^n \\nu_{t+n}$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6JajZqo7xRIb"
   },
   "outputs": [],
   "source": [
    "class Game(object):\n",
    "    \"\"\"A single episode of interaction with the environment. (One trajectory)\"\"\"\n",
    "    def __init__(self, action_space_size: int, discount: float):\n",
    "        self.env = Environment()\n",
    "        self.history = [] # actual actions a\n",
    "        self.rewards = [] # observed rewards u\n",
    "        self.child_visits = [] # search tree action distributions pi\n",
    "        self.root_values = [] # values ν\n",
    "        self.action_space_size = action_space_size\n",
    "        self.gamma = discount\n",
    "    \n",
    "    def terminal(self) -> bool:\n",
    "        return self.env.terminal()\n",
    "    \n",
    "    def legal_actions(self) -> List[int]:\n",
    "        return self.env.legal_actions()\n",
    "    \n",
    "    def apply(self, action: int):\n",
    "        reward = self.env.step(action)\n",
    "        self.rewards.append(reward)\n",
    "        self.history.append(action)\n",
    "        \n",
    "    def store_search_statistics(self, root: Node):\n",
    "        \"\"\"Stores the MCTS search value of node and search policy (visits ratio of children)\"\"\"\n",
    "        sum_visits = sum(child.visit_count for child in root.children.values())\n",
    "        action_space = (index for index in range(self.action_space_size))\n",
    "        # search policy π = [0.1,0.5,0.4] probability over children\n",
    "        self.child_visits.append([\n",
    "            root.children[a].visit_count / sum_visits if a in root.children else 0\n",
    "            for a in action_space\n",
    "        ])\n",
    "        # search value ν\n",
    "        self.root_values.append(root.value())\n",
    "    \n",
    "    def make_image(self, t: int, feat_history_len:int = 32):\n",
    "        \"\"\"Observation at chosen position w/ history\"\"\"\n",
    "        # Game specific feature planes\n",
    "        # For Atari we have the 32 most recent RGB frames at resolution 96x96\n",
    "        # Instead I use 80x80x1 B&W\n",
    "        frames = self.env.get_obs(t-feat_history_len+1, t+2) # We want 32 frames up to and including t\n",
    "        # Cast to tensor and add dummy batch dim\n",
    "        # Todo: figure out how to stack RGB images - i.e. colour & time dimensions\n",
    "        frame_tensor = tf.convert_to_tensor(np.stack(frames,axis=-1))\n",
    "        # If we're missing a channel dim add one\n",
    "        # if len(frame_tensor.shape)==3:\n",
    "        #     frame_tensor = frame_tensor.expand_dims(-1) # this is wrong\n",
    "        # Pad out sequences with insufficient history\n",
    "        padding_size = feat_history_len-frame_tensor.shape[-1]\n",
    "        padded_frames = tf.pad(frame_tensor, paddings=[[0, 0], [0, 0], [padding_size, 0]], constant_values=0)\n",
    "        padded_frames = tf.expand_dims(padded_frames, 0) # dummy batch dim for 4D\n",
    "        return padded_frames\n",
    "    \n",
    "    def make_target(self, t: int, K: int, td: int):\n",
    "        \"\"\"\n",
    "        (value,reward,policy) target for each unroll step t to t+K\n",
    "        This is taken from actuals results of the game (to be stored in replay buffer)\n",
    "        Uses TD learning to calculate value target via n step bootstrap, see above\n",
    "        \"\"\"\n",
    "        # The value target is the discounted root value of the search tree N steps\n",
    "        # into the future, plus the discounted sum of all rewards until then.\n",
    "        # Returns target tuple (value, reward, policy) i.e. (z,u,pi)\n",
    "        # K=5\n",
    "        targets = []\n",
    "        for current_index in range(t, t + K + 1):\n",
    "            ## Value Target z_{t+K} ##\n",
    "            \n",
    "            bootstrap_index = current_index + td \n",
    "            # If our TD lookahead is still before the end of the game, the update with that \n",
    "            # future game state value estimate ν_{t+N}\n",
    "            if bootstrap_index < len(self.root_values):\n",
    "                # γ^N*ν_{t+N}\n",
    "                value =  self.root_values[bootstrap_index] * self.gamma**td\n",
    "            else:\n",
    "                value = 0\n",
    "            \n",
    "            # Rest of the TD value estimate from observed rewards: u_{t+1} + ... + γ^{N-1} u_{t+n}\n",
    "            for i, reward in enumerate(self.rewards[current_index:bootstrap_index]):\n",
    "                value += reward * self.gamma**i\n",
    "                \n",
    "            ## Reward u_{t+K} and Action π_{t+K} Targets ##\n",
    "            \n",
    "            # For simplicity the network always predicts the most recently received\n",
    "            # reward, even for the initial representation network where we already\n",
    "            # know this reward.\n",
    "            if current_index > 0 and current_index <= len(self.rewards):\n",
    "                last_reward = self.rewards[current_index - 1]\n",
    "            else:\n",
    "                last_reward = 0\n",
    "            \n",
    "            if current_index < len(self.root_values):\n",
    "                targets.append((value, last_reward, self.child_visits[current_index]))\n",
    "            else:\n",
    "                # States past the end are treated as absorbing states\n",
    "                targets.append((0, last_reward, []))\n",
    "        return targets \n",
    "    \n",
    "    # @tf.function\n",
    "    def action_history(self) -> ActionHistory:\n",
    "        return ActionHistory(self.history, self.action_space_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Hc3EWn8gxRId"
   },
   "outputs": [],
   "source": [
    "class ReplayBuffer(object):\n",
    "    \"\"\"Stores the target tuples to sample from later during training\"\"\"\n",
    "    def __init__(self, config: MuZeroConfig):\n",
    "        self.window_size = config.window_size\n",
    "        self.batch_size = config.batch_size\n",
    "        self.buffer = []\n",
    "\n",
    "    def save_game(self, game):\n",
    "        # Pop off oldest replays to make space for new ones\n",
    "        if len(self.buffer) > self.window_size:\n",
    "            self.buffer.pop(0)\n",
    "        self.buffer.append(game)\n",
    "\n",
    "    def sample_batch(self, K: int, td: int):\n",
    "        \"\"\"\n",
    "        Inputs\n",
    "            K: num unroll steps\n",
    "            td: num TD steps\n",
    "        Outputs\n",
    "            (observation, next K actions, target tuple)\n",
    "        \"\"\"\n",
    "        games = [self.sample_game() for _ in range(self.batch_size)]\n",
    "        game_pos = [(g, self.sample_position(g)) for g in games]\n",
    "        return [(g.make_image(i), g.history[i:i + K],\n",
    "                 g.make_target(i, K, td))\n",
    "                for (g, i) in game_pos]\n",
    "\n",
    "    def sample_game(self) -> Game:\n",
    "        # TODO: figure out sampling regime\n",
    "        # Sample game from buffer either uniformly or according to some priority e.g. importance sampling.\n",
    "        game_ix = random.randint(0,len(self.buffer)-1) # random uniform\n",
    "        return self.buffer[game_ix]\n",
    "\n",
    "    def sample_position(self, game) -> int:\n",
    "        # Sample position from game either uniformly or according to some priority.\n",
    "        pos_ix = random.randint(0,len(game.root_values)-1) # random uniform\n",
    "        return pos_ix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Y0pD-gdHxRIg"
   },
   "source": [
    "## NN Networks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "qgNoaX2qxRIh"
   },
   "source": [
    "#### 1. Prediction Network (F)\n",
    "The prediction function pk, vk = fθ(sk) uses the same architecture as AlphaZero: one or two convolutional layers that preserve the resolution but reduce the number of planes, followed by a fully connected layer to the size of the output.\n",
    "\n",
    "#### 2. Dynamics Network (G)\n",
    "Both the representation and dynamics function use the same architecture as AlphaZero, but with 16 instead of 20 residual blocks. We use 3x3 kernels and 256 hidden planes for each convolution. \n",
    "\n",
    "For the dynamics function (which always operates at the downsampled resolution of 6x6), the action is first encoded as an image, then stacked with the hidden state of the previous step along the plane dimension\n",
    "\n",
    "#### 3. Representation Network (H)\n",
    "For Atari, where observations have large spatial resolution, the representation function starts with a sequence\n",
    "of convolutions with stride 2 to reduce the spatial resolution. Specifically, starting with an input observation of resolution 96x96 and 128 planes (32 history frames of 3 color channels each, concatenated with the corresponding 32 actions broadcast to planes), we downsample as follows:\n",
    "* 1 convolution with stride 2 and 128 output planes, output resolution 48x48. \n",
    "* 2 residual blocks with 128 planes \n",
    "* 1 convolution with stride 2 and 256 output planes, output resolution 24x24. \n",
    "* 3 residual blocks with 256 planes. \n",
    "* Average pooling with stride 2, output resolution 12x12. \n",
    "* 3 residual blocks with 256 planes. \n",
    "* Average pooling with stride 2, output resolution 6x6.\n",
    "The kernel size is 3x3 for all operations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "cZYPap-uzHMa"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.layers import Conv2D, BatchNormalization, Dense, Add, ReLU, Input, Flatten\n",
    "from tensorflow.keras.models import Model\n",
    "\n",
    "def ResBlock(x_in, nf=128):\n",
    "    x = Conv2D(nf, 3, padding='same', use_bias=False)(x_in)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = ReLU()(x)\n",
    "    x = Conv2D(nf, 3, padding='same', use_bias=False)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Add()([x, input_data])\n",
    "    x = ReLU()(x)\n",
    "    return x\n",
    "\n",
    "def ConvBlock(x_in, nf, s=1, bn=True):\n",
    "    x = Conv2D(nf, 3, padding='same', strides=s, use_bias=not bn)(x_in)\n",
    "    if bn: x = BatchNormalization()(x)\n",
    "    x = ReLU()(x)\n",
    "    return x\n",
    "\n",
    "def ReprNet(input_shape=(80,80,32), nf=128):\n",
    "    o = Input(shape=input_shape)\n",
    "    x = ConvBlock(o, nf, 2)\n",
    "    x = ConvBlock(x, nf, 2)\n",
    "    x = ConvBlock(x, nf, 2)\n",
    "    s = ConvBlock(x, nf, 2)\n",
    "    return Model(o, s)\n",
    "\n",
    "def DynaNet(input_shape=(5,5,129), nf=128):\n",
    "    s = Input(shape=input_shape)\n",
    "    x = ConvBlock(s, nf)\n",
    "    x = ConvBlock(x, nf)\n",
    "    x = ConvBlock(x, nf)\n",
    "    s_new = ConvBlock(x, nf)\n",
    "    \n",
    "    r = Flatten()(s_new)\n",
    "    r = Dense(1)(r)\n",
    "    return Model(s, [s_new, r])\n",
    "\n",
    "def PredNet(input_shape=(5,5,128), nf=64, num_actions=4):\n",
    "    s = Input(shape=input_shape)\n",
    "    x = ConvBlock(s, nf)\n",
    "    x = ConvBlock(x, nf//2) \n",
    "    x = Flatten()(x)\n",
    "\n",
    "    a = Dense(num_actions)(x)\n",
    "    v = Dense(1)(x)\n",
    "    return Model(s, [a, v])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 605
    },
    "colab_type": "code",
    "id": "3hOyHqx-xRIj",
    "outputId": "1ab688f0-0965-4a6e-e389-0232362b5335"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_9\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_10 (InputLayer)        [(None, 80, 80, 32)]      0         \n",
      "_________________________________________________________________\n",
      "conv2d_30 (Conv2D)           (None, 40, 40, 128)       36864     \n",
      "_________________________________________________________________\n",
      "batch_normalization_30 (Batc (None, 40, 40, 128)       512       \n",
      "_________________________________________________________________\n",
      "re_lu_30 (ReLU)              (None, 40, 40, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_31 (Conv2D)           (None, 20, 20, 128)       147456    \n",
      "_________________________________________________________________\n",
      "batch_normalization_31 (Batc (None, 20, 20, 128)       512       \n",
      "_________________________________________________________________\n",
      "re_lu_31 (ReLU)              (None, 20, 20, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_32 (Conv2D)           (None, 10, 10, 128)       147456    \n",
      "_________________________________________________________________\n",
      "batch_normalization_32 (Batc (None, 10, 10, 128)       512       \n",
      "_________________________________________________________________\n",
      "re_lu_32 (ReLU)              (None, 10, 10, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_33 (Conv2D)           (None, 5, 5, 128)         147456    \n",
      "_________________________________________________________________\n",
      "batch_normalization_33 (Batc (None, 5, 5, 128)         512       \n",
      "_________________________________________________________________\n",
      "re_lu_33 (ReLU)              (None, 5, 5, 128)         0         \n",
      "=================================================================\n",
      "Total params: 481,280\n",
      "Trainable params: 480,256\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Input: (32,80,80) - 80x80 game in B&W with 32 length obs history\n",
    "# Output: (128,5,5) - s0\n",
    "ReprNet().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 674
    },
    "colab_type": "code",
    "id": "mVUBy_d8xRIm",
    "outputId": "b7068355-4985-44cb-cf79-ad34aa8912f4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_10\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_11 (InputLayer)        [(None, 5, 5, 129)]       0         \n",
      "_________________________________________________________________\n",
      "conv2d_34 (Conv2D)           (None, 5, 5, 128)         148608    \n",
      "_________________________________________________________________\n",
      "batch_normalization_34 (Batc (None, 5, 5, 128)         512       \n",
      "_________________________________________________________________\n",
      "re_lu_34 (ReLU)              (None, 5, 5, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_35 (Conv2D)           (None, 5, 5, 128)         147456    \n",
      "_________________________________________________________________\n",
      "batch_normalization_35 (Batc (None, 5, 5, 128)         512       \n",
      "_________________________________________________________________\n",
      "re_lu_35 (ReLU)              (None, 5, 5, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_36 (Conv2D)           (None, 5, 5, 128)         147456    \n",
      "_________________________________________________________________\n",
      "batch_normalization_36 (Batc (None, 5, 5, 128)         512       \n",
      "_________________________________________________________________\n",
      "re_lu_36 (ReLU)              (None, 5, 5, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_37 (Conv2D)           (None, 5, 5, 128)         147456    \n",
      "_________________________________________________________________\n",
      "batch_normalization_37 (Batc (None, 5, 5, 128)         512       \n",
      "_________________________________________________________________\n",
      "re_lu_37 (ReLU)              (None, 5, 5, 128)         0         \n",
      "_________________________________________________________________\n",
      "flatten_6 (Flatten)          (None, 3200)              0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 1)                 3201      \n",
      "=================================================================\n",
      "Total params: 596,225\n",
      "Trainable params: 595,201\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Input: (129,5,5) - 5x5 internal state with nf channels +1 action dim\n",
    "# Output: (128,5,5), (1,) - s', r\n",
    "DynaNet().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 501
    },
    "colab_type": "code",
    "id": "E5TKxIctxRIo",
    "outputId": "9568d2cf-86db-4c57-ebe7-2496059100e1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_11\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_12 (InputLayer)           [(None, 5, 5, 128)]  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_38 (Conv2D)              (None, 5, 5, 64)     73728       input_12[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_38 (BatchNo (None, 5, 5, 64)     256         conv2d_38[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_38 (ReLU)                 (None, 5, 5, 64)     0           batch_normalization_38[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_39 (Conv2D)              (None, 5, 5, 32)     18432       re_lu_38[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_39 (BatchNo (None, 5, 5, 32)     128         conv2d_39[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_39 (ReLU)                 (None, 5, 5, 32)     0           batch_normalization_39[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "flatten_7 (Flatten)             (None, 800)          0           re_lu_39[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_10 (Dense)                (None, 4)            3204        flatten_7[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_11 (Dense)                (None, 1)            801         flatten_7[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 96,549\n",
      "Trainable params: 96,357\n",
      "Non-trainable params: 192\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Input: (128,5,5) - 5x5 internal state with nf channels\n",
    "# Output: (4,), (1,) - p, v\n",
    "PredNet().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ZiHSAcpFxRIq"
   },
   "outputs": [],
   "source": [
    "class NetworkOutput(typing.NamedTuple):\n",
    "    value: float\n",
    "    reward: float\n",
    "    policy_logits: tf.Tensor # Dict[Action, float]\n",
    "    hidden_state: List[float] # not sure about this one lol"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "35S6OKqbxRIs"
   },
   "outputs": [],
   "source": [
    "class Network(object):\n",
    "    \"\"\"TODO: Implement in Pytorch/TF\"\"\"\n",
    "    def __init__(self, h_in=5, w_in=5, c_in=32, nf=128, n_acts=4):\n",
    "        # Initialise a uniform network - should I init these networks explicitly?\n",
    "        self.f = PredNet((5,5,nf), nf, n_acts)\n",
    "        self.g = DynaNet((5,5,nf+1), nf)\n",
    "        self.h = ReprNet((80,80,32), nf)\n",
    "        self.steps = 0\n",
    "\n",
    "#     @tf.function\n",
    "    def initial_inference(self, obs) -> NetworkOutput:\n",
    "        # representation + prediction function\n",
    "        # input: 32x80x80 observation\n",
    "        state = self.h(obs)\n",
    "        policy_logits, value = self.f(state)\n",
    "        # state, policy_logits, value = self.initial_inference_compiled(obs)\n",
    "        # policy = {Action(i):p for i,p in enumerate(policy_logits[0])}\n",
    "        policy = policy_logits[0]\n",
    "        return NetworkOutput(value[0], 0.0, policy, state) # keep state 4D\n",
    "    \n",
    "#     @tf.function\n",
    "    def recurrent_inference(self, state, action) -> NetworkOutput:\n",
    "        # dynamics + prediction function\n",
    "        # Input: hidden state nfx5x5\n",
    "        # Concat/pad action to channel dim of states\n",
    "        state_action = tf.pad(state, paddings=[[0, 0], [0, 0], [0, 0], [0, 1]], constant_values=action)\n",
    "        next_state, reward =  self.g(state_action)\n",
    "        policy_logits, value = self.f(next_state)\n",
    "        # next_state, reward, policy_logits, value = self.recurrent_inference_compiled(state, action)\n",
    "        # policy = {Action(i):p for i,p in enumerate(policy_logits[0])}\n",
    "        policy = policy_logits[0]\n",
    "        return NetworkOutput(value[0], reward[0], policy, next_state)\n",
    "\n",
    "    def get_weights(self):\n",
    "        # Returns the weights of this network.\n",
    "        self.steps += 1 # probably not ideal\n",
    "        return [self.f.parameters(), self.g.parameters(), self.h.parameters()]\n",
    "\n",
    "    def training_steps(self) -> int:\n",
    "        # How many steps / batches the network has been trained for.\n",
    "        return self.steps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "zNRwMH0AxRIu"
   },
   "outputs": [],
   "source": [
    "class SharedStorage(object):\n",
    "\n",
    "    def __init__(self):\n",
    "        self._networks = {}\n",
    "\n",
    "    def latest_network(self) -> Network:\n",
    "        if self._networks:\n",
    "            return self._networks[max(self._networks.keys())]\n",
    "        else:\n",
    "            # policy -> uniform, value -> 0, reward -> 0\n",
    "            return make_uniform_network()\n",
    "\n",
    "    def save_network(self, step: int, network: Network):\n",
    "        self._networks[step] = network\n",
    "\n",
    "def make_uniform_network():\n",
    "    return Network()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "mPaehAh9xRIx"
   },
   "source": [
    "# 1. Self-play"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "v5MiJYXOxRIx"
   },
   "outputs": [],
   "source": [
    "# Each self-play job is independent of all others; it takes the latest network\n",
    "# snapshot, produces a game and makes it available to the training job by\n",
    "# writing it to a shared replay buffer.\n",
    "@ray.remote\n",
    "def run_selfplay(config: MuZeroConfig, storage: SharedStorage,\n",
    "                 replay_buffer: ReplayBuffer):\n",
    "    \"\"\"TODO: Multithread this\"\"\"\n",
    "    # tf.summary.trace_on()\n",
    "    for i in tqdm(range(config.selfplay_iterations), desc='Self-play iter'):\n",
    "        network = storage.latest_network()\n",
    "        game = play_game(config, network)\n",
    "        print(game.root_values)\n",
    "        replay_buffer.save_game(game)\n",
    "        # tf.summary.trace_export(\"Selfplay\", step=i, profiler_outdir='logs')\n",
    "    # tf.summary.trace_off()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "5ijhRSlVxRIz"
   },
   "source": [
    "### Run 1 Game/Trajectory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "rSB8CTSwxRIz"
   },
   "outputs": [],
   "source": [
    "# Each game is produced by starting at the initial board position, then\n",
    "# repeatedly executing a Monte Carlo Tree Search to generate moves until the end\n",
    "# of the game is reached.\n",
    "def play_game(config: MuZeroConfig, network: Network) -> Game:\n",
    "    game = config.new_game()\n",
    "\n",
    "    while not game.terminal() and len(game.history) < config.max_moves:\n",
    "        # At the root of the search tree we use the representation function h to\n",
    "        # obtain a hidden state given the current observation.\n",
    "        root = Node(0)\n",
    "        current_observation = game.make_image(-1) # 80x80x32\n",
    "        expand_node(root, game.legal_actions(),\n",
    "                    network.initial_inference(current_observation))\n",
    "        add_exploration_noise(config, root)\n",
    "\n",
    "        # We then run a Monte Carlo Tree Search using only action sequences and the\n",
    "        # model learned by the network.\n",
    "        run_mcts(config, root, game.action_history(), network)\n",
    "        action = select_action(config, len(game.history), root, network)\n",
    "        game.apply(action)\n",
    "        game.store_search_statistics(root)\n",
    "    return game"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "jbsyVFyRxRI1"
   },
   "source": [
    "#### Exploration Noise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "AegOB2i0xRI1"
   },
   "outputs": [],
   "source": [
    "# At the start of each search, we add dirichlet noise to the prior of the root\n",
    "# to encourage the search to explore new actions.\n",
    "# @tf.function\n",
    "def add_exploration_noise(config: MuZeroConfig, node: Node):\n",
    "    actions = list(node.children.keys())\n",
    "    noise = np.random.dirichlet([config.root_dirichlet_alpha] * len(actions))\n",
    "    frac = config.root_exploration_fraction\n",
    "    for a, n in zip(actions, noise):\n",
    "        node.children[a].prior = node.children[a].prior * (1 - frac) + n * frac"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "DvObjpTlxRI3"
   },
   "source": [
    "### Softmax search policy $\\pi$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6HYHEl9WxRI4"
   },
   "outputs": [],
   "source": [
    "# @tf.function\n",
    "def select_action(config: MuZeroConfig, num_moves: int, node: Node,\n",
    "                  network: Network) -> int:\n",
    "    \"\"\"Search policy: softmax probability over actions dictated by visited counts\"\"\"\n",
    "    # Visit counts of chilren nodes - policy proportional to counts\n",
    "    visit_counts = [\n",
    "        (child.visit_count, action) for action, child in node.children.items()\n",
    "    ]\n",
    "    # Get softmax temp\n",
    "    t = config.visit_softmax_temperature_fn(\n",
    "        num_moves=num_moves, training_steps=network.training_steps())\n",
    "    _, action = softmax_sample(visit_counts, t)\n",
    "    return action\n",
    "\n",
    "# @tf.function\n",
    "def softmax_sample(distribution, T: float):\n",
    "    counts = np.array([d[0] for d in distribution])\n",
    "    actions = [d[1] for d in distribution]\n",
    "    softmax_probs = softmax(counts/T)\n",
    "    sampled_action = np.random.choice(actions, size=1, p=softmax_probs)[0]\n",
    "    return _, sampled_action\n",
    "#     return nn.Softmax(dim=0)(distribution/T)\n",
    "\n",
    "# @tf.function\n",
    "def softmax(x, axis=-1):\n",
    "    e_x = np.exp(x - np.max(x))\n",
    "    return e_x / e_x.sum(axis=axis, keepdims=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "cb7XvbKBxRI6"
   },
   "source": [
    "### MCTS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ZB69k9kHxRI6"
   },
   "outputs": [],
   "source": [
    "# Core Monte Carlo Tree Search algorithm.\n",
    "# To decide on an action, we run N simulations, always starting at the root of\n",
    "# the search tree and traversing the tree according to the UCB formula until we\n",
    "# reach a leaf node.\n",
    "def run_mcts(config: MuZeroConfig, root: Node, action_history: ActionHistory,\n",
    "             network: Network):\n",
    "    \"\"\"TODO: Multithread\"\"\"\n",
    "    min_max_stats = MinMaxStats(config.known_bounds)\n",
    "\n",
    "    for _ in range(config.num_simulations):\n",
    "        history = action_history.clone()\n",
    "        node = root\n",
    "        search_path = [node]\n",
    "\n",
    "        # Traverse tree, expanding by highest UCB until leaf reached\n",
    "        while node.expanded():\n",
    "            action, node = select_child(config, node, min_max_stats) # UCB selection\n",
    "            history.add_action(action)\n",
    "            search_path.append(node)\n",
    "\n",
    "        # Inside the search tree we use the dynamics function to obtain the next\n",
    "        # hidden state given an action and the previous hidden state.\n",
    "        parent = search_path[-2] # parent of leaf\n",
    "        # Dynamics: g(s_{k-1},a_k) = r_k, s_k\n",
    "        # Predictions: f(s_k) = v_k, p_k\n",
    "        # -> (v,r,p,s)\n",
    "        network_output = network.recurrent_inference(parent.hidden_state,\n",
    "                                                     history.last_action())\n",
    "        # expand node using v,r,p predictions from NN\n",
    "        expand_node(node, history.action_space(), network_output)\n",
    "\n",
    "        # back up values to the root node\n",
    "        backpropagate(search_path, network_output.value, config.discount, \n",
    "                      min_max_stats)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "5FUTdI0axRI9"
   },
   "source": [
    "#### i. Selection: UCB Child Selection\n",
    "We select the child node/state based on the action that maximises over an upper confidence bound (UCB):\n",
    "$$a^{k}=\\arg \\max _{a}\\left[Q(s, a)+P(s, a) \\cdot \\frac{\\sqrt{\\sum_{b} N(s, b)}}{1+N(s, a)}\\left(c_{1}+\\log \\left(\\frac{\\sum_{b} N(s, b)+c_{2}+1}{c_{2}}\\right)\\right)\\right]$$\n",
    "To clarify: $\\sum_{b} N(s, b)$ is just the sum of counts over all child nodes, i.e. the count of the parent node.\n",
    "\n",
    "\"The constants c1 and c2 are used to control the influence of the prior P(s, a) relative to the value Q(s, a) as\n",
    "nodes are visited more often. In our experiments, $c_1$ = 1.25 and $c_2$ = 19652\" - (pg. 12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KyKU0IngxRI9"
   },
   "outputs": [],
   "source": [
    "# Select the child with the highest UCB score.\n",
    "def select_child(config: MuZeroConfig, node: Node,\n",
    "                 min_max_stats: MinMaxStats):\n",
    "    _, action, child = max(\n",
    "        (ucb_score(config, node, child, min_max_stats), action,\n",
    "         child) for action, child in node.children.items())\n",
    "    return action, child\n",
    "\n",
    "# The score for a node is based on its value, plus an exploration bonus based on\n",
    "# the prior.\n",
    "# @tf.function\n",
    "def ucb_score(config: MuZeroConfig, parent: Node, child: Node,\n",
    "              min_max_stats: MinMaxStats) -> float:\n",
    "    pb_c = math.log((parent.visit_count + config.pb_c_base + 1) /\n",
    "                  config.pb_c_base) + config.pb_c_init\n",
    "    pb_c *= math.sqrt(parent.visit_count) / (child.visit_count + 1)\n",
    "\n",
    "    # P(s,a)*pb_c\n",
    "    prior_score = pb_c * child.prior\n",
    "    # Q(s,a)\n",
    "    if child.visit_count > 0:\n",
    "        value_score = child.reward + config.discount * min_max_stats.normalize(\n",
    "            child.value())\n",
    "    else:\n",
    "        value_score = 0\n",
    "    return prior_score + value_score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "jh_PczcTxRI_"
   },
   "source": [
    "#### ii. Expansion: Leaf Node Expansion + Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "3r9K4PfZxRI_"
   },
   "outputs": [],
   "source": [
    "# We expand a node using the value, reward and policy prediction obtained from\n",
    "# the neural network.\n",
    "# @tf.function\n",
    "def expand_node(node: Node, actions: List[int], network_output: NetworkOutput):\n",
    "    \"\"\"Updates predictions for state s, reward r and policy p for node based on NN outputs\"\"\"\n",
    "    # Update leaf with predictions from parent\n",
    "    node.hidden_state = network_output.hidden_state # s\n",
    "    node.reward = network_output.reward # r\n",
    "    # policy = {a: tf.math.exp(network_output.policy_logits[a]) for a in actions} # unnormalised probabilities\n",
    "    policy = [tf.math.exp(network_output.policy_logits[a]) for a in actions] # unnormalised probabilities\n",
    "    policy_sum = tf.reduce_sum(policy) \n",
    "    for action in range(len(policy)):\n",
    "        p = policy[action]\n",
    "        node.children[action] = Node(p / policy_sum) # p"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Yh-o6iq9xRJE"
   },
   "source": [
    "#### iii. Backup: Search Tree Update/Backprop\n",
    "$$G^{k}=\\sum_{\\tau=0}^{l-1-k} \\gamma^{\\tau} r_{k+1+\\tau}+\\gamma^{l-k} v^{l}$$\n",
    "For $k = l...1$, we update the statistics for each edge $\\left(s^{k-1}, a^{k}\\right)$ in the simulation\n",
    "$$Q\\left(s^{k-1}, a^{k}\\right):=\\frac{N\\left(s^{k-1}, a^{k}\\right) \\cdot Q\\left(s^{k-1}, a^{k}\\right)+G^{k}}{N\\left(s^{k-1}, a^{k}\\right)+1}$$\n",
    "$$N\\left(s^{k-1}, a^{k}\\right):=N\\left(s^{k-1}, a^{k}\\right)+1$$\n",
    "Q value estimates are normalised so that we can combine value estimates with probabilities via pUCT rule (above)\n",
    "$$\\bar{Q}\\left(s^{k-1}, a^{k}\\right)=\\frac{Q\\left(s^{k-1}, a^{k}\\right)-\\min _{s, a \\in \\operatorname{Tree}} Q(s, a)}{\\max _{s, a \\in \\operatorname{Tree}} Q(s, a)-\\min _{s, a \\in \\operatorname{Tree}} Q(s, a)}$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "N3VBP2LnxRJE"
   },
   "outputs": [],
   "source": [
    "# At the end of a simulation, we propagate the evaluation all the way up the\n",
    "# tree to the root.\n",
    "# @tf.function\n",
    "def backpropagate(search_path: List[Node], value: float,\n",
    "                  discount: float, min_max_stats: MinMaxStats):\n",
    "    # Traverse back up UCB search path\n",
    "    for node in reversed(search_path):\n",
    "        node.value_sum += value # if node.to_play == to_play else -value\n",
    "        node.visit_count += 1\n",
    "        min_max_stats.update(node.value())\n",
    "\n",
    "        value = discount * value + node.reward"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [],
   "source": [
    "##################################\n",
    "####### Part 2: Training #########\n",
    "from tensorflow.keras.optimizers import Adam, Optimizer\n",
    "from tensorflow.compat.v1.train import get_global_step\n",
    "\n",
    "\n",
    "def train_network(config: MuZeroConfig, storage: SharedStorage,\n",
    "                  replay_buffer: ReplayBuffer):\n",
    "#   while len(replay_buffer.buffer)==0: pass\n",
    "  network = Network()\n",
    "  learning_rate = config.lr_init * config.lr_decay_rate**(\n",
    "#       get_global_step() / config.lr_decay_steps)\n",
    "      1)\n",
    "  optimizer = Adam(learning_rate, config.momentum)\n",
    "\n",
    "  for i in range(config.training_steps):\n",
    "    if i % config.checkpoint_interval == 0:\n",
    "      storage.save_network(i, network)\n",
    "    batch = replay_buffer.sample_batch(config.num_unroll_steps, config.td_steps)\n",
    "    update_weights(optimizer, network, batch, config.weight_decay)\n",
    "  storage.save_network(config.training_steps, network)\n",
    "\n",
    "\n",
    "def scale_gradient(tensor, scale):\n",
    "  \"\"\"Scales the gradient for the backward pass.\"\"\"\n",
    "  return tensor * scale + tf.stop_gradient(tensor) * (1 - scale)\n",
    "\n",
    "\n",
    "def update_weights(optimizer: Optimizer, network: Network, batch,\n",
    "                   weight_decay: float):\n",
    "  loss = 0\n",
    "  for image, actions, targets in batch:\n",
    "    # Initial step, from the real observation.\n",
    "    value, reward, policy_logits, hidden_state = network.initial_inference(\n",
    "        image)\n",
    "    predictions = [(1.0, value, reward, policy_logits)]\n",
    "\n",
    "    # Recurrent steps, from action and previous hidden state.\n",
    "    for action in actions:\n",
    "      value, reward, policy_logits, hidden_state = network.recurrent_inference(\n",
    "          hidden_state, action)\n",
    "      predictions.append((1.0 / len(actions), value, reward, policy_logits))\n",
    "\n",
    "      hidden_state = scale_gradient(hidden_state, 0.5)\n",
    "\n",
    "    for prediction, target in zip(predictions, targets):\n",
    "      gradient_scale, value, reward, policy_logits = prediction\n",
    "      target_value, target_reward, target_policy = target\n",
    "\n",
    "      l = (\n",
    "          scalar_loss(value, target_value) +\n",
    "          scalar_loss(reward, target_reward) +\n",
    "          tf.nn.softmax_cross_entropy_with_logits(\n",
    "              logits=policy_logits, labels=target_policy))\n",
    "\n",
    "      loss += scale_gradient(l, gradient_scale)\n",
    "\n",
    "  for weights in network.get_weights():\n",
    "    loss += weight_decay * tf.nn.l2_loss(weights)\n",
    "\n",
    "  optimizer.minimize(loss)\n",
    "\n",
    "\n",
    "def scalar_loss(prediction, target) -> float:\n",
    "  # MSE in board games, cross entropy between categorical values in Atari.\n",
    "  return -1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "8W988LWPxRJG"
   },
   "source": [
    "### Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [],
   "source": [
    "# MuZero training is split into two independent parts: Network training and\n",
    "# self-play data generation.\n",
    "# These two parts only communicate by transferring the latest network checkpoint\n",
    "# from the training to the self-play, and the finished games from the self-play\n",
    "# to the training.\n",
    "class Muzero(object):\n",
    "    def __init__(self, config: MuZeroConfig):\n",
    "        self.config = config\n",
    "        self.storage = SharedStorage()\n",
    "        self.replay_buffer = ReplayBuffer(config)\n",
    "        \n",
    "    def launch_job(self, f, *args):\n",
    "    #   f(*args)\n",
    "        f.remote(*args)\n",
    "    \n",
    "    def run(self):\n",
    "        for _ in range(self.config.num_actors):\n",
    "            self.launch_job(run_selfplay, self.config, self.storage, self.replay_buffer)\n",
    "        train_network(self.config, self.storage, self.replay_buffer)\n",
    "        return self.storage.latest_network()\n",
    "\n",
    "        \n",
    "# def muzero(config: MuZeroConfig):\n",
    "#     storage = SharedStorage()\n",
    "#     replay_buffer = ReplayBuffer(config)\n",
    "\n",
    "#     for _ in range(config.num_actors):\n",
    "#         launch_job(run_selfplay, config, storage, replay_buffer)\n",
    "\n",
    "# #     train_network(config, storage, replay_buffer)\n",
    "\n",
    "#     return storage.latest_network()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [],
   "source": [
    "config = make_atari_config()\n",
    "mz = Muzero(config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-04-30 23:39:17,352\tINFO resource_spec.py:212 -- Starting Ray with 9.08 GiB memory available for workers and up to 4.56 GiB for objects. You can adjust these settings with ray.init(memory=<bytes>, object_store_memory=<bytes>).\n",
      "2020-04-30 23:39:17,639\tINFO services.py:1148 -- View the Ray dashboard at \u001b[1m\u001b[32mlocalhost:8265\u001b[39m\u001b[22m\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'node_ip_address': '192.168.0.24',\n",
       " 'redis_address': '192.168.0.24:65228',\n",
       " 'object_store_address': '/tmp/ray/session_2020-04-30_23-39-17_334462_11196/sockets/plasma_store',\n",
       " 'raylet_socket_name': '/tmp/ray/session_2020-04-30_23-39-17_334462_11196/sockets/raylet',\n",
       " 'webui_url': 'localhost:8265',\n",
       " 'session_dir': '/tmp/ray/session_2020-04-30_23-39-17_334462_11196'}"
      ]
     },
     "execution_count": 173,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ray.init()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Self-play iter:   0%|          | 0/1 [00:00<?, ?it/s]2020-04-30 23:39:28.721783: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m 2020-04-30 23:39:28.756070: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7f93385bfdc0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m 2020-04-30 23:39:28.756118: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version\n",
      "Self-play iter:   0%|          | 0/1 [00:00<?, ?it/s]\n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m 2020-04-30 23:39:28.776501: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA\n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m 2020-04-30 23:39:28.797378: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fa2f3fe3c30 initialized for platform Host (this does not guarantee that XLA will be used). Devices:\n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m 2020-04-30 23:39:28.797421: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version\n",
      "Self-play iter:   0%|          | 0/1 [00:00<?, ?it/s]2020-04-30 23:39:28.950556: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA\n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m 2020-04-30 23:39:28.980927: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fcc9adc03a0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:\n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m 2020-04-30 23:39:28.980978: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version\n",
      "Self-play iter:   0%|          | 0/1 [00:00<?, ?it/s]2020-04-30 23:39:29.299706: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA\n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m 2020-04-30 23:39:29.322882: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fb1f3d20de0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:\n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m 2020-04-30 23:39:29.322927: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m WARNING:tensorflow:Layer conv2d_6 is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m WARNING:tensorflow:Layer conv2d_6 is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m WARNING:tensorflow:Layer conv2d_6 is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m WARNING:tensorflow:Layer conv2d_6 is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m \n",
      "\u001b[2m\u001b[36m(pid=11808)\u001b[0m [<tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01625987], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01730009], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01532874], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01461578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01609246], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0164874], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01778697], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01684255], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688369], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01701035], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01600514], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01716549], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01731716], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01730009], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01707115], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01483417], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01722653], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0165587], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01571291], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01687381], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0166847], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01803908], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01453499], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01679387], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01730007], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01647183], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01754177], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01746436], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01628561], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01690666], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0162466], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01634086], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01690666], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01677313], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0157327], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01750897], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01625987], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01625987], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01701791], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01645399], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01754914], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01765563], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01640906], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01661474], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01687381], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01625987], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01681672], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0150422], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01453499], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01453499], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01677313], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688369], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01659638], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01669119], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01663902], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01754914], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01681672], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01756596], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01634086], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01597648], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01646278], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01706664], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01532874], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01641805], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01658651], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01784716], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01730009], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01571291], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01643873], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01754177], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01677313], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01658651], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01733362], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01641805], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01609246], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01749952], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0180457], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01728865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01707115], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0180457], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01681672], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01646278], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01552402], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0177231], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0162466], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01752926], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0178431], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01681672], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0165587], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01634561], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01634086], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01634086], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01749952], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0150422], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01688579], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01679387], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01679387], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01643873], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01716549], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0154237], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01681672], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01744934], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01559991], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01754914], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01805795], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01765412], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01703234], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01716549], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01679387], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01687382], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01677313], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01617614], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01716765], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01743229], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01735247], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01847727], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01803908], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01461578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01564568], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01680139], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01613126], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01716549], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01641805], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01618196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.01617614], dtype=float32)>]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-04-30 23:41:50,184\tERROR worker.py:1012 -- Possible unhandled error from worker: \u001b[36mray::__main__.run_selfplay()\u001b[39m (pid=11808, ip=192.168.0.24)\n",
      "  File \"python/ray/_raylet.pyx\", line 452, in ray._raylet.execute_task\n",
      "  File \"<ipython-input-132-9d4fc6776c71>\", line 13, in run_selfplay\n",
      "AttributeError: 'Game' object has no attribute 'actions'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[2m\u001b[36m(pid=11801)\u001b[0m [<tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267143], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05259096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05439528], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05447539], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05223686], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05373451], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05373451], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05394021], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05515242], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05284924], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528056], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05515241], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05356538], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05301155], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05501192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05720939], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05444961], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05278027], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528056], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05526311], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05296574], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0542049], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05300927], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05250737], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05511779], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05327417], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0520589], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05394021], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05377366], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267142], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528055], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05243358], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05170518], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05526311], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05242238], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05304968], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05262563], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05253381], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267142], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0542942], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05278027], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05259096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05300928], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05614078], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528056], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528056], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05259097], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427456], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0552452], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05243359], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05288652], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230055], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05300927], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05319211], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05253381], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05346294], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05466395], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528055], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05373903], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05241464], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05507268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0528392], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0528392], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05412662], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05218114], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05515242], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05524519], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05429419], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05390481], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05515242], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05216376], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05207631], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05300928], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05720939], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05284924], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05526311], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05242238], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267142], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05373451], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05399103], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05284924], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05250737], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05288652], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05373451], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05354022], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05466395], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0544925], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05250737], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05182404], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05354023], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05488791], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05319211], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213704], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05188213], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05319211], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05300928], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05429419], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05231567], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05284924], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05399102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05507268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05724717], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05524519], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.052248], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05592478], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05831578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0547742], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05144927], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05250737], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05444961], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05259096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0528392], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05592478], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427457], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0520589], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05412662], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05524519], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267142], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267143], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05346294], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0528392], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05278028], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213704], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267142], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05267143], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05373451], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05501192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0523675], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05331256], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05259096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05445499], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05406022], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0523144], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05515242], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05328261], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05397201], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05427878], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05242238], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213704], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05258296], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05354023], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05376853], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230054], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05214712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05242238], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05198157], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0539048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05274786], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05318582], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05354023], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05520301], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05169303], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0539048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528056], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05429418], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05242238], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0539048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05526311], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05439528], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05223686], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05117945], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05524519], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05354023], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05247488], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05829331], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05216376], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05318582], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.0547742], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05284924], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05223686], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05213704], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05230683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05528056], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([-0.05242238], dtype=float32)>]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-04-30 23:42:05,235\tERROR worker.py:1012 -- Possible unhandled error from worker: \u001b[36mray::__main__.run_selfplay()\u001b[39m (pid=11801, ip=192.168.0.24)\n",
      "  File \"python/ray/_raylet.pyx\", line 452, in ray._raylet.execute_task\n",
      "  File \"<ipython-input-132-9d4fc6776c71>\", line 13, in run_selfplay\n",
      "AttributeError: 'Game' object has no attribute 'actions'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[2m\u001b[36m(pid=11812)\u001b[0m [<tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06555449], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08929123], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06952062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07175034], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06642777], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08070306], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08234865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06755091], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06164849], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0843541], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06623083], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08131976], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825941], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06772219], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0773619], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0654632], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06758095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0666336], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09184889], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06983564], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09552702], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06952062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825942], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08450232], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0654707], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09075077], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06290729], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08242095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09184889], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07576827], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.04602185], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08248479], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06937961], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07967154], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09594928], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08998832], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08351842], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07198076], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06758095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06156943], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0820369], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0673048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06698397], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06679784], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08518815], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0825622], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08157526], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07990912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08813328], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0507804], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07198076], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09005384], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08242095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07780829], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08248479], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06952062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0673048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09596607], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07198076], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06698397], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08933847], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825942], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06735053], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07861236], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08157527], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0692902], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09461212], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0673048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08157526], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.080718], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06952062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08145008], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08998832], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08088902], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06517226], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09098516], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06907479], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08144342], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825941], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07198076], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0843541], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08402412], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08144342], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08873189], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0685567], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08402412], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08813328], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09535591], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06907479], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06164849], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06758095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09252212], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469357], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09146572], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0666336], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07967154], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0593088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09128189], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08183961], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09190791], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06878713], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09552703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06662213], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09058215], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07576827], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06620062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08192685], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825941], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09511055], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06679784], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06179473], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09461214], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09075077], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06766244], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08284626], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08088902], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06517226], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07126797], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06789286], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06711794], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09049548], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08303158], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09081297], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07198076], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07175034], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07701702], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09552703], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0673048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06080843], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08583729], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09224799], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0692902], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0929151], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06643105], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06952062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0692902], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09142662], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07198076], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825942], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.05832552], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.05608454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09594928], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08303158], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0553667], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06745663], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08366659], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08638542], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08242095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09075077], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469357], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06735053], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07887571], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08638542], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09288792], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0654632], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07990912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06642777], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08157526], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06626765], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07576827], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825941], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08482323], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469358], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09150196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08306585], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06247135], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469358], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08731044], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0671498], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08700299], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08608576], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06735834], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08392335], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06262927], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06517226], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08432865], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08760563], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08632438], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06334434], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469357], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0673048], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07070105], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0685567], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09100544], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08485575], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08873189], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09471148], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09150196], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09334303], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07126797], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07377444], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08366127], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06789286], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07329715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09461212], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07175034], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08196004], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08450232], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08518815], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06364395], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06429578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06745662], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08157526], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07175034], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0843541], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08145008], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08204719], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06276324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06758095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469357], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08402412], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07250998], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08955526], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06789286], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07576827], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825942], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0666336], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08469358], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0654707], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06517226], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06758095], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.05737133], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06517226], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0743088], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07064192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06952062], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08813415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09552702], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09128189], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07284111], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08402412], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09634838], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06080843], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0824848], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0919212], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08124746], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08059683], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06335129], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06983564], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825941], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07250998], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0674776], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07985231], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09219534], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825941], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09491239], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09252212], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.08340546], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.09326454], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.0674599], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.06825942], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.07967154], dtype=float32)>]\n",
      "\u001b[2m\u001b[36m(pid=11810)\u001b[0m [<tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3454038], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32905495], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31654432], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34300864], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31565478], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.279644], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32451516], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.24997893], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35423175], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34411898], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33085945], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3156548], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35500127], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33085945], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3411501], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.279644], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3437782], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32905492], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31676635], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32531253], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34411898], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3283805], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32905495], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33784202], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3503166], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31514567], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33182114], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35500127], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.36519527], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3072224], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34422082], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3550013], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.26054323], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3270283], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3156548], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.341051], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35423175], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35338458], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912712], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31505057], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3437782], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3008768], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33182114], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32905495], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33488905], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34383258], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3140381], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30722237], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27282935], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.36442578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3366258], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28138074], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30189204], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373953], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33085945], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.29641297], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2983706], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3160919], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.341051], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30364802], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32905495], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27282938], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3280224], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31403807], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34422082], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3216882], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.26725438], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.341051], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32357582], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31403807], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.330079], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33353692], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33085945], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33007896], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.26826957], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.26826957], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.337842], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2739397], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3072224], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32973817], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.26551762], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30364802], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3282854], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32451516], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2748292], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33662578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34377816], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33353692], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28138074], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33662578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2567799], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32905495], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35423175], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28138074], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3427668], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32913268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33488905], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3411501], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33919984], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31565478], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.279644], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27282935], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.279644], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34411898], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3216882], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31676635], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.341051], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33353692], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3427668], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31006944], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.29338935], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3550013], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30845273], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31676635], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33488905], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.335268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3072224], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.36442578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34411898], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32913268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3550013], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30756322], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27282935], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31654432], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33085945], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27473414], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33919984], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33353692], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.36442578], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31505057], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3282854], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34383258], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2739397], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32255748], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34377816], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3100694], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3411501], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34377816], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28331965], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3442886], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.29641297], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31609192], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33007896], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34377816], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32973817], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3373002], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34460208], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31006944], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.374315], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3503166], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31565478], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.279644], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.279644], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34411898], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3360375], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34383258], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3520324], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32913268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3660425], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27473414], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27282935], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31403807], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32357582], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31654432], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31654432], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3427668], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.335268], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33182114], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32973817], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31676635], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2804912], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2709708], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27473414], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.27282935], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.32357582], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35095415], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.3282854], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31676635], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.34377816], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.31565478], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.323314], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.28738102], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.35389096], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.33488905], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.2747341], dtype=float32)>, <tf.Tensor: shape=(1,), dtype=float32, numpy=array([0.30912715], dtype=float32)>]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2020-04-30 23:43:29,424\tERROR worker.py:1012 -- Possible unhandled error from worker: \u001b[36mray::__main__.run_selfplay()\u001b[39m (pid=11812, ip=192.168.0.24)\n",
      "  File \"python/ray/_raylet.pyx\", line 452, in ray._raylet.execute_task\n",
      "  File \"<ipython-input-132-9d4fc6776c71>\", line 13, in run_selfplay\n",
      "AttributeError: 'Game' object has no attribute 'actions'\n",
      "2020-04-30 23:43:32,531\tERROR worker.py:1012 -- Possible unhandled error from worker: \u001b[36mray::__main__.run_selfplay()\u001b[39m (pid=11810, ip=192.168.0.24)\n",
      "  File \"python/ray/_raylet.pyx\", line 452, in ray._raylet.execute_task\n",
      "  File \"<ipython-input-132-9d4fc6776c71>\", line 13, in run_selfplay\n",
      "AttributeError: 'Game' object has no attribute 'actions'\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-174-cdd65be29286>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mmz\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m<ipython-input-171-c9135bfb6b94>\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m     17\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0m_\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_actors\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     18\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlaunch_job\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_selfplay\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstorage\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreplay_buffer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 19\u001b[0;31m         \u001b[0mtrain_network\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstorage\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreplay_buffer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     20\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstorage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlatest_network\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-170-08433b5cadac>\u001b[0m in \u001b[0;36mtrain_network\u001b[0;34m(config, storage, replay_buffer)\u001b[0m\n\u001b[1;32m      7\u001b[0m def train_network(config: MuZeroConfig, storage: SharedStorage,\n\u001b[1;32m      8\u001b[0m                   replay_buffer: ReplayBuffer):\n\u001b[0;32m----> 9\u001b[0;31m   \u001b[0;32mwhile\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mreplay_buffer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuffer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m==\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mpass\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m   \u001b[0mnetwork\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mNetwork\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m   learning_rate = config.lr_init * config.lr_decay_rate**(\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "mz.run()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 167,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(mz.replay_buffer.buffer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 168,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mz.replay_buffer.buffer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {},
   "outputs": [],
   "source": [
    "ray.shutdown()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7adUOccIoVFY"
   },
   "outputs": [],
   "source": [
    "# !pip install memory-profiler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "3ifGNTeLxRJH",
    "outputId": "d453ec46-4e86-4b5c-c960-c16c7cccf62f"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Self-play iter: 100%|██████████| 1/1 [02:23<00:00, 143.04s/it]\n"
     ]
    }
   ],
   "source": [
    "# with strategy.scope():\n",
    "config = make_atari_config()\n",
    "ss = SharedStorage()\n",
    "rbuf = ReplayBuffer(config)\n",
    "\n",
    "run_selfplay(config, ss, rbuf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Oc_qMcEnxDSs"
   },
   "source": [
    "**ncalls**\n",
    "for the number of calls,\n",
    "\n",
    "**tottime**\n",
    "for the total time spent in the given function (and excluding time made in calls to sub-functions)\n",
    "\n",
    "**percall**\n",
    "is the quotient of tottime divided by ncalls\n",
    "\n",
    "**cumtime**\n",
    "is the cumulative time spent in this and all subfunctions (from invocation till exit). This figure is accurate even for recursive functions.\n",
    "\n",
    "**percall**\n",
    "is the quotient of cumtime divided by primitive calls\n",
    "\n",
    "**filename:lineno(function)**\n",
    "provides the respective data of each function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "kbK6w8jqppng",
    "outputId": "48756d95-b82a-4ecb-d4d6-71f8b9225216"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Self-play iter: 100%|██████████| 1/1 [02:28<00:00, 148.28s/it]\n"
     ]
    }
   ],
   "source": [
    "import cProfile\n",
    "cProfile.run('run_selfplay(config, ss, rbuf)', filename='trace.prof')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 380
    },
    "colab_type": "code",
    "id": "jmD5mDqp4IYR",
    "outputId": "b367ab01-0dfd-46a8-f266-7d0970e47ab3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wed Apr 29 14:37:33 2020    trace.prof\n",
      "\n",
      "         55230475 function calls (55072481 primitive calls) in 157.398 seconds\n",
      "\n",
      "   Ordered by: internal time\n",
      "   List reduced from 2058 to 10 due to restriction <10>\n",
      "\n",
      "   ncalls  tottime  percall  cumtime  percall filename:lineno(function)\n",
      "   456966   52.023    0.000   52.023    0.000 {method '_numpy_internal' of 'tensorflow.python.framework.ops.EagerTensor' objects}\n",
      "  1678974   39.434    0.000   39.434    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_FastPathExecute}\n",
      "    24617    8.728    0.000    8.728    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_Execute}\n",
      "1261332/1261058    4.339    0.000   18.527    0.000 /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py:1290(convert_to_tensor)\n",
      "   746908    4.268    0.000    5.069    0.000 /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py:68(convert_to_eager_tensor)\n",
      "   752389    2.797    0.000   32.070    0.000 /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:981(binary_op_wrapper)\n",
      "   116131    2.625    0.000  106.026    0.001 {built-in method builtins.max}\n",
      "  1399943    2.493    0.000    3.712    0.000 /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py:6189(name_scope)\n",
      "   181460    2.111    0.000   54.528    0.000 <ipython-input-24-9602d5b20ed6>:11(ucb_score)\n",
      "   323012    1.819    0.000   22.952    0.000 /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:1007(r_binary_op_wrapper)\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<pstats.Stats at 0x7f250c6fcba8>"
      ]
     },
     "execution_count": 31,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pstats\n",
    "p = pstats.Stats('trace.prof')\n",
    "p.sort_stats('tottime').print_stats(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "cellView": "form",
    "colab": {},
    "colab_type": "code",
    "id": "1f34wq3bry0f"
   },
   "outputs": [],
   "source": [
    "#@title Default title text\n",
    "trace=\"\"\"ncalls  tottime  percall  cumtime  percall filename:lineno(function)\n",
    "       47    0.000    0.000    0.001    0.000 <__array_function__ internals>:2(all)\n",
    "      294    0.001    0.000    0.011    0.000 <__array_function__ internals>:2(amax)\n",
    "      338    0.001    0.000    0.008    0.000 <__array_function__ internals>:2(any)\n",
    "      297    0.001    0.000    0.017    0.000 <__array_function__ internals>:2(concatenate)\n",
    "        2    0.000    0.000    0.001    0.000 <__array_function__ internals>:2(copyto)\n",
    "      370    0.001    0.000    0.009    0.000 <__array_function__ internals>:2(prod)\n",
    "      294    0.001    0.000    0.027    0.000 <__array_function__ internals>:2(stack)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:103(release)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:143(__init__)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:147(__enter__)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:151(__exit__)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:157(_get_module_lock)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:176(cb)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:194(_lock_unlock_module)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:211(_call_with_frames_removed)\n",
    "        6    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:222(_verbose_message)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:35(_new_module)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:369(__init__)\n",
    "      190    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:391(__eq__)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:403(cached)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:416(parent)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:424(has_location)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:504(_init_module_attrs)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:564(module_from_spec)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:58(__init__)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:78(acquire)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:917(_sanity_check)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:966(_find_and_load)\n",
    "        1    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:982(_gcd_import)\n",
    "      129    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap>:997(_handle_fromlist)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:106(_write_atomic)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:1359(_get_supported_file_loaders)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:263(cache_from_source)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:361(_get_cached)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:373(_calc_mode)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:393(_check_name_wrapper)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:47(_w_long)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:497(_code_to_bytecode)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:524(spec_from_file_location)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:57(_path_join)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:59(<listcomp>)\n",
    "        8    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:63(_path_split)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:661(is_package)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:669(create_module)\n",
    "        2    0.000    0.000    0.001    0.000 <frozen importlib._bootstrap_external>:672(exec_module)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:735(source_to_code)\n",
    "        2    0.000    0.000    0.001    0.000 <frozen importlib._bootstrap_external>:743(get_code)\n",
    "        6    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:75(_path_stat)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:800(__init__)\n",
    "      190    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:806(__eq__)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:825(get_filename)\n",
    "        4    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:830(get_data)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:840(path_stats)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:845(_cache_bytecode)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:85(_path_is_mode_type)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:850(set_data)\n",
    "        2    0.000    0.000    0.000    0.000 <frozen importlib._bootstrap_external>:99(_path_isdir)\n",
    "    71628    0.252    0.000   24.637    0.000 <ipython-input-40-ef54d4819491>:13(update)\n",
    "   102386    1.625    0.000   30.763    0.000 <ipython-input-40-ef54d4819491>:17(normalize)\n",
    "      294    0.000    0.000    0.000    0.000 <ipython-input-40-ef54d4819491>:9(__init__)\n",
    "        1    0.000    0.000    0.201    0.201 <ipython-input-41-b6219d23a42f>:56(new_game)\n",
    "      294    0.000    0.000    0.000    0.000 <ipython-input-41-b6219d23a42f>:61(visit_softmax_temperature)\n",
    "      882    0.000    0.000    0.000    0.000 <ipython-input-42-30c928fcc991>:12(__gt__)\n",
    "   121128    0.042    0.000    0.042    0.000 <ipython-input-42-30c928fcc991>:3(__init__)\n",
    "   259602    0.044    0.000    0.044    0.000 <ipython-input-42-30c928fcc991>:6(__hash__)\n",
    "    77893    0.024    0.000    0.024    0.000 <ipython-input-42-30c928fcc991>:9(__eq__)\n",
    "    71628    0.088    0.000    0.109    0.000 <ipython-input-43-28c03be0c141>:12(expanded)\n",
    "   174308    0.529    0.000   18.052    0.000 <ipython-input-43-28c03be0c141>:15(value)\n",
    "    60270    0.050    0.000    0.050    0.000 <ipython-input-43-28c03be0c141>:3(__init__)\n",
    "    14700    0.027    0.000    0.070    0.000 <ipython-input-44-ef803af33755>:11(clone)\n",
    "    56928    0.079    0.000    0.096    0.000 <ipython-input-44-ef803af33755>:14(add_action)\n",
    "    14700    0.006    0.000    0.006    0.000 <ipython-input-44-ef803af33755>:17(last_action)\n",
    "    14700    0.036    0.000    0.083    0.000 <ipython-input-44-ef803af33755>:20(action_space)\n",
    "    14700    0.034    0.000    0.047    0.000 <ipython-input-44-ef803af33755>:21(<listcomp>)\n",
    "    14994    0.043    0.000    0.043    0.000 <ipython-input-44-ef803af33755>:7(__init__)\n",
    "      295    0.000    0.000    0.000    0.000 <ipython-input-45-8c7d8a00adce>:13(terminal)\n",
    "      294    0.002    0.000    0.003    0.000 <ipython-input-45-8c7d8a00adce>:16(legal_actions)\n",
    "      294    0.001    0.000    0.002    0.000 <ipython-input-45-8c7d8a00adce>:19(<listcomp>)\n",
    "      295    0.025    0.000    0.035    0.000 <ipython-input-45-8c7d8a00adce>:21(prepro)\n",
    "      294    0.001    0.000    0.001    0.000 <ipython-input-45-8c7d8a00adce>:29(get_obs)\n",
    "        1    0.000    0.000    0.201    0.201 <ipython-input-45-8c7d8a00adce>:3(__init__)\n",
    "      294    0.003    0.000    0.254    0.001 <ipython-input-45-8c7d8a00adce>:8(step)\n",
    "      295    0.001    0.000    0.001    0.000 <ipython-input-46-1362272aa653>:12(terminal)\n",
    "      294    0.001    0.000    0.004    0.000 <ipython-input-46-1362272aa653>:15(legal_actions)\n",
    "      294    0.002    0.000    0.256    0.001 <ipython-input-46-1362272aa653>:18(apply)\n",
    "      294    0.003    0.000    0.049    0.000 <ipython-input-46-1362272aa653>:23(store_search_statistics)\n",
    "     1470    0.000    0.000    0.000    0.000 <ipython-input-46-1362272aa653>:25(<genexpr>)\n",
    "     1470    0.001    0.000    0.001    0.000 <ipython-input-46-1362272aa653>:26(<genexpr>)\n",
    "      294    0.002    0.000    0.005    0.000 <ipython-input-46-1362272aa653>:29(<listcomp>)\n",
    "        1    0.000    0.000    0.201    0.201 <ipython-input-46-1362272aa653>:3(__init__)\n",
    "      294    0.006    0.000    0.200    0.001 <ipython-input-46-1362272aa653>:35(make_image)\n",
    "      294    0.001    0.000    0.001    0.000 <ipython-input-46-1362272aa653>:97(action_history)\n",
    "        1    0.000    0.000    0.000    0.000 <ipython-input-47-5f58c4392d1b>:8(save_game)\n",
    "       10    0.000    0.000    2.271    0.227 <ipython-input-48-a3aee8082806>:14(ConvBlock)\n",
    "        1    0.000    0.000    0.221    0.221 <ipython-input-48-a3aee8082806>:20(ReprNet)\n",
    "        1    0.000    0.000    1.955    1.955 <ipython-input-48-a3aee8082806>:28(DynaNet)\n",
    "        1    0.000    0.000    0.149    0.149 <ipython-input-48-a3aee8082806>:39(PredNet)\n",
    "      294    0.008    0.000    0.754    0.003 <ipython-input-53-ed847ba989f4>:16(initial_inference)\n",
    "      294    0.004    0.000    0.005    0.000 <ipython-input-53-ed847ba989f4>:22(<dictcomp>)\n",
    "        1    0.000    0.000    2.325    2.325 <ipython-input-53-ed847ba989f4>:3(__init__)\n",
    "    14700    0.421    0.000   27.850    0.002 <ipython-input-53-ed847ba989f4>:32(recurrent_inference)\n",
    "    14700    0.188    0.000    0.230    0.000 <ipython-input-53-ed847ba989f4>:40(<dictcomp>)\n",
    "      294    0.000    0.000    0.000    0.000 <ipython-input-53-ed847ba989f4>:48(training_steps)\n",
    "        1    0.000    0.000    2.325    2.325 <ipython-input-54-0886533d695a>:16(make_uniform_network)\n",
    "        1    0.000    0.000    2.325    2.325 <ipython-input-54-0886533d695a>:6(latest_network)\n",
    "        1    0.427    0.427  190.490  190.490 <ipython-input-55-37c85f058a24>:1(run_selfplay)\n",
    "        1    0.086    0.086  181.856  181.856 <ipython-input-56-aaae3305ad34>:1(play_game)\n",
    "      294    0.004    0.000    0.021    0.000 <ipython-input-57-fd1ed6f6ed97>:1(add_exploration_noise)\n",
    "      294    0.003    0.000    0.074    0.000 <ipython-input-58-4ec0544e5e1c>:1(select_action)\n",
    "      294    0.008    0.000    0.070    0.000 <ipython-input-58-4ec0544e5e1c>:14(softmax_sample)\n",
    "      294    0.000    0.000    0.000    0.000 <ipython-input-58-4ec0544e5e1c>:15(<listcomp>)\n",
    "      294    0.000    0.000    0.000    0.000 <ipython-input-58-4ec0544e5e1c>:16(<listcomp>)\n",
    "      294    0.006    0.000    0.020    0.000 <ipython-input-58-4ec0544e5e1c>:22(softmax)\n",
    "      294    0.001    0.000    0.001    0.000 <ipython-input-58-4ec0544e5e1c>:6(<listcomp>)\n",
    "      294    0.561    0.002  180.199    0.613 <ipython-input-59-6a2401fc79df>:1(run_mcts)\n",
    "    56928    0.192    0.000  107.099    0.002 <ipython-input-60-c9d2d71ff63b>:1(select_child)\n",
    "   227712    2.436    0.000   67.272    0.000 <ipython-input-60-c9d2d71ff63b>:10(ucb_score)\n",
    "   284640    0.616    0.000   67.888    0.000 <ipython-input-60-c9d2d71ff63b>:4(<genexpr>)\n",
    "    14994    0.118    0.000    0.358    0.000 <ipython-input-61-5f8fc43a9ebe>:1(expand_node)\n",
    "    14994    0.089    0.000    0.158    0.000 <ipython-input-61-5f8fc43a9ebe>:6(<dictcomp>)\n",
    "    14700    1.077    0.000   43.962    0.003 <ipython-input-62-24dec825edc5>:1(backpropagate)\n",
    "        1    0.000    0.000  190.543  190.543 <string>:1(<module>)\n",
    "     1049    0.000    0.000    0.001    0.000 <string>:12(__new__)\n",
    "      128    0.000    0.000    0.000    0.000 <string>:16(_make)\n",
    "       18    0.000    0.000    0.000    0.000 <string>:24(_replace)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:108(import_module)\n",
    "        2    0.000    0.000    0.001    0.001 __init__.py:11(unparse)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:120(getLevelName)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:1310(warning)\n",
    "       36    0.000    0.000    0.000    0.000 __init__.py:1359(log)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:1406(makeRecord)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:1421(_log)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:1446(handle)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:1500(callHandlers)\n",
    "       37    0.000    0.000    0.000    0.000 __init__.py:1530(getEffectiveLevel)\n",
    "       37    0.000    0.000    0.000    0.000 __init__.py:1544(isEnabledFor)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:251(__init__)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:329(getMessage)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:387(usesTime)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:390(format)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:4(_game_dir)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:458(close)\n",
    "        3    0.000    0.000    0.000    0.000 __init__.py:517(__init__)\n",
    "       38    0.000    0.000    0.000    0.000 __init__.py:537(__missing__)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:542(usesTime)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:548(formatMessage)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:564(format)\n",
    "        3    0.000    0.000    0.000    0.000 __init__.py:586(update)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:639(__del__)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:7(get_game_path)\n",
    "        2    0.000    0.000    0.000    0.000 __init__.py:705(filter)\n",
    "        2    0.000    0.000    0.000    0.000 __init__.py:809(acquire)\n",
    "        2    0.000    0.000    0.000    0.000 __init__.py:816(release)\n",
    "        1    0.000    0.000    0.000    0.000 __init__.py:829(format)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:852(handle)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:971(flush)\n",
    "        1    0.000    0.000    0.004    0.004 __init__.py:982(emit)\n",
    "       23    0.000    0.000    0.000    0.000 _asarray.py:16(asarray)\n",
    "      294    0.000    0.000    0.001    0.000 _asarray.py:88(asanyarray)\n",
    "      926    0.001    0.000    0.004    0.000 _collections_abc.py:657(get)\n",
    "      656    0.001    0.000    0.002    0.000 _collections_abc.py:664(__contains__)\n",
    "     2484    0.003    0.000    0.004    0.000 _collections_abc.py:676(items)\n",
    "     2484    0.001    0.000    0.001    0.000 _collections_abc.py:698(__init__)\n",
    "     7585    0.016    0.000    0.016    0.000 _collections_abc.py:742(__iter__)\n",
    "      336    0.000    0.000    0.001    0.000 _collections_abc.py:879(__iter__)\n",
    "       20    0.000    0.000    0.000    0.000 _internal.py:830(npy_ctypes_check)\n",
    "      294    0.000    0.000    0.002    0.000 _methods.py:36(_sum)\n",
    "        3    0.000    0.000    0.000    0.000 _methods.py:44(_any)\n",
    "        1    0.000    0.000    0.000    0.000 _monitor.py:98(report)\n",
    "        6    0.000    0.000    0.000    0.000 _ufunc_config.py:139(geterr)\n",
    "        6    0.000    0.000    0.000    0.000 _ufunc_config.py:39(seterr)\n",
    "        3    0.000    0.000    0.000    0.000 _ufunc_config.py:441(__enter__)\n",
    "        3    0.000    0.000    0.000    0.000 _ufunc_config.py:446(__exit__)\n",
    "        1    0.000    0.000    0.000    0.000 _weakrefset.py:106(remove)\n",
    "      173    0.000    0.000    0.001    0.000 _weakrefset.py:116(update)\n",
    "      909    0.000    0.000    0.000    0.000 _weakrefset.py:16(__init__)\n",
    "      909    0.000    0.000    0.001    0.000 _weakrefset.py:20(__enter__)\n",
    "      909    0.001    0.000    0.001    0.000 _weakrefset.py:26(__exit__)\n",
    "      397    0.000    0.000    0.001    0.000 _weakrefset.py:36(__init__)\n",
    "      227    0.000    0.000    0.000    0.000 _weakrefset.py:38(_remove)\n",
    "      698    0.000    0.000    0.000    0.000 _weakrefset.py:52(_commit_removals)\n",
    "     1231    0.001    0.000    0.003    0.000 _weakrefset.py:58(__iter__)\n",
    "        1    0.000    0.000    0.000    0.000 _weakrefset.py:67(__len__)\n",
    "   432634    0.450    0.000    0.450    0.000 _weakrefset.py:70(__contains__)\n",
    "      351    0.389    0.001    1.716    0.005 _weakrefset.py:81(add)\n",
    "   239505    0.548    0.000    0.998    0.000 abc.py:180(__instancecheck__)\n",
    "      174    0.000    0.000    0.000    0.000 activations.py:338(linear)\n",
    "       23    0.000    0.000    0.000    0.000 activations.py:425(get)\n",
    "     18/6    0.000    0.000    0.000    0.000 activity.py:126(referenced)\n",
    "      404    0.001    0.000    0.002    0.000 activity.py:172(finalize)\n",
    "       55    0.000    0.000    0.000    0.000 activity.py:191(mark_param)\n",
    "       44    0.000    0.000    0.000    0.000 activity.py:210(__init__)\n",
    "       22    0.000    0.000    0.000    0.000 activity.py:224(__init__)\n",
    "      170    0.000    0.000    0.000    0.000 activity.py:232(_in_constructor)\n",
    "      710    0.002    0.000    0.006    0.000 activity.py:251(_track_symbol)\n",
    "      404    0.000    0.000    0.003    0.000 activity.py:303(_enter_scope)\n",
    "      404    0.000    0.000    0.002    0.000 activity.py:306(_exit_scope)\n",
    "      374    0.000    0.000    0.003    0.000 activity.py:312(_exit_and_record_scope)\n",
    "      173    0.000    0.000    0.035    0.000 activity.py:317(_process_statement)\n",
    "        8    0.000    0.000    0.000    0.000 activity.py:334(visit_Expr)\n",
    "        8    0.000    0.000    0.000    0.000 activity.py:337(visit_Raise)\n",
    "       22    0.000    0.000    0.004    0.000 activity.py:340(visit_Return)\n",
    "       95    0.000    0.000    0.022    0.000 activity.py:343(visit_Assign)\n",
    "      540    0.000    0.000    0.012    0.000 activity.py:366(visit_Name)\n",
    "      170    0.000    0.000    0.009    0.000 activity.py:384(visit_Attribute)\n",
    "  139/107    0.000    0.000    0.021    0.000 activity.py:409(visit_Call)\n",
    "       22    0.000    0.000    0.003    0.000 activity.py:477(visit_arguments)\n",
    "       22    0.000    0.000    0.041    0.002 activity.py:499(visit_FunctionDef)\n",
    "       18    0.000    0.000    0.027    0.002 activity.py:533(visit_With)\n",
    "       18    0.000    0.000    0.006    0.000 activity.py:539(visit_withitem)\n",
    "        8    0.000    0.000    0.001    0.000 activity.py:581(visit_ExceptHandler)\n",
    "       22    0.000    0.000    0.042    0.002 activity.py:592(resolve)\n",
    "      426    0.001    0.000    0.003    0.000 activity.py:92(__init__)\n",
    "       10    0.000    0.000    0.004    0.000 advanced_activations.py:344(__init__)\n",
    "      118    0.000    0.000    0.063    0.001 advanced_activations.py:360(call)\n",
    "      146    0.000    0.000    0.000    0.000 ag_ctx.py:29(_control_ctx)\n",
    "       38    0.000    0.000    0.000    0.000 ag_ctx.py:35(control_status_ctx)\n",
    "       36    0.000    0.000    0.000    0.000 ag_ctx.py:49(__init__)\n",
    "       36    0.000    0.000    0.000    0.000 ag_ctx.py:53(__enter__)\n",
    "       36    0.000    0.000    0.000    0.000 ag_ctx.py:61(__exit__)\n",
    "      270    0.000    0.000    0.002    0.000 ag_logging.py:114(get_verbosity)\n",
    "      270    0.000    0.000    0.002    0.000 ag_logging.py:121(has_verbosity)\n",
    "      231    0.000    0.000    0.002    0.000 ag_logging.py:138(log)\n",
    "        3    0.000    0.000    0.000    0.000 ale_python_interface.py:115(_as_bytes)\n",
    "        1    0.000    0.000    0.000    0.000 ale_python_interface.py:127(__init__)\n",
    "        1    0.000    0.000    0.000    0.000 ale_python_interface.py:141(setInt)\n",
    "        1    0.000    0.000    0.000    0.000 ale_python_interface.py:145(setFloat)\n",
    "        1    0.188    0.188    0.188    0.188 ale_python_interface.py:148(loadROM)\n",
    "      880    0.152    0.000    0.152    0.000 ale_python_interface.py:151(act)\n",
    "      294    0.001    0.000    0.001    0.000 ale_python_interface.py:154(game_over)\n",
    "        1    0.010    0.010    0.010    0.010 ale_python_interface.py:157(reset_game)\n",
    "        1    0.000    0.000    0.000    0.000 ale_python_interface.py:196(getMinimalActionSet)\n",
    "      294    0.000    0.000    0.000    0.000 ale_python_interface.py:205(lives)\n",
    "        1    0.000    0.000    0.000    0.000 ale_python_interface.py:211(getScreenDims)\n",
    "      295    0.023    0.000    0.050    0.000 ale_python_interface.py:251(getScreenRGB2)\n",
    "        3    0.000    0.000    0.000    0.000 ale_python_interface.py:353(__del__)\n",
    "     7010    0.004    0.000    0.008    0.000 anno.py:111(getanno)\n",
    "    28849    0.014    0.000    0.029    0.000 anno.py:119(hasanno)\n",
    "     3155    0.004    0.000    0.006    0.000 anno.py:123(setanno)\n",
    "     5010    0.002    0.000    0.007    0.000 anno.py:141(copyanno)\n",
    "        2    0.000    0.000    0.001    0.000 anno.py:150(dup)\n",
    "       35    0.000    0.000    0.000    0.000 api.py:126(__init__)\n",
    "     1835    0.001    0.000    0.001    0.000 api.py:129(get_effective_source_map)\n",
    "       35    0.000    0.000    0.000    0.000 api.py:155(autograph_artifact)\n",
    "       38    0.000    0.000    0.000    0.000 api.py:160(is_autograph_artifact)\n",
    "       53    0.000    0.000    1.279    0.024 api.py:337(_call_unconverted)\n",
    "      114    0.000    0.000    0.000    0.000 api.py:351(_is_known_loaded_type)\n",
    "    88/18    0.002    0.000    1.561    0.087 api.py:378(converted_call)\n",
    "      228    0.001    0.000    0.002    0.000 api.py:480(<genexpr>)\n",
    "        2    0.000    0.000    0.000    0.000 arg_defaults.py:72(visit_FunctionDef)\n",
    "        2    0.000    0.000    0.000    0.000 arg_defaults.py:79(visit_arguments)\n",
    "        2    0.000    0.000    0.000    0.000 arg_defaults.py:94(transform)\n",
    "    44688    0.095    0.000    1.588    0.000 array_ops.py:1041(strided_slice)\n",
    "   134064    0.106    0.000    2.791    0.000 array_ops.py:1281(stack)\n",
    "   134388    0.158    0.000    0.583    0.000 array_ops.py:1432(_should_not_autopack)\n",
    "   270398    0.085    0.000    0.085    0.000 array_ops.py:1438(<genexpr>)\n",
    "   134388    0.067    0.000    0.650    0.000 array_ops.py:1442(_autopacking_conversion_function)\n",
    "       43    0.000    0.000    0.003    0.000 array_ops.py:198(fill)\n",
    "      226    0.001    0.000    0.107    0.000 array_ops.py:239(identity)\n",
    "       23    0.000    0.000    0.003    0.000 array_ops.py:2676(wrapped)\n",
    "       23    0.000    0.000    0.003    0.000 array_ops.py:2684(zeros)\n",
    "       20    0.000    0.000    0.003    0.000 array_ops.py:2934(ones)\n",
    "        3    0.000    0.000    0.003    0.001 array_ops.py:2984(placeholder)\n",
    "       10    0.000    0.000    0.010    0.001 array_ops.py:3029(placeholder_with_default)\n",
    "      311    0.001    0.000    0.164    0.001 array_ops.py:3144(pad_v2)\n",
    "      311    0.003    0.000    0.163    0.001 array_ops.py:3202(pad)\n",
    "   221/17    0.000    0.000    0.000    0.000 array_ops.py:3297(_get_paddings_constant)\n",
    "    85/17    0.000    0.000    0.000    0.000 array_ops.py:3314(<listcomp>)\n",
    "      294    0.001    0.000    0.011    0.000 array_ops.py:347(expand_dims_v2)\n",
    "       20    0.000    0.000    0.011    0.001 array_ops.py:4093(squeeze)\n",
    "       20    0.000    0.000    0.013    0.001 array_ops.py:4148(squeeze_v2)\n",
    "       37    0.000    0.000    0.023    0.001 array_ops.py:58(reshape)\n",
    "    44688    0.073    0.000    0.227    0.000 array_ops.py:823(_check_index)\n",
    "    44688    0.787    0.000    6.693    0.000 array_ops.py:843(_slice_helper)\n",
    "      385    0.000    0.000    0.001    0.000 ast.py:118(copy_location)\n",
    "    57878    0.015    0.000    0.020    0.000 ast.py:166(iter_fields)\n",
    "      256    0.000    0.000    0.000    0.000 ast.py:178(iter_child_nodes)\n",
    "      131    0.000    0.000    0.001    0.000 ast.py:215(walk)\n",
    "18632/170    0.013    0.000    0.231    0.001 ast.py:249(visit)\n",
    " 2676/149    0.005    0.000    0.014    0.000 ast.py:255(generic_visit)\n",
    "       35    0.000    0.000    0.002    0.000 ast.py:30(parse)\n",
    "14149/335    0.033    0.000    0.177    0.001 ast.py:302(generic_visit)\n",
    "       35    0.000    0.000    0.011    0.000 ast3.py:10(visit_Module)\n",
    "    43/38    0.000    0.000    0.004    0.000 ast3.py:114(visit_Call)\n",
    "       34    0.000    0.000    0.000    0.000 ast3.py:139(visit_NameConstant)\n",
    "       12    0.000    0.000    0.001    0.000 ast3.py:149(visit_arguments)\n",
    "      213    0.000    0.000    0.003    0.000 ast3.py:162(visit_Name)\n",
    "       16    0.000    0.000    0.000    0.000 ast3.py:17(visit_Num)\n",
    "       22    0.000    0.000    0.000    0.000 ast3.py:172(visit_arg)\n",
    "        4    0.000    0.000    0.000    0.000 ast3.py:187(visit_ExceptHandler)\n",
    "       29    0.000    0.000    0.000    0.000 ast3.py:218(visit_Constant)\n",
    "       23    0.000    0.000    0.000    0.000 ast3.py:234(_make_arg)\n",
    "       83    0.000    0.000    0.001    0.000 ast3.py:250(visit_Name)\n",
    "        2    0.000    0.000    0.000    0.000 ast3.py:258(visit_ExceptHandler)\n",
    "      6/2    0.000    0.000    0.002    0.001 ast3.py:310(visit_FunctionDef)\n",
    "        4    0.000    0.000    0.000    0.000 ast3.py:34(visit_Str)\n",
    "        2    0.000    0.000    0.001    0.001 ast3.py:353(visit_With)\n",
    "    19/15    0.000    0.000    0.001    0.000 ast3.py:369(visit_Call)\n",
    "        6    0.000    0.000    0.000    0.000 ast3.py:378(visit_arguments)\n",
    "        6    0.000    0.000    0.000    0.000 ast3.py:380(<listcomp>)\n",
    "        6    0.000    0.000    0.000    0.000 ast3.py:392(<listcomp>)\n",
    "       35    0.000    0.000    0.011    0.000 ast3.py:398(ast_to_gast)\n",
    "        2    0.000    0.000    0.002    0.001 ast3.py:402(gast_to_ast)\n",
    "     12/6    0.000    0.000    0.006    0.001 ast3.py:50(visit_FunctionDef)\n",
    "        4    0.000    0.000    0.003    0.001 ast3.py:96(visit_With)\n",
    "      395    0.002    0.000    0.003    0.000 ast_util.py:278(parallel_walk)\n",
    "       82    0.000    0.000    0.000    0.000 ast_util.py:34(__init__)\n",
    "  2017/82    0.006    0.000    0.017    0.000 ast_util.py:38(copy)\n",
    "   228/41    0.000    0.000    0.014    0.000 ast_util.py:42(<listcomp>)\n",
    "        5    0.000    0.000    0.006    0.001 ast_util.py:44(<genexpr>)\n",
    "       82    0.000    0.000    0.017    0.000 ast_util.py:64(copy_clean)\n",
    "   476/43    0.000    0.000    0.013    0.000 astn.py:11(<listcomp>)\n",
    "  876/180    0.002    0.000    0.011    0.000 astn.py:17(generic_visit)\n",
    "  2710/45    0.002    0.000    0.013    0.000 astn.py:9(_visit)\n",
    "      294    0.004    0.000    0.214    0.001 atari_env.py:111(step)\n",
    "      295    0.001    0.000    0.051    0.000 atari_env.py:125(_get_image)\n",
    "      295    0.001    0.000    0.052    0.000 atari_env.py:135(_get_obs)\n",
    "        1    0.000    0.000    0.010    0.010 atari_env.py:143(reset)\n",
    "        1    0.000    0.000    0.191    0.191 atari_env.py:26(__init__)\n",
    "        1    0.000    0.000    0.188    0.188 atari_env.py:83(seed)\n",
    "     4962    0.003    0.000    0.008    0.000 auto_control_deps.py:121(op_is_stateful)\n",
    "       93    0.000    0.000    0.000    0.000 auto_control_deps.py:148(__init__)\n",
    "      186    0.001    0.000    0.091    0.000 auto_control_deps.py:152(mark_as_return)\n",
    "       93    0.000    0.000    0.002    0.000 auto_control_deps.py:195(__enter__)\n",
    "       93    0.018    0.000    0.284    0.003 auto_control_deps.py:270(__exit__)\n",
    "     2481    0.000    0.000    0.000    0.000 auto_control_deps.py:391(<listcomp>)\n",
    "       10    0.000    0.000    0.000    0.000 auto_control_deps.py:401(<listcomp>)\n",
    "     2481    0.012    0.000    0.241    0.000 auto_control_deps.py:452(_get_resource_inputs)\n",
    "     2481    0.002    0.000    0.003    0.000 auto_control_deps.py:471(<listcomp>)\n",
    "     2481    0.001    0.000    0.002    0.000 auto_control_deps.py:472(<listcomp>)\n",
    "     1484    0.001    0.000    0.003    0.000 auto_control_deps_utils.py:102(<genexpr>)\n",
    "     1719    0.002    0.000    0.004    0.000 auto_control_deps_utils.py:109(<genexpr>)\n",
    "       20    0.000    0.000    0.002    0.000 auto_control_deps_utils.py:33(get_read_only_resource_input_indices_graph)\n",
    "       80    0.000    0.000    0.000    0.000 auto_control_deps_utils.py:51(<listcomp>)\n",
    "       80    0.000    0.000    0.000    0.000 auto_control_deps_utils.py:60(_get_read_only_resource_input_indices_op)\n",
    "       80    0.000    0.000    0.000    0.000 auto_control_deps_utils.py:63(<listcomp>)\n",
    "     2481    0.011    0.000    0.186    0.000 auto_control_deps_utils.py:86(get_read_write_resource_inputs)\n",
    "        3    0.000    0.000    0.004    0.001 backend.py:1024(placeholder)\n",
    "       36    0.000    0.000    0.000    0.000 backend.py:1091(is_placeholder)\n",
    "       10    0.000    0.000    0.011    0.001 backend.py:1110(freezable_variable)\n",
    "       84    0.000    0.000    0.001    0.000 backend.py:1181(int_shape)\n",
    "       20    0.000    0.000    0.000    0.000 backend.py:169(cast_to_floatx)\n",
    "        3    0.000    0.000    0.000    0.000 backend.py:207(get_uid)\n",
    "       71    0.000    0.000    0.005    0.000 backend.py:304(learning_phase)\n",
    "       71    0.000    0.000    0.000    0.000 backend.py:334(global_learning_phase_is_set)\n",
    "       71    0.000    0.000    0.000    0.000 backend.py:338(_mark_func_graph_as_unsaveable)\n",
    "      118    0.004    0.000    0.062    0.001 backend.py:4400(relu)\n",
    "     1989    0.002    0.000    0.007    0.000 backend.py:531(get_graph)\n",
    "       38    0.000    0.000    0.000    0.000 backend.py:598(get_default_graph_uid_map)\n",
    "       71    0.000    0.000    0.000    0.000 backend.py:6123(_key)\n",
    "       81    0.000    0.000    0.000    0.000 backend.py:6142(_get_recursive)\n",
    "       81    0.000    0.000    0.000    0.000 backend.py:6155(__getitem__)\n",
    "       56    0.000    0.000    0.000    0.000 backend.py:725(is_sparse)\n",
    "      483    0.001    0.000    0.001    0.000 backend.py:776(name_scope)\n",
    "       56    0.000    0.000    0.000    0.000 backend.py:862(track_variable)\n",
    "       38    0.000    0.000    0.000    0.000 backend.py:870(unique_object_name)\n",
    "       12    0.000    0.000    0.000    0.000 backend_config.py:112(image_data_format)\n",
    "       61    0.000    0.000    0.000    0.000 backend_config.py:65(floatx)\n",
    "     1856    0.002    0.000    0.004    0.000 backprop.py:165(_must_record_gradient)\n",
    "2512/1932    0.005    0.000    0.040    0.000 base.py:452(_method_wrapper)\n",
    "      310    0.000    0.000    0.001    0.000 base.py:576(_maybe_initialize_trackable)\n",
    "      122    0.000    0.000    0.000    0.000 base.py:655(_deferred_dependencies)\n",
    "      122    0.000    0.000    0.000    0.000 base.py:670(_lookup_dependency)\n",
    "       56    0.001    0.000    0.061    0.001 base.py:684(_add_variable_with_custom_getter)\n",
    "       56    0.000    0.000    0.000    0.000 base.py:756(_preload_simple_restoration)\n",
    "      122    0.001    0.000    0.003    0.000 base.py:793(_track_trackable)\n",
    "       66    0.000    0.000    0.001    0.000 base.py:847(_handle_deferred_dependencies)\n",
    "      396    0.000    0.000    0.000    0.000 base_layer.py:1005(trainable)\n",
    "      482    0.000    0.000    0.000    0.000 base_layer.py:1025(input_spec)\n",
    "       38    0.000    0.000    0.000    0.000 base_layer.py:1060(input_spec)\n",
    "   519/71    0.001    0.000    0.012    0.000 base_layer.py:1291(_clear_losses)\n",
    "       20    0.000    0.000    0.006    0.000 base_layer.py:1380(add_update)\n",
    "       20    0.000    0.000    0.004    0.000 base_layer.py:1433(process_update)\n",
    "       20    0.000    0.000    0.004    0.000 base_layer.py:1459(<listcomp>)\n",
    "       20    0.000    0.000    0.000    0.000 base_layer.py:1463(<listcomp>)\n",
    "        3    0.000    0.000    0.000    0.000 base_layer.py:1792(input)\n",
    "       41    0.000    0.000    0.000    0.000 base_layer.py:1916(inbound_nodes)\n",
    "       35    0.000    0.000    0.000    0.000 base_layer.py:1922(outbound_nodes)\n",
    "       41    0.000    0.000    0.002    0.000 base_layer.py:1983(_set_dtype_policy)\n",
    "     1022    0.001    0.000    0.001    0.000 base_layer.py:2012(_compute_dtype)\n",
    "      483    0.002    0.000    0.014    0.000 base_layer.py:2025(_maybe_cast_inputs)\n",
    "      447    0.003    0.000    0.005    0.000 base_layer.py:2040(f)\n",
    "      483    0.000    0.000    0.000    0.000 base_layer.py:2099(_name_scope)\n",
    "       41    0.000    0.000    0.002    0.000 base_layer.py:2102(_init_set_name)\n",
    "      483    0.000    0.000    0.000    0.000 base_layer.py:2201(_handle_activity_regularization)\n",
    "      483    0.003    0.000    0.013    0.000 base_layer.py:2218(_set_mask_metadata)\n",
    "      894    0.000    0.000    0.001    0.000 base_layer.py:2223(<genexpr>)\n",
    "       36    0.000    0.000    0.000    0.000 base_layer.py:2233(<listcomp>)\n",
    "      273    0.000    0.000    0.000    0.000 base_layer.py:2235(<listcomp>)\n",
    "      174    0.000    0.000    0.000    0.000 base_layer.py:2241(<listcomp>)\n",
    "      483    0.002    0.000    0.012    0.000 base_layer.py:2258(_collect_input_masks)\n",
    "      483    0.000    0.000    0.000    0.000 base_layer.py:2266(<lambda>)\n",
    "      966    0.002    0.000    0.004    0.000 base_layer.py:2272(_call_arg_was_passed)\n",
    "      108    0.000    0.000    0.000    0.000 base_layer.py:2283(_get_call_arg_value)\n",
    "       35    0.000    0.000    0.005    0.000 base_layer.py:2293(_set_connectivity_metadata_)\n",
    "       35    0.000    0.000    0.004    0.000 base_layer.py:2317(_add_inbound_node)\n",
    "       35    0.000    0.000    0.000    0.000 base_layer.py:2329(<lambda>)\n",
    "       35    0.000    0.000    0.000    0.000 base_layer.py:2331(<lambda>)\n",
    "       35    0.000    0.000    0.000    0.000 base_layer.py:2333(<lambda>)\n",
    "        3    0.000    0.000    0.000    0.000 base_layer.py:2358(_get_node_attribute_at_index)\n",
    "      483    0.001    0.000    1.800    0.004 base_layer.py:2394(_maybe_build)\n",
    "       70    0.000    0.000    0.000    0.000 base_layer.py:2408(<genexpr>)\n",
    "       35    0.000    0.000    0.000    0.000 base_layer.py:2409(<lambda>)\n",
    "     41/3    0.000    0.000    0.001    0.000 base_layer.py:2446(_get_trainable_state)\n",
    "     1312    0.002    0.000    0.008    0.000 base_layer.py:2469(_obj_reference_counts)\n",
    "     1557    0.001    0.000    0.002    0.000 base_layer.py:2476(_maybe_create_attribute)\n",
    "      656    0.002    0.000    0.009    0.000 base_layer.py:2493(__delattr__)\n",
    "7486/4292    0.011    0.000    1.751    0.000 base_layer.py:2543(__setattr__)\n",
    "       10    0.000    0.000    0.000    0.000 base_layer.py:2579(<genexpr>)\n",
    "       85    0.000    0.000    0.000    0.000 base_layer.py:2604(<genexpr>)\n",
    "       50    0.000    0.000    0.000    0.000 base_layer.py:2608(<genexpr>)\n",
    "       41    0.000    0.000    0.000    0.000 base_layer.py:2654(_attribute_sentinel)\n",
    "       41    0.000    0.000    0.004    0.000 base_layer.py:2665(_init_call_fn_args)\n",
    "       41    0.000    0.000    0.003    0.000 base_layer.py:2677(_call_full_argspec)\n",
    "       41    0.000    0.000    0.003    0.000 base_layer.py:2684(_call_fn_args)\n",
    "       38    0.000    0.000    0.000    0.000 base_layer.py:2693(_call_accepts_kwargs)\n",
    "       38    0.000    0.000    0.000    0.000 base_layer.py:2698(_should_compute_mask)\n",
    "      519    0.001    0.000    0.001    0.000 base_layer.py:2716(_eager_losses)\n",
    "       41    0.001    0.000    0.010    0.000 base_layer.py:278(__init__)\n",
    "       35    0.000    0.000    0.000    0.000 base_layer.py:386(build)\n",
    "       56    0.001    0.000    0.063    0.001 base_layer.py:440(add_weight)\n",
    "      174    0.000    0.000    0.000    0.000 base_layer.py:741(compute_mask)\n",
    "   483/71    0.018    0.000    3.512    0.049 base_layer.py:763(__call__)\n",
    "      966    0.001    0.000    0.001    0.000 base_layer.py:812(<genexpr>)\n",
    "      966    0.001    0.000    0.003    0.000 base_layer.py:887(<genexpr>)\n",
    "      142    0.000    0.000    0.000    0.000 base_layer.py:976(dtype)\n",
    "     1154    0.000    0.000    0.000    0.000 base_layer.py:981(name)\n",
    "       38    0.000    0.000    0.000    0.000 base_layer.py:986(dynamic)\n",
    "       56    0.000    0.000    0.011    0.000 base_layer_utils.py:121(<lambda>)\n",
    "      483    0.001    0.000    0.005    0.000 base_layer_utils.py:160(have_all_keras_metadata)\n",
    "      966    0.000    0.000    0.001    0.000 base_layer_utils.py:161(<genexpr>)\n",
    "      483    0.001    0.000    0.013    0.000 base_layer_utils.py:272(needs_keras_history)\n",
    "      142    0.000    0.000    0.000    0.000 base_layer_utils.py:294(<genexpr>)\n",
    "     1069    0.001    0.000    0.008    0.000 base_layer_utils.py:301(is_in_keras_graph)\n",
    "      986    0.001    0.000    0.043    0.000 base_layer_utils.py:306(is_in_eager_or_tf_function)\n",
    "      951    0.003    0.000    0.040    0.000 base_layer_utils.py:311(is_in_tf_function)\n",
    "       36    0.002    0.000    0.011    0.000 base_layer_utils.py:329(uses_keras_history)\n",
    "       36    0.000    0.000    0.001    0.000 base_layer_utils.py:374(mark_checked)\n",
    "       36    0.000    0.000    0.000    0.000 base_layer_utils.py:384(_mark_checked)\n",
    "     2091    0.001    0.000    0.002    0.000 base_layer_utils.py:390(call_context)\n",
    "      966    0.002    0.000    0.005    0.000 base_layer_utils.py:420(enter)\n",
    "     1069    0.002    0.000    0.007    0.000 base_layer_utils.py:453(in_keras_graph)\n",
    "      483    0.001    0.000    0.004    0.000 base_layer_utils.py:471(autocast_context_manager)\n",
    "      483    0.001    0.000    0.001    0.000 base_layer_utils.py:488(is_subclassed)\n",
    "       56    0.001    0.000    0.057    0.001 base_layer_utils.py:51(make_variable)\n",
    "       35    0.000    0.000    0.019    0.001 base_layer_utils.py:627(mark_as_return)\n",
    "       35    0.000    0.000    0.018    0.001 base_layer_utils.py:630(_mark_as_return)\n",
    "      117    0.000    0.000    0.000    0.000 base_layer_utils.py:703(v2_dtype_behavior_enabled)\n",
    "        1    0.000    0.000    0.002    0.002 box.py:24(__init__)\n",
    "        3    0.000    0.000    0.000    0.000 box.py:39(_get_precision)\n",
    "        2    0.000    0.000    0.002    0.001 break_statements.py:152(transform)\n",
    "     3788    0.003    0.000    0.008    0.000 c_api_util.py:157(tf_buffer)\n",
    "     3144    0.012    0.000    0.012    0.000 c_api_util.py:190(tf_output)\n",
    "       60    0.000    0.000    0.002    0.000 c_api_util.py:45(__init__)\n",
    "      544    0.001    0.000    0.229    0.000 c_api_util.py:53(__del__)\n",
    "       58    0.000    0.000    0.000    0.000 c_api_util.py:86(__init__)\n",
    "      538    0.002    0.000    0.090    0.000 c_api_util.py:94(__del__)\n",
    "        2    0.000    0.000    0.009    0.004 call_trees.py:111(visit_FunctionDef)\n",
    "        2    0.000    0.000    0.008    0.004 call_trees.py:141(visit_With)\n",
    "        6    0.000    0.000    0.000    0.000 call_trees.py:146(_args_to_tuple)\n",
    "        6    0.000    0.000    0.001    0.000 call_trees.py:164(_kwargs_to_dict)\n",
    "    12/11    0.000    0.000    0.007    0.001 call_trees.py:175(visit_Call)\n",
    "        2    0.000    0.000    0.009    0.004 call_trees.py:226(transform)\n",
    "        2    0.000    0.000    0.000    0.000 call_trees.py:43(__init__)\n",
    "        6    0.000    0.000    0.000    0.000 call_trees.py:58(__init__)\n",
    "        6    0.000    0.000    0.000    0.000 call_trees.py:63(_consume_args)\n",
    "        5    0.000    0.000    0.000    0.000 call_trees.py:69(add_arg)\n",
    "        6    0.000    0.000    0.000    0.000 call_trees.py:81(finalize)\n",
    "        6    0.000    0.000    0.000    0.000 call_trees.py:85(to_ast)\n",
    "       44    0.000    0.000    0.001    0.000 cfg.py:175(__init__)\n",
    "       44    0.000    0.000    0.001    0.000 cfg.py:202(reset)\n",
    "       44    0.000    0.000    0.000    0.000 cfg.py:203(<dictcomp>)\n",
    "       44    0.000    0.000    0.000    0.000 cfg.py:206(<dictcomp>)\n",
    "       44    0.001    0.000    0.016    0.000 cfg.py:210(_visit_internal)\n",
    "       22    0.000    0.000    0.012    0.001 cfg.py:234(visit_forward)\n",
    "       22    0.000    0.000    0.004    0.000 cfg.py:237(visit_reverse)\n",
    "       22    0.000    0.000    0.000    0.000 cfg.py:277(__init__)\n",
    "       44    0.000    0.000    0.000    0.000 cfg.py:281(reset)\n",
    "      151    0.000    0.000    0.000    0.000 cfg.py:322(_connect_nodes)\n",
    "      173    0.001    0.000    0.001    0.000 cfg.py:337(_add_new_node)\n",
    "       16    0.000    0.000    0.000    0.000 cfg.py:361(begin_statement)\n",
    "       16    0.000    0.000    0.000    0.000 cfg.py:370(end_statement)\n",
    "      143    0.000    0.000    0.001    0.000 cfg.py:380(add_ordinary_node)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:395(_add_jump_node)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:414(_connect_jump_to_finally_sections)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:427(add_exit_node)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:458(connect_raise_node)\n",
    "       22    0.000    0.000    0.000    0.000 cfg.py:473(enter_section)\n",
    "       22    0.000    0.000    0.000    0.000 cfg.py:485(exit_section)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:528(enter_cond_section)\n",
    "       16    0.000    0.000    0.000    0.000 cfg.py:542(new_cond_branch)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:556(exit_cond_section)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:563(enter_except_section)\n",
    "       22    0.000    0.000    0.002    0.000 cfg.py:588(build)\n",
    "       22    0.000    0.000    0.000    0.000 cfg.py:642(__init__)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:651(_enter_lexical_scope)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:654(_exit_lexical_scope)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:658(_get_enclosing_finally_scopes)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:667(_get_enclosing_except_scopes)\n",
    "      143    0.000    0.000    0.010    0.000 cfg.py:676(_process_basic_statement)\n",
    "       30    0.000    0.000    0.000    0.000 cfg.py:680(_process_exit_statement)\n",
    "      173    0.000    0.000    0.000    0.000 cfg.py:73(__init__)\n",
    "       22    0.000    0.000    0.013    0.001 cfg.py:730(visit_FunctionDef)\n",
    "       22    0.000    0.000    0.000    0.000 cfg.py:754(visit_Return)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:763(visit_Expr)\n",
    "       95    0.000    0.000    0.006    0.000 cfg.py:766(visit_Assign)\n",
    "      173    0.000    0.000    0.001    0.000 cfg.py:78(freeze)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:787(visit_Raise)\n",
    "        8    0.000    0.000    0.000    0.000 cfg.py:879(visit_ExceptHandler)\n",
    "        8    0.000    0.000    0.001    0.000 cfg.py:893(visit_Try)\n",
    "       18    0.000    0.000    0.008    0.000 cfg.py:935(visit_With)\n",
    "       22    0.000    0.000    0.013    0.001 cfg.py:943(build)\n",
    "        2    0.000    0.000    0.000    0.000 codecs.py:185(__init__)\n",
    "    19851    0.010    0.000    0.021    0.000 compat.py:114(as_str)\n",
    "      246    0.000    0.000    0.000    0.000 compat.py:39(_date_to_date_number)\n",
    "      246    0.000    0.000    0.000    0.000 compat.py:63(forward_compatible)\n",
    "     6209    0.007    0.000    0.012    0.000 compat.py:64(as_bytes)\n",
    "    19851    0.008    0.000    0.011    0.000 compat.py:90(as_text)\n",
    "        4    0.000    0.000    0.000    0.000 compilerop.py:137(check_linecache_ipython)\n",
    "       20    0.000    0.000    0.004    0.000 cond_v2.py:1127(_set_read_only_resource_inputs_attr)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:1147(<listcomp>)\n",
    "       20    0.001    0.000    0.076    0.004 cond_v2.py:191(_build_cond)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:257(<listcomp>)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:260(<listcomp>)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:269(<listcomp>)\n",
    "       20    0.000    0.000    0.019    0.001 cond_v2.py:295(<listcomp>)\n",
    "       20    0.000    0.000    0.001    0.000 cond_v2.py:496(_make_inputs_match)\n",
    "       40    0.000    0.000    0.000    0.000 cond_v2.py:524(<listcomp>)\n",
    "       20    0.001    0.000    0.310    0.015 cond_v2.py:58(cond_v2)\n",
    "       20    0.000    0.000    0.001    0.000 cond_v2.py:619(_make_indexed_slices_indices_types_match)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:629(<listcomp>)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:632(<listcomp>)\n",
    "      120    0.000    0.000    0.000    0.000 cond_v2.py:637(<genexpr>)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:683(_get_op_and_outputs)\n",
    "       20    0.000    0.000    0.001    0.000 cond_v2.py:692(_pack_sequence_as)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:760(_check_same_outputs)\n",
    "       20    0.000    0.000    0.001    0.000 cond_v2.py:804(_get_output_shapes)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:814(verify_captures)\n",
    "       20    0.000    0.000    0.000    0.000 cond_v2.py:820(<dictcomp>)\n",
    "        2    0.000    0.000    0.003    0.002 conditional_expressions.py:44(transform)\n",
    "      805    0.000    0.000    0.001    0.000 config_lib.py:33(matches)\n",
    "      768    0.000    0.000    0.001    0.000 config_lib.py:50(get_action)\n",
    "       37    0.000    0.000    0.000    0.000 config_lib.py:62(get_action)\n",
    "   615293    0.545    0.000    7.645    0.000 constant_op.py:164(constant)\n",
    "   615293    0.953    0.000    7.101    0.000 constant_op.py:265(_constant_impl)\n",
    "   134388    0.070    0.000    1.363    0.000 constant_op.py:318(_constant_tensor_conversion_function)\n",
    "       43    0.000    0.000    0.001    0.000 constant_op.py:330(_tensor_shape_tensor_conversion_function)\n",
    "   615191    4.950    0.000    5.772    0.000 constant_op.py:68(convert_to_eager_tensor)\n",
    "      122    0.000    0.000    0.000    0.000 constraints.py:280(get)\n",
    "    15014    0.010    0.000    0.010    0.000 context.py:1021(function_call_options)\n",
    "       58    0.000    0.000    0.005    0.000 context.py:1050(add_function)\n",
    "      538    0.002    0.000    0.543    0.001 context.py:1096(remove_function)\n",
    "     2382    0.002    0.000    0.002    0.000 context.py:1146(op_callbacks)\n",
    "    15014    0.004    0.000    0.004    0.000 context.py:127(executor_type)\n",
    "    29988    0.008    0.000    0.008    0.000 context.py:135(config_proto_serialized)\n",
    "        2    0.000    0.000    0.000    0.000 context.py:1562(disable_run_metadata)\n",
    "        1    0.000    0.000    2.249    2.249 context.py:1583(export_run_metadata)\n",
    "     9026    0.002    0.000    0.002    0.000 context.py:1601(context_switches)\n",
    "        1    0.000    0.000    0.000    0.000 context.py:1616(__init__)\n",
    "        1    0.000    0.000    0.000    0.000 context.py:1621(__enter__)\n",
    "        1    0.000    0.000    0.000    0.000 context.py:1655(__exit__)\n",
    "  2599334    0.638    0.000    0.638    0.000 context.py:1702(context)\n",
    "   123774    0.034    0.000    0.034    0.000 context.py:1709(context_safe)\n",
    "    15053    0.026    0.000    0.043    0.000 context.py:1714(ensure_initialized)\n",
    "       33    0.000    0.000    0.000    0.000 context.py:1724(global_seed)\n",
    "   123658    0.196    0.000    0.334    0.000 context.py:1734(executing_eagerly)\n",
    "       56    0.000    0.000    0.000    0.000 context.py:1864(shared_name)\n",
    "     2119    0.002    0.000    0.006    0.000 context.py:1885(graph_mode)\n",
    "     1596    0.002    0.000    0.005    0.000 context.py:1890(eager_mode)\n",
    "        1    0.000    0.000    0.000    0.000 context.py:1900(device)\n",
    "     3715    0.005    0.000    0.010    0.000 context.py:218(push)\n",
    "       58    0.000    0.000    0.005    0.000 context.py:2230(add_function)\n",
    "      538    0.001    0.000    0.545    0.001 context.py:2235(remove_function)\n",
    "      116    0.000    0.000    0.000    0.000 context.py:2248(_tmp_in_graph_mode)\n",
    "     3715    0.003    0.000    0.004    0.000 context.py:238(pop)\n",
    "   646128    0.286    0.000    0.286    0.000 context.py:492(ensure_initialized)\n",
    "   631075    0.213    0.000    0.213    0.000 context.py:723(_handle)\n",
    "     7430    0.008    0.000    0.013    0.000 context.py:747(_mode)\n",
    "  2710312    1.143    0.000    1.143    0.000 context.py:765(executing_eagerly)\n",
    "    44845    0.014    0.000    0.014    0.000 context.py:777(scope_name)\n",
    "    89688    0.047    0.000    0.047    0.000 context.py:782(scope_name)\n",
    "   630480    0.216    0.000    0.216    0.000 context.py:787(device_name)\n",
    "        2    0.000    0.000    0.000    0.000 context.py:792(device_spec)\n",
    "        2    0.000    0.000    0.000    0.000 context.py:797(_set_device)\n",
    "        1    0.000    0.000    0.000    0.000 context.py:801(device)\n",
    "     2119    0.005    0.000    0.022    0.000 contextlib.py:129(contextmanager)\n",
    "    18838    0.012    0.000    0.043    0.000 contextlib.py:157(helper)\n",
    "        3    0.000    0.000    0.000    0.000 contextlib.py:36(_recreate_cm)\n",
    "        3    0.000    0.000    0.000    0.000 contextlib.py:49(inner)\n",
    "    18838    0.026    0.000    0.031    0.000 contextlib.py:59(__init__)\n",
    "18838/11274    0.008    0.000    0.159    0.000 contextlib.py:79(__enter__)\n",
    "18838/11297    0.017    0.000    0.075    0.000 contextlib.py:85(__exit__)\n",
    "        2    0.000    0.000    0.001    0.001 continue_statements.py:117(_visit_non_loop_body)\n",
    "        2    0.000    0.000    0.002    0.001 continue_statements.py:143(visit_With)\n",
    "        2    0.000    0.000    0.002    0.001 continue_statements.py:161(transform)\n",
    "        2    0.000    0.000    0.000    0.000 continue_statements.py:29(__init__)\n",
    "        4    0.000    0.000    0.000    0.000 continue_statements.py:52(__init__)\n",
    "        7    0.000    0.000    0.000    0.000 continue_statements.py:77(_postprocess_statement)\n",
    "        2    0.000    0.000    0.003    0.002 control_flow.py:49(visit_FunctionDef)\n",
    "        2    0.000    0.000    0.003    0.002 control_flow.py:528(transform)\n",
    "       20    0.000    0.000    0.310    0.015 control_flow_ops.py:1098(cond)\n",
    "     2423    0.001    0.000    0.002    0.000 control_flow_util.py:180(GetOutputContext)\n",
    "     2481    0.001    0.000    0.001    0.000 control_flow_util.py:191(GetContainingWhileContext)\n",
    "       80    0.000    0.000    0.000    0.000 control_flow_util.py:213(GetContainingXLAContext)\n",
    "     2423    0.003    0.000    0.005    0.000 control_flow_util.py:266(CheckInputFromValidContext)\n",
    "       20    0.000    0.000    0.000    0.000 control_flow_util.py:50(EnableControlFlowV2)\n",
    "       80    0.000    0.000    0.000    0.000 control_flow_util.py:68(InXlaContext)\n",
    "       40    0.000    0.000    0.000    0.000 control_flow_util.py:73(GraphOrParentsInXlaContext)\n",
    "     2481    0.002    0.000    0.004    0.000 control_flow_util.py:82(IsInWhileLoop)\n",
    "       20    0.000    0.000    0.000    0.000 control_flow_util_v2.py:113(maybe_propagate_compile_time_consts_in_xla)\n",
    "       20    0.000    0.000    0.000    0.000 control_flow_util_v2.py:231(output_all_intermediates)\n",
    "       20    0.000    0.000    0.000    0.000 control_flow_util_v2.py:42(in_defun)\n",
    "       40    0.000    0.000    0.017    0.000 control_flow_util_v2.py:53(create_new_tf_function)\n",
    "       40    0.000    0.000    0.000    0.000 control_flow_util_v2.py:68(unique_fn_name)\n",
    "       20    0.000    0.000    0.001    0.000 control_flow_util_v2.py:85(maybe_set_lowering_attr)\n",
    "       12    0.000    0.000    0.000    0.000 conv_utils.py:189(normalize_data_format)\n",
    "       10    0.000    0.000    0.000    0.000 conv_utils.py:200(normalize_padding)\n",
    "       10    0.000    0.000    0.000    0.000 conv_utils.py:28(convert_data_format)\n",
    "       30    0.000    0.000    0.000    0.000 conv_utils.py:51(normalize_tuple)\n",
    "      123    0.001    0.000    0.002    0.000 conversion.py:119(has)\n",
    "       38    0.000    0.000    0.000    0.000 conversion.py:125(__getitem__)\n",
    "       70    0.000    0.000    0.000    0.000 conversion.py:147(_get_key)\n",
    "       91    0.000    0.000    0.000    0.000 conversion.py:163(_get_key)\n",
    "        2    0.000    0.000    0.012    0.006 conversion.py:182(_wrap_into_dynamic_factory)\n",
    "       35    0.000    0.000    0.264    0.008 conversion.py:254(_convert_with_cache)\n",
    "       35    0.001    0.000    0.001    0.000 conversion.py:298(_instantiate)\n",
    "       35    0.000    0.000    0.000    0.000 conversion.py:316(<genexpr>)\n",
    "       35    0.000    0.000    0.266    0.008 conversion.py:340(convert)\n",
    "    54/20    0.001    0.000    0.004    0.000 conversion.py:366(is_whitelisted)\n",
    "       88    0.000    0.000    0.001    0.000 conversion.py:471(is_in_whitelist_cache)\n",
    "        3    0.000    0.000    0.000    0.000 conversion.py:479(cache_whitelisted)\n",
    "        2    0.000    0.000    0.231    0.115 conversion.py:488(convert_entity_to_ast)\n",
    "        2    0.000    0.000    0.000    0.000 conversion.py:635(_add_reserved_symbol)\n",
    "        2    0.000    0.000    0.000    0.000 conversion.py:646(_add_self_references)\n",
    "        2    0.000    0.000    0.231    0.115 conversion.py:669(convert_func_to_ast)\n",
    "        2    0.000    0.000    0.216    0.108 conversion.py:729(node_to_graph)\n",
    "       35    0.000    0.000    0.000    0.000 conversion.py:90(get_module)\n",
    "       35    0.000    0.000    0.000    0.000 conversion.py:93(get_factory)\n",
    "       71    0.000    0.000    0.000    0.000 converter.py:158(__init__)\n",
    "      357    0.000    0.000    0.000    0.000 converter.py:175(as_tuple)\n",
    "      121    0.000    0.000    0.000    0.000 converter.py:179(__hash__)\n",
    "      118    0.000    0.000    0.000    0.000 converter.py:182(__eq__)\n",
    "       74    0.000    0.000    0.000    0.000 converter.py:189(uses)\n",
    "       35    0.000    0.000    0.000    0.000 converter.py:193(call_options)\n",
    "        2    0.000    0.000    0.002    0.001 converter.py:201(to_ast)\n",
    "        2    0.000    0.000    0.000    0.000 converter.py:221(list_of_features)\n",
    "        2    0.000    0.000    0.000    0.000 converter.py:223(<genexpr>)\n",
    "        2    0.000    0.000    0.000    0.000 converter.py:268(__init__)\n",
    "       22    0.000    0.000    0.000    0.000 converter.py:282(__init__)\n",
    "  2124/22    0.002    0.000    0.049    0.002 converter.py:337(visit)\n",
    "      201    0.000    0.000    0.000    0.000 converter.py:352(__init__)\n",
    "       22    0.000    0.000    0.166    0.008 converter.py:357(standard_analysis)\n",
    "       20    0.000    0.000    0.205    0.010 converter.py:387(apply_)\n",
    "       10    0.000    0.000    0.009    0.001 convolutional.py:102(__init__)\n",
    "       10    0.000    0.000    0.025    0.003 convolutional.py:151(build)\n",
    "      118    0.001    0.000    0.213    0.002 convolutional.py:193(call)\n",
    "       20    0.000    0.000    0.000    0.000 convolutional.py:284(_get_channel_axis)\n",
    "       10    0.000    0.000    0.000    0.000 convolutional.py:290(_get_input_channel)\n",
    "       10    0.000    0.000    0.000    0.000 convolutional.py:297(_get_padding_op)\n",
    "      118    0.001    0.000    0.003    0.000 convolutional.py:306(_recreate_conv_op)\n",
    "       10    0.000    0.000    0.010    0.001 convolutional.py:565(__init__)\n",
    "        3    0.000    0.000    0.002    0.001 core.py:1116(__init__)\n",
    "        3    0.000    0.000    0.007    0.002 core.py:1147(build)\n",
    "       56    0.001    0.000    0.178    0.003 core.py:1179(call)\n",
    "        1    0.000    0.000    0.000    0.000 core.py:132(unwrapped)\n",
    "        1    0.000    0.000    0.000    0.000 core.py:208(__init__)\n",
    "        2    0.000    0.000    0.001    0.000 core.py:629(__init__)\n",
    "       37    0.001    0.000    0.046    0.001 core.py:634(call)\n",
    "       37    0.000    0.000    0.002    0.000 core.py:669(compute_output_shape)\n",
    "      296    0.016    0.000    0.016    0.000 ctypeslib.py:350(_ctype_ndarray)\n",
    "      296    0.002    0.000    0.003    0.000 ctypeslib.py:376(_ctype_from_dtype_scalar)\n",
    "      296    0.001    0.000    0.004    0.000 ctypeslib.py:455(_ctype_from_dtype)\n",
    "      296    0.001    0.000    0.005    0.000 ctypeslib.py:464(as_ctypes_type)\n",
    "      296    0.005    0.000    0.026    0.000 ctypeslib.py:526(as_ctypes)\n",
    "      802    0.011    0.000    0.039    0.000 custom_gradient.py:44(copy_handle_data)\n",
    "      782    0.003    0.000    0.003    0.000 custom_gradient.py:74(<listcomp>)\n",
    "      782    0.001    0.000    0.001    0.000 custom_gradient.py:76(<listcomp>)\n",
    "      782    0.001    0.000    0.002    0.000 custom_gradient.py:77(<listcomp>)\n",
    "  696/676    0.001    0.000    0.004    0.000 data_structures.py:103(sticky_attribute_assignment)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:140(__init__)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:152(__init__)\n",
    "       30    0.000    0.000    0.000    0.000 data_structures.py:161(_attribute_sentinel)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:173(_track_value)\n",
    "      118    0.000    0.000    0.001    0.000 data_structures.py:195(_layers)\n",
    "      118    0.000    0.000    0.002    0.000 data_structures.py:208(layers)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:309(__init__)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:330(_name_element)\n",
    "      118    0.000    0.000    0.000    0.000 data_structures.py:333(_values)\n",
    "      336    0.000    0.000    0.000    0.000 data_structures.py:377(__getitem__)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:383(__len__)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:414(__init__)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:431(_non_append_mutation)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:447(_external_modification)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:475(_make_storage)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:479(_check_external_modification)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:487(_update_snapshot)\n",
    "       10    0.000    0.000    0.000    0.000 data_structures.py:525(__setitem__)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:576(__eq__)\n",
    "       20    0.000    0.000    0.000    0.000 data_structures.py:614(_track_value)\n",
    "       33    0.000    0.000    0.000    0.000 data_structures.py:66(_should_wrap_tuple)\n",
    "  768/676    0.001    0.000    0.002    0.000 data_structures.py:80(wrap_or_unwrap)\n",
    "     2481    0.002    0.000    0.007    0.000 dataset_ops.py:4530(_resource_resolver)\n",
    "        2    0.000    0.000    0.000    0.000 def_function.py:318(__init__)\n",
    "        6    0.000    0.000    0.000    0.000 def_function.py:321(__del__)\n",
    "        2    0.000    0.000    0.000    0.000 def_function.py:338(__init__)\n",
    "        4    0.000    0.000    0.001    0.000 def_function.py:420(_defun_with_scope)\n",
    "       18    0.000    0.000    1.562    0.087 def_function.py:424(wrapped_fn)\n",
    "        4    0.000    0.000    0.001    0.000 def_function.py:448(_defun)\n",
    "        2    0.000    0.000    0.437    0.218 def_function.py:472(_initialize)\n",
    "        2    0.000    0.000    0.000    0.000 def_function.py:53(__init__)\n",
    "    29988    0.045    0.000    0.045    0.000 def_function.py:555(_get_tracing_count)\n",
    "    14994    0.099    0.000   17.339    0.001 def_function.py:560(__call__)\n",
    "       18    0.000    0.000    0.000    0.000 def_function.py:58(called_with_tracing)\n",
    "    14994    0.075    0.000   17.126    0.001 def_function.py:602(_call)\n",
    "    14976    0.026    0.000    0.026    0.000 def_function.py:68(called_without_tracing)\n",
    "        8    0.000    0.000    0.000    0.000 def_function.py:710(python_function)\n",
    "        6    0.000    0.000    0.000    0.000 def_function.py:715(input_signature)\n",
    "       18    0.000    0.000    0.000    0.000 def_function.py:76(get_tracing_count)\n",
    "    14994    0.058    0.000    0.121    0.000 def_function.py:963(__get__)\n",
    "   116/96    0.001    0.000    0.323    0.003 deprecation.py:473(new_func)\n",
    "      138    0.000    0.000    0.000    0.000 deprecation.py:583(deprecated_argument_lookup)\n",
    "        1    0.000    0.000    0.000    0.000 device.py:45(is_device_spec)\n",
    "      118    0.000    0.000    0.000    0.000 device_context.py:21(enclosing_tpu_context)\n",
    "       41    0.000    0.000    0.001    0.000 directives.py:117(visit_Name)\n",
    "       14    0.000    0.000    0.001    0.000 directives.py:126(visit_Attribute)\n",
    "        5    0.000    0.000    0.001    0.000 directives.py:134(visit_Assign)\n",
    "        2    0.000    0.000    0.002    0.001 directives.py:177(transform)\n",
    "        2    0.000    0.000    0.000    0.000 directives.py:50(__init__)\n",
    "        1    0.000    0.000    0.000    0.000 discrete.py:13(__init__)\n",
    "134726/134706    0.081    0.000    3.013    0.000 dispatch.py:177(wrapper)\n",
    "      486    0.000    0.000    0.000    0.000 distribute_lib.py:2677(value_container)\n",
    "      486    0.000    0.000    0.000    0.000 distribute_lib.py:627(extended)\n",
    "      144    0.000    0.000    0.001    0.000 distribution_strategy_context.py:179(get_strategy)\n",
    "      141    0.000    0.000    0.001    0.000 distribution_strategy_context.py:199(has_strategy)\n",
    "      486    0.001    0.000    0.003    0.000 distribution_strategy_context.py:215(get_strategy_and_replica_context)\n",
    "      141    0.000    0.000    0.000    0.000 distribution_strategy_context.py:312(_get_default_strategy)\n",
    "      630    0.000    0.000    0.000    0.000 distribution_strategy_context.py:337(_get_default_replica_mode)\n",
    "      630    0.002    0.000    0.004    0.000 distribution_strategy_context.py:79(_get_per_thread_mode)\n",
    "       74    0.000    0.000    0.000    0.000 dtypes.py:103(as_numpy_dtype)\n",
    "   309793    0.664    0.000    1.100    0.000 dtypes.py:172(is_compatible_with)\n",
    "   941272    1.326    0.000    1.329    0.000 dtypes.py:192(__eq__)\n",
    "   452980    0.282    0.000    0.842    0.000 dtypes.py:205(__ne__)\n",
    "   805476    0.365    0.000    0.482    0.000 dtypes.py:606(as_dtype)\n",
    "  1699685    1.610    0.000    1.610    0.000 dtypes.py:71(_is_ref_dtype)\n",
    "  1699685    0.844    0.000    2.454    0.000 dtypes.py:84(base_dtype)\n",
    "      597    0.001    0.000    0.001    0.000 enum.py:267(__call__)\n",
    "        2    0.000    0.000    0.000    0.000 enum.py:297(__contains__)\n",
    "      597    0.000    0.000    0.000    0.000 enum.py:517(__new__)\n",
    "    25780    0.007    0.000    0.010    0.000 enum.py:581(__hash__)\n",
    "     1719    0.001    0.000    0.002    0.000 errors_impl.py:117(__str__)\n",
    "     1719    0.002    0.000    0.005    0.000 errors_impl.py:264(__init__)\n",
    "     1719    0.003    0.000    0.003    0.000 errors_impl.py:64(__init__)\n",
    "     1719    0.000    0.000    0.000    0.000 errors_impl.py:86(message)\n",
    "      256    0.000    0.000    0.001    0.000 execute.py:159(make_float)\n",
    "      944    0.001    0.000    0.001    0.000 execute.py:166(make_int)\n",
    "      420    0.000    0.000    0.001    0.000 execute.py:177(make_str)\n",
    "      359    0.000    0.000    0.000    0.000 execute.py:184(make_bool)\n",
    "      786    0.001    0.000    0.003    0.000 execute.py:191(make_type)\n",
    "       53    0.000    0.000    0.000    0.000 execute.py:201(make_shape)\n",
    "      588    0.002    0.000    0.019    0.000 execute.py:236(args_to_matching_eager)\n",
    "      294    0.000    0.000    0.001    0.000 execute.py:271(<listcomp>)\n",
    "    15288    0.055    0.000   12.284    0.001 execute.py:33(quick_execute)\n",
    "        1    0.000    0.000    0.000    0.000 ezpickle.py:20(__init__)\n",
    "      338    0.000    0.000    0.000    0.000 fromnumeric.py:2232(_any_dispatcher)\n",
    "      338    0.001    0.000    0.006    0.000 fromnumeric.py:2236(any)\n",
    "       47    0.000    0.000    0.000    0.000 fromnumeric.py:2320(_all_dispatcher)\n",
    "       47    0.000    0.000    0.001    0.000 fromnumeric.py:2324(all)\n",
    "      294    0.000    0.000    0.000    0.000 fromnumeric.py:2546(_amax_dispatcher)\n",
    "      294    0.002    0.000    0.008    0.000 fromnumeric.py:2551(amax)\n",
    "      370    0.000    0.000    0.000    0.000 fromnumeric.py:2838(_prod_dispatcher)\n",
    "      370    0.001    0.000    0.007    0.000 fromnumeric.py:2843(prod)\n",
    "     1049    0.005    0.000    0.019    0.000 fromnumeric.py:73(_wrapreduction)\n",
    "     1049    0.001    0.000    0.001    0.000 fromnumeric.py:74(<dictcomp>)\n",
    "       58    0.000    0.000    0.001    0.000 func_graph.py:1011(<listcomp>)\n",
    "      209    0.000    0.000    0.000    0.000 func_graph.py:1018(<genexpr>)\n",
    "     2063    0.003    0.000    0.008    0.000 func_graph.py:1048(device_stack_has_callable)\n",
    "     2063    0.001    0.000    0.002    0.000 func_graph.py:1050(<genexpr>)\n",
    "      116    0.000    0.000    0.001    0.000 func_graph.py:1054(check_mutation)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:1074(flatten)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:1086(<listcomp>)\n",
    "       20    0.000    0.000    0.001    0.000 func_graph.py:1091(pack_sequence_as)\n",
    "      802    0.005    0.000    0.310    0.000 func_graph.py:1116(_create_substitute_placeholder)\n",
    "       58    0.000    0.000    0.014    0.000 func_graph.py:1129(_get_defun_inputs_from_args)\n",
    "       35    0.000    0.000    0.000    0.000 func_graph.py:1135(_get_composite_tensor_spec)\n",
    "      116    0.001    0.000    0.016    0.000 func_graph.py:1141(_get_defun_inputs)\n",
    "       35    0.000    0.000    0.000    0.000 func_graph.py:1192(<listcomp>)\n",
    "       35    0.000    0.000    0.000    0.000 func_graph.py:1194(<listcomp>)\n",
    "       58    0.000    0.000    0.001    0.000 func_graph.py:1258(_get_defun_inputs_from_kwargs)\n",
    "      544    0.006    0.000    0.560    0.001 func_graph.py:1269(dismantle_func_graph)\n",
    "       18    0.000    0.000    0.000    0.000 func_graph.py:127(<listcomp>)\n",
    "      116    0.000    0.000    0.002    0.000 func_graph.py:130(<listcomp>)\n",
    "       60    0.001    0.000    0.008    0.000 func_graph.py:165(__init__)\n",
    "      742    0.002    0.000    0.009    0.000 func_graph.py:265(watch_variable)\n",
    "      822    0.001    0.000    0.003    0.000 func_graph.py:321(control_dependencies)\n",
    "     2119    0.006    0.000    0.060    0.000 func_graph.py:362(as_default)\n",
    "     4238    0.025    0.000    0.123    0.000 func_graph.py:365(inner_cm)\n",
    "     1368    0.001    0.000    0.001    0.000 func_graph.py:429(outer_graph)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:493(variables)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:495(<listcomp>)\n",
    "2484/2464    0.019    0.000    0.717    0.000 func_graph.py:530(_create_op_internal)\n",
    "     3356    0.006    0.000    0.330    0.000 func_graph.py:597(capture)\n",
    "      802    0.004    0.000    0.319    0.000 func_graph.py:646(_capture_helper)\n",
    "      802    0.002    0.000    0.002    0.000 func_graph.py:664(add_capture)\n",
    "       40    0.000    0.000    0.000    0.000 func_graph.py:678(reset_captures)\n",
    "      116    0.000    0.000    0.007    0.000 func_graph.py:69(convert_structure_to_signature)\n",
    "      544    0.007    0.000    0.016    0.000 func_graph.py:692(clear_captures)\n",
    "       98    0.000    0.000    0.000    0.000 func_graph.py:730(external_captures)\n",
    "       98    0.000    0.000    0.000    0.000 func_graph.py:733(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:735(internal_captures)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:738(<listcomp>)\n",
    "       18    0.000    0.000    0.000    0.000 func_graph.py:740(deferred_external_captures)\n",
    "       18    0.000    0.000    0.000    0.000 func_graph.py:743(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:745(deferred_internal_captures)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:748(<listcomp>)\n",
    "       58    0.003    0.000    2.048    0.035 func_graph.py:796(func_graph_from_py_func)\n",
    "       35    0.000    0.000    0.002    0.000 func_graph.py:82(encode_arg)\n",
    "       58    0.000    0.000    0.000    0.000 func_graph.py:914(<listcomp>)\n",
    "      151    0.001    0.000    0.084    0.001 func_graph.py:925(convert)\n",
    "       18    0.000    0.000    1.561    0.087 func_graph.py:953(wrapper)\n",
    "       36    0.000    0.000    0.000    0.000 func_graph.py:97(<genexpr>)\n",
    "44688/29988    0.027    0.000    0.123    0.000 function.py:105(<lambda>)\n",
    "       18    0.000    0.000    0.029    0.002 function.py:1498(__init__)\n",
    "       89    0.000    0.000    0.002    0.000 function.py:1551(<genexpr>)\n",
    "    14994    0.073    0.000   14.068    0.001 function.py:1647(_filtered_call)\n",
    "    29988    0.033    0.000    0.047    0.000 function.py:1662(<genexpr>)\n",
    "    14994    0.195    0.000   13.851    0.001 function.py:1667(_call_flat)\n",
    "    14994    0.049    0.000    0.073    0.000 function.py:1811(captured_inputs)\n",
    "    14994    0.004    0.000    0.004    0.000 function.py:1817(<listcomp>)\n",
    "    14994    0.132    0.000    0.879    0.000 function.py:1954(_build_call_outputs)\n",
    "        6    0.000    0.000    0.001    0.000 function.py:1998(from_function_and_signature)\n",
    "        6    0.000    0.000    0.000    0.000 function.py:2081(__init__)\n",
    "        6    0.000    0.000    0.000    0.000 function.py:2100(<dictcomp>)\n",
    "        6    0.000    0.000    0.000    0.000 function.py:2106(<dictcomp>)\n",
    "    30032    0.011    0.000    0.011    0.000 function.py:2142(input_signature)\n",
    "    14996    0.054    0.000    0.257    0.000 function.py:2155(canonicalize_function_inputs)\n",
    "    14996    0.069    0.000    0.169    0.000 function.py:2242(_convert_numpy_inputs)\n",
    "        4    0.000    0.000    0.000    0.000 function.py:2320(__init__)\n",
    "        4    0.000    0.000    0.001    0.000 function.py:2358(__init__)\n",
    "    14992    0.062    0.000   16.544    0.001 function.py:2416(__call__)\n",
    "    30026    0.022    0.000    0.033    0.000 function.py:2431(input_signature)\n",
    "        2    0.000    0.000    0.436    0.218 function.py:2441(_get_concrete_function_internal_garbage_collected)\n",
    "    14994    0.143    0.000    0.621    0.000 function.py:2570(_cache_key)\n",
    "       18    0.000    0.000    1.862    0.103 function.py:2639(_create_graph_function)\n",
    "       18    0.000    0.000    0.000    0.000 function.py:2653(<listcomp>)\n",
    "    14994    0.081    0.000    2.904    0.000 function.py:2711(_maybe_define_function)\n",
    "       18    0.000    0.000    0.000    0.000 function.py:277(_parse_func_attrs)\n",
    "    14994    0.009    0.000    0.009    0.000 function.py:313(__init__)\n",
    "        4    0.000    0.000    0.001    0.000 function.py:3159(defun_with_attributes)\n",
    "    14994    0.003    0.000    0.003    0.000 function.py:316(__enter__)\n",
    "    14994    0.011    0.000    0.011    0.000 function.py:319(__exit__)\n",
    "        4    0.000    0.000    0.001    0.000 function.py:3196(decorated)\n",
    "        2    0.000    0.000    0.000    0.000 function.py:3240(__init__)\n",
    "       18    0.000    0.000    0.000    0.000 function.py:3244(target)\n",
    "        2    0.000    0.000    0.000    0.000 function.py:3264(class_method_to_instance_method)\n",
    "       18    0.000    0.000    1.561    0.087 function.py:3282(bound_method_wrapper)\n",
    "       12    0.000    0.000    0.000    0.000 function.py:3324(__init__)\n",
    "       36    0.145    0.004    1.365    0.038 function.py:3327(__del__)\n",
    "       18    0.000    0.000    0.000    0.000 function.py:3341(__init__)\n",
    "      538    0.002    0.000    0.562    0.001 function.py:3348(__del__)\n",
    "       18    0.000    0.000    0.000    0.000 function.py:404(_inference_name)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:429(__init__)\n",
    "      538    0.001    0.000    0.546    0.001 function.py:432(__del__)\n",
    "       58    0.003    0.000    0.040    0.001 function.py:460(__init__)\n",
    "      822    0.000    0.000    0.000    0.000 function.py:470(<genexpr>)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:471(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:489(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:490(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:491(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:493(<listcomp>)\n",
    "       58    0.000    0.000    0.000    0.000 function.py:522(<listcomp>)\n",
    "       58    0.000    0.000    0.001    0.000 function.py:523(<listcomp>)\n",
    "      784    0.000    0.000    0.000    0.000 function.py:532(<genexpr>)\n",
    "       40    0.000    0.000    0.002    0.000 function.py:537(add_to_graph)\n",
    "      138    0.000    0.000    0.000    0.000 function.py:549(name)\n",
    "    14994    0.260    0.000   12.551    0.001 function.py:557(call)\n",
    "       18    0.000    0.000    0.026    0.001 function.py:645(__init__)\n",
    "       18    0.000    0.000    0.000    0.000 function.py:807(forward)\n",
    "59682/14994    0.109    0.000    0.184    0.000 function.py:87(_make_input_signature_hashable)\n",
    "        2    0.000    0.000    0.011    0.005 function_scopes.py:131(transform)\n",
    "        4    0.000    0.000    0.000    0.000 function_scopes.py:31(__init__)\n",
    "        2    0.000    0.000    0.002    0.001 function_scopes.py:38(visit_Return)\n",
    "        2    0.000    0.000    0.000    0.000 function_scopes.py:46(_function_scope_options)\n",
    "        2    0.000    0.000    0.011    0.005 function_scopes.py:90(visit_FunctionDef)\n",
    "       35    0.000    0.000    0.001    0.000 function_wrappers.py:43(__init__)\n",
    "       35    0.000    0.000    0.000    0.000 function_wrappers.py:70(__enter__)\n",
    "       35    0.000    0.000    0.000    0.000 function_wrappers.py:79(__exit__)\n",
    "       35    0.000    0.000    0.000    0.000 function_wrappers.py:87(mark_return_value)\n",
    "     2121    0.008    0.000    0.015    0.000 functools.py:44(update_wrapper)\n",
    "     2121    0.002    0.000    0.002    0.000 functools.py:74(wraps)\n",
    "     2015    0.004    0.000    0.005    0.000 gast.py:15(create_node)\n",
    "       35    0.000    0.000    0.012    0.000 gast.py:297(parse)\n",
    "      125    0.000    0.000    0.000    0.000 gast.py:321(copy_location)\n",
    "      294    0.001    0.000    0.010    0.000 gen_array_ops.py:2211(expand_dims)\n",
    "       43    0.000    0.000    0.002    0.000 gen_array_ops.py:3265(fill)\n",
    "      226    0.001    0.000    0.105    0.000 gen_array_ops.py:3873(identity)\n",
    "      303    0.003    0.000    0.142    0.000 gen_array_ops.py:6383(pad)\n",
    "      294    0.003    0.000    0.082    0.000 gen_array_ops.py:6449(pad_eager_fallback)\n",
    "        8    0.000    0.000    0.013    0.002 gen_array_ops.py:6463(pad_v2)\n",
    "        3    0.000    0.000    0.003    0.001 gen_array_ops.py:6637(placeholder)\n",
    "       10    0.000    0.000    0.010    0.001 gen_array_ops.py:6771(placeholder_with_default)\n",
    "       37    0.000    0.000    0.021    0.001 gen_array_ops.py:7997(reshape)\n",
    "       20    0.000    0.000    0.011    0.001 gen_array_ops.py:9817(squeeze)\n",
    "       20    0.000    0.000    0.000    0.000 gen_array_ops.py:9873(<listcomp>)\n",
    "    44688    0.086    0.000    1.483    0.000 gen_array_ops.py:9998(strided_slice)\n",
    "       10    0.000    0.000    0.024    0.002 gen_functional_ops.py:272(_if)\n",
    "       10    0.000    0.000    0.000    0.000 gen_functional_ops.py:321(<listcomp>)\n",
    "       10    0.000    0.000    0.000    0.000 gen_functional_ops.py:328(<listcomp>)\n",
    "       10    0.000    0.000    0.007    0.001 gen_functional_ops.py:646(stateless_if)\n",
    "       10    0.000    0.000    0.000    0.000 gen_functional_ops.py:699(<listcomp>)\n",
    "       10    0.000    0.000    0.000    0.000 gen_functional_ops.py:706(<listcomp>)\n",
    "       56    0.000    0.000    0.009    0.000 gen_logging_ops.py:23(_assert)\n",
    "   204785    0.370    0.000    6.447    0.000 gen_math_ops.py:10071(sub)\n",
    "        1    0.000    0.000    0.001    0.001 gen_math_ops.py:1944(cast)\n",
    "       13    0.000    0.000    0.000    0.000 gen_math_ops.py:306(add)\n",
    "   109860    0.208    0.000    4.469    0.000 gen_math_ops.py:3175(equal)\n",
    "   255856    0.551    0.000    7.082    0.000 gen_math_ops.py:3911(greater)\n",
    "   348028    0.436    0.000   13.686    0.000 gen_math_ops.py:451(add_v2)\n",
    "    99646    0.355    0.000    4.657    0.000 gen_math_ops.py:4830(less)\n",
    "       10    0.000    0.000    0.009    0.001 gen_math_ops.py:5330(logical_and)\n",
    "       56    0.000    0.000    0.011    0.000 gen_math_ops.py:5389(logical_not)\n",
    "       56    0.001    0.000    0.088    0.002 gen_math_ops.py:5537(mat_mul)\n",
    "   174027    0.259    0.000    5.247    0.000 gen_math_ops.py:6060(mul)\n",
    "   276694    0.401    0.000    7.796    0.000 gen_math_ops.py:7225(real_div)\n",
    "      118    0.001    0.000    0.059    0.000 gen_nn_ops.py:10328(relu)\n",
    "      128    0.002    0.000    0.361    0.003 gen_nn_ops.py:4200(fused_batch_norm_v3)\n",
    "       56    0.000    0.000    0.031    0.001 gen_nn_ops.py:642(bias_add)\n",
    "      118    0.002    0.000    0.207    0.002 gen_nn_ops.py:856(conv2d)\n",
    "      118    0.000    0.000    0.001    0.000 gen_nn_ops.py:943(<listcomp>)\n",
    "      118    0.000    0.000    0.000    0.000 gen_nn_ops.py:954(<listcomp>)\n",
    "      118    0.000    0.000    0.000    0.000 gen_nn_ops.py:964(<listcomp>)\n",
    "       13    0.000    0.000    0.001    0.000 gen_random_ops.py:687(random_uniform)\n",
    "       56    0.000    0.000    0.002    0.000 gen_resource_variable_ops.py:1147(var_handle_op)\n",
    "       56    0.000    0.000    0.002    0.000 gen_resource_variable_ops.py:121(assign_variable_op)\n",
    "       56    0.000    0.000    0.001    0.000 gen_resource_variable_ops.py:1229(var_is_initialized_op)\n",
    "      742    0.006    0.000    0.665    0.001 gen_resource_variable_ops.py:445(read_variable_op)\n",
    "       66    0.000    0.000    0.001    0.000 generic_utils.py:350(deserialize_keras_object)\n",
    "       20    0.000    0.000    0.000    0.000 generic_utils.py:742(to_list)\n",
    "       38    0.000    0.000    0.001    0.000 generic_utils.py:759(to_snake_case)\n",
    "      483    0.000    0.000    0.001    0.000 generic_utils.py:769(is_all_none)\n",
    "       47    0.000    0.000    0.000    0.000 generic_utils.py:786(validate_kwargs)\n",
    "        1    0.000    0.000    0.000    0.000 genericpath.py:117(_splitext)\n",
    "       42    0.000    0.000    0.001    0.000 genericpath.py:16(exists)\n",
    "      588    0.001    0.000    0.002    0.000 getlimits.py:365(__new__)\n",
    "       37    0.000    0.000    0.000    0.000 getlimits.py:497(__init__)\n",
    "       37    0.000    0.000    0.000    0.000 getlimits.py:521(max)\n",
    "      820    0.013    0.000    0.257    0.000 graph_only_ops.py:29(graph_placeholder)\n",
    "        3    0.000    0.000    0.000    0.000 graph_view.py:143(__init__)\n",
    "       13    0.000    0.000    0.000    0.000 init_ops_v2.py:1021(_assert_float_dtype)\n",
    "       13    0.000    0.000    0.000    0.000 init_ops_v2.py:1044(__init__)\n",
    "       13    0.000    0.000    0.004    0.000 init_ops_v2.py:1061(random_uniform)\n",
    "       23    0.000    0.000    0.003    0.000 init_ops_v2.py:117(__call__)\n",
    "       20    0.000    0.000    0.003    0.000 init_ops_v2.py:157(__call__)\n",
    "       13    0.000    0.000    0.000    0.000 init_ops_v2.py:503(__init__)\n",
    "       13    0.000    0.000    0.004    0.000 init_ops_v2.py:525(__call__)\n",
    "       13    0.000    0.000    0.000    0.000 init_ops_v2.py:749(__init__)\n",
    "       13    0.000    0.000    0.000    0.000 init_ops_v2.py:994(_compute_fans)\n",
    "       66    0.000    0.000    0.003    0.000 initializers.py:169(deserialize)\n",
    "       66    0.001    0.000    0.001    0.000 initializers.py:176(<dictcomp>)\n",
    "      142    0.000    0.000    0.004    0.000 initializers.py:189(get)\n",
    "        3    0.000    0.000    0.007    0.002 input_layer.py:194(Input)\n",
    "        3    0.000    0.000    0.007    0.002 input_layer.py:88(__init__)\n",
    "      518    0.004    0.000    0.013    0.000 input_spec.py:132(assert_input_compatibility)\n",
    "       38    0.000    0.000    0.000    0.000 input_spec.py:57(__init__)\n",
    "       38    0.000    0.000    0.000    0.000 input_spec.py:75(<dictcomp>)\n",
    "      145    0.002    0.000    0.008    0.000 inspect.py:1089(getfullargspec)\n",
    "       35    0.000    0.000    0.000    0.000 inspect.py:1495(currentframe)\n",
    "      482    0.000    0.000    0.000    0.000 inspect.py:159(isfunction)\n",
    "       17    0.000    0.000    0.000    0.000 inspect.py:172(isgeneratorfunction)\n",
    "      145    0.002    0.000    0.005    0.000 inspect.py:2102(_signature_from_function)\n",
    "  230/145    0.001    0.000    0.005    0.000 inspect.py:2183(_signature_from_callable)\n",
    "       66    0.000    0.000    0.000    0.000 inspect.py:229(istraceback)\n",
    "       66    0.000    0.000    0.000    0.000 inspect.py:239(isframe)\n",
    "      485    0.001    0.000    0.002    0.000 inspect.py:2452(__init__)\n",
    "      970    0.000    0.000    0.000    0.000 inspect.py:2502(name)\n",
    "      705    0.000    0.000    0.000    0.000 inspect.py:2506(default)\n",
    "      485    0.000    0.000    0.000    0.000 inspect.py:2510(annotation)\n",
    "      485    0.000    0.000    0.000    0.000 inspect.py:2514(kind)\n",
    "       64    0.000    0.000    0.000    0.000 inspect.py:253(iscode)\n",
    "      145    0.001    0.000    0.001    0.000 inspect.py:2732(__init__)\n",
    "       38    0.000    0.000    0.000    0.000 inspect.py:277(isbuiltin)\n",
    "      630    0.000    0.000    0.000    0.000 inspect.py:2781(<genexpr>)\n",
    "      145    0.000    0.000    0.000    0.000 inspect.py:2817(parameters)\n",
    "      145    0.000    0.000    0.000    0.000 inspect.py:2821(return_annotation)\n",
    "       17    0.000    0.000    0.000    0.000 inspect.py:479(getmro)\n",
    "        2    0.000    0.000    0.000    0.000 inspect.py:485(unwrap)\n",
    "        2    0.000    0.000    0.000    0.000 inspect.py:502(_is_wrapper)\n",
    "      172    0.000    0.000    0.000    0.000 inspect.py:64(ismodule)\n",
    "       60    0.000    0.000    0.001    0.000 inspect.py:643(getfile)\n",
    "       58    0.000    0.000    0.002    0.000 inspect.py:680(getsourcefile)\n",
    "      123    0.000    0.000    0.000    0.000 inspect.py:687(<genexpr>)\n",
    "      164    0.000    0.000    0.000    0.000 inspect.py:690(<genexpr>)\n",
    "       17    0.000    0.000    0.000    0.000 inspect.py:702(getabsfile)\n",
    "       99    0.000    0.000    0.001    0.000 inspect.py:714(getmodule)\n",
    "      307    0.000    0.000    0.000    0.000 inspect.py:73(isclass)\n",
    "        4    0.000    0.000    0.000    0.000 inspect.py:760(findsource)\n",
    "      618    0.000    0.000    0.000    0.000 inspect.py:81(ismethod)\n",
    "        4    0.000    0.000    0.000    0.000 inspect.py:882(__init__)\n",
    "      282    0.000    0.000    0.000    0.000 inspect.py:891(tokeneater)\n",
    "        4    0.000    0.000    0.002    0.000 inspect.py:935(getblock)\n",
    "        2    0.000    0.000    0.001    0.001 inspect.py:946(getsourcelines)\n",
    "        2    0.002    0.001    0.006    0.003 inspect_utils.py:116(_fix_linecache_record)\n",
    "        2    0.000    0.000    0.008    0.004 inspect_utils.py:143(getimmediatesource)\n",
    "        2    0.000    0.000    0.000    0.000 inspect_utils.py:151(getnamespace)\n",
    "       51    0.000    0.000    0.000    0.000 inspect_utils.py:240(_get_unbound_function)\n",
    "       17    0.000    0.000    0.000    0.000 inspect_utils.py:250(getdefiningclass)\n",
    "       17    0.000    0.000    0.000    0.000 inspect_utils.py:265(getmethodclass)\n",
    "        2    0.000    0.000    0.000    0.000 inspect_utils.py:339(getfutureimports)\n",
    "        2    0.000    0.000    0.000    0.000 inspect_utils.py:350(<genexpr>)\n",
    "        2    0.000    0.000    0.000    0.000 inspect_utils.py:60(islambda)\n",
    "       51    0.000    0.000    0.000    0.000 inspect_utils.py:68(isnamedtuple)\n",
    "       38    0.000    0.000    0.002    0.000 inspect_utils.py:82(isbuiltin)\n",
    "     5928    0.001    0.000    0.001    0.000 inspect_utils.py:84(<genexpr>)\n",
    "       38    0.000    0.000    0.000    0.000 inspect_utils.py:96(isconstructor)\n",
    "       36    0.000    0.000    0.003    0.000 iostream.py:195(schedule)\n",
    "       11    0.000    0.000    0.000    0.000 iostream.py:307(_is_master_process)\n",
    "       11    0.000    0.000    0.001    0.000 iostream.py:320(_schedule_flush)\n",
    "        8    0.000    0.000    0.009    0.001 iostream.py:334(flush)\n",
    "       11    0.000    0.000    0.002    0.000 iostream.py:382(write)\n",
    "       36    0.000    0.000    0.000    0.000 iostream.py:93(_event_pipe)\n",
    "  521/483    0.002    0.000    0.006    0.000 layer_utils.py:126(wrapped)\n",
    "       44    0.000    0.000    0.000    0.000 layer_utils.py:161(mark_as)\n",
    "      521    0.000    0.000    0.000    0.000 layer_utils.py:166(in_cached_state)\n",
    "       51    0.000    0.000    0.000    0.000 layer_utils.py:179(__init__)\n",
    "       48    0.000    0.000    1.716    0.036 layer_utils.py:194(add_parent)\n",
    "      521    0.001    0.000    0.001    0.000 layer_utils.py:205(get)\n",
    "    44/41    0.000    0.000    0.000    0.000 layer_utils.py:209(_set)\n",
    "       41    0.000    0.000    0.000    0.000 layer_utils.py:216(mark_cached)\n",
    "        3    0.000    0.000    0.000    0.000 layer_utils.py:220(invalidate)\n",
    "    68/58    0.000    0.000    0.001    0.000 layer_utils.py:224(invalidate_all)\n",
    "1246/1128    0.002    0.000    0.008    0.000 layer_utils.py:234(filter_empty_layer_containers)\n",
    "      798    0.000    0.000    0.001    0.000 layer_utils.py:37(is_layer)\n",
    "      912    0.001    0.000    0.001    0.000 layer_utils.py:43(has_weights)\n",
    "        4    0.000    0.000    0.000    0.000 linecache.py:37(getlines)\n",
    "        4    0.000    0.000    0.000    0.000 linecache.py:53(checkcache)\n",
    "       22    0.000    0.000    0.000    0.000 liveness.py:106(__init__)\n",
    "       22    0.000    0.000    0.029    0.001 liveness.py:112(visit_FunctionDef)\n",
    "       22    0.000    0.000    0.000    0.000 liveness.py:155(__init__)\n",
    "  2407/22    0.002    0.000    0.029    0.001 liveness.py:160(visit)\n",
    "       22    0.000    0.000    0.028    0.001 liveness.py:170(visit_FunctionDef)\n",
    "       16    0.000    0.000    0.000    0.000 liveness.py:178(_block_statement_live_out)\n",
    "       34    0.000    0.000    0.000    0.000 liveness.py:186(_block_statement_live_in)\n",
    "        8    0.000    0.000    0.003    0.000 liveness.py:213(visit_Try)\n",
    "        8    0.000    0.000    0.001    0.000 liveness.py:218(visit_ExceptHandler)\n",
    "       18    0.000    0.000    0.020    0.001 liveness.py:223(visit_With)\n",
    "        8    0.000    0.000    0.000    0.000 liveness.py:227(visit_Expr)\n",
    "       22    0.000    0.000    0.058    0.003 liveness.py:235(resolve)\n",
    "       22    0.000    0.000    0.000    0.000 liveness.py:43(__init__)\n",
    "      346    0.000    0.000    0.000    0.000 liveness.py:49(init_state)\n",
    "      241    0.001    0.000    0.002    0.000 liveness.py:52(visit_node)\n",
    "        2    0.000    0.000    0.002    0.001 loader.py:36(load_source)\n",
    "        2    0.000    0.000    0.020    0.010 loader.py:56(load_ast)\n",
    "     2484    0.004    0.000    0.008    0.000 lock_util.py:106(_another_group_active)\n",
    "     4968    0.002    0.000    0.002    0.000 lock_util.py:108(<genexpr>)\n",
    "     7452    0.003    0.000    0.003    0.000 lock_util.py:110(_validate_group_id)\n",
    "     2484    0.001    0.000    0.001    0.000 lock_util.py:119(__init__)\n",
    "     2484    0.002    0.000    0.018    0.000 lock_util.py:123(__enter__)\n",
    "     2484    0.002    0.000    0.026    0.000 lock_util.py:126(__exit__)\n",
    "       60    0.000    0.000    0.001    0.000 lock_util.py:54(__init__)\n",
    "     2484    0.004    0.000    0.006    0.000 lock_util.py:74(group)\n",
    "     2484    0.007    0.000    0.016    0.000 lock_util.py:86(acquire)\n",
    "     2484    0.008    0.000    0.023    0.000 lock_util.py:96(release)\n",
    "        1    0.000    0.000    0.000    0.000 logger.py:24(info)\n",
    "        2    0.000    0.000    0.003    0.002 logical_expressions.py:135(transform)\n",
    "       41    0.000    0.000    0.000    0.000 loss_scale.py:421(get)\n",
    "       41    0.000    0.000    0.000    0.000 loss_scale.py:46(get)\n",
    "   306094    1.668    0.000   27.875    0.000 math_ops.py:1007(r_binary_op_wrapper)\n",
    "   276694    2.108    0.000   13.183    0.000 math_ops.py:1074(_truediv_python3)\n",
    "   348028    0.706    0.000   15.197    0.000 math_ops.py:1271(_add_dispatch)\n",
    "   174027    0.293    0.000    5.572    0.000 math_ops.py:1279(_mul_dispatch)\n",
    "       10    0.000    0.000    0.009    0.001 math_ops.py:1360(logical_and)\n",
    "   109860    0.435    0.000    5.735    0.000 math_ops.py:1484(tensor_equals)\n",
    "       92    0.000    0.000    0.002    0.000 math_ops.py:732(cast)\n",
    "   697440    3.196    0.000   37.093    0.000 math_ops.py:981(binary_op_wrapper)\n",
    "     1668    0.001    0.000    0.002    0.000 memory.py:24(dismantle_ordered_dict)\n",
    "      392    0.000    0.000    0.000    0.000 module_wrapper.py:154(__getattribute__)\n",
    "        3    0.000    0.000    0.000    0.000 monitoring.py:134(get_cell)\n",
    "        3    0.000    0.000    0.000    0.000 monitoring.py:293(__init__)\n",
    "        3    0.000    0.000    0.000    0.000 monitoring.py:301(set)\n",
    "        3    0.000    0.000    0.000    0.000 monitoring.py:333(get_cell)\n",
    "        2    0.000    0.000    0.000    0.000 multiarray.py:1043(copyto)\n",
    "      297    0.000    0.000    0.000    0.000 multiarray.py:145(concatenate)\n",
    "       10    0.000    0.000    0.000    0.000 naming.py:103(new_symbol)\n",
    "        4    0.000    0.000    0.000    0.000 naming.py:35(__init__)\n",
    "        2    0.000    0.000    0.000    0.000 naming.py:39(_as_symbol_name)\n",
    "        2    0.000    0.000    0.000    0.000 naming.py:90(function_name)\n",
    "    30816    0.015    0.000    0.081    0.000 nest.py:104(_is_namedtuple)\n",
    "      151    0.000    0.000    0.002    0.000 nest.py:1288(yield_flat_paths)\n",
    "    15954    0.082    0.000    0.141    0.000 nest.py:129(_sequence_like)\n",
    "      116    0.000    0.000    0.003    0.000 nest.py:1355(flatten_with_tuple_paths)\n",
    "    76124    0.038    0.000    0.302    0.000 nest.py:191(_yield_value)\n",
    "    76426    0.073    0.000    0.265    0.000 nest.py:196(_yield_sorted_items)\n",
    "   203612    0.109    0.000    0.580    0.000 nest.py:259(flatten)\n",
    "      136    0.000    0.000    0.001    0.000 nest.py:341(assert_same_structure)\n",
    "    15918    0.075    0.000    0.409    0.000 nest.py:444(_packed_nest_with_indices)\n",
    "    17653    0.056    0.000    0.636    0.000 nest.py:479(_pack_sequence_as)\n",
    "    17653    0.016    0.000    0.652    0.000 nest.py:519(pack_sequence_as)\n",
    "     2181    0.008    0.000    0.142    0.000 nest.py:555(map_structure)\n",
    "     2181    0.001    0.000    0.005    0.000 nest.py:613(<listcomp>)\n",
    "     2181    0.002    0.000    0.113    0.000 nest.py:617(<listcomp>)\n",
    "  221/151    0.000    0.000    0.002    0.000 nest.py:702(_yield_flat_up_to)\n",
    "     1288    0.001    0.000    0.002    0.000 nest.py:96(_sorted)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:1356(_validate_graph_inputs_and_outputs)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:1359(<setcomp>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:1394(<listcomp>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:1499(_compute_tensor_usage_count)\n",
    "        6    0.000    0.000    0.000    0.000 network.py:1507(<genexpr>)\n",
    "       35    0.000    0.000    0.000    0.000 network.py:1515(<setcomp>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:1584(_set_save_spec)\n",
    "       38    0.000    0.000    0.000    0.000 network.py:1620(_make_node_key)\n",
    "        3    0.000    0.000    0.002    0.001 network.py:1624(_map_graph_network)\n",
    "     40/5    0.000    0.000    0.001    0.000 network.py:1649(build_map)\n",
    "        3    0.000    0.000    0.007    0.002 network.py:167(__init__)\n",
    "       38    0.000    0.000    0.000    0.000 network.py:1766(<lambda>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:1799(<listcomp>)\n",
    "        3    0.000    0.000    0.001    0.000 network.py:193(_base_init)\n",
    "        3    0.000    0.000    0.006    0.002 network.py:229(_init_graph_network)\n",
    "        8    0.000    0.000    0.000    0.000 network.py:253(<genexpr>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:262(<lambda>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:346(_set_output_names)\n",
    "        3    0.000    0.000    0.001    0.000 network.py:378(dynamic)\n",
    "       41    0.000    0.000    0.000    0.000 network.py:382(<genexpr>)\n",
    "  471/387    0.001    0.000    0.004    0.000 network.py:438(__setattr__)\n",
    "      210    0.000    0.000    0.000    0.000 network.py:444(<genexpr>)\n",
    "        3    0.000    0.000    0.000    0.000 network.py:512(_should_compute_mask)\n",
    "        6    0.000    0.000    0.001    0.000 network.py:527(layers)\n",
    "       36    0.000    0.000    0.000    0.000 network.py:584(input_spec)\n",
    "       36    0.000    0.000    1.206    0.033 network.py:695(call)\n",
    "       36    0.008    0.000    1.206    0.033 network.py:800(_run_internal_graph)\n",
    "       36    0.000    0.000    0.000    0.000 network.py:828(<listcomp>)\n",
    "      824    0.001    0.000    0.001    0.000 network.py:853(<genexpr>)\n",
    "      412    0.001    0.000    0.001    0.000 network.py:858(<lambda>)\n",
    "      108    0.000    0.000    0.000    0.000 network.py:877(_map_tensor_if_from_keras_layer)\n",
    "       36    0.000    0.000    0.000    0.000 network.py:905(<listcomp>)\n",
    "       36    0.000    0.000    0.000    0.000 network.py:918(_flatten_to_reference_inputs)\n",
    "       36    0.000    0.000    0.005    0.000 network.py:931(_conform_to_reference_input)\n",
    "      128    0.001    0.000    0.618    0.005 nn_impl.py:1425(fused_batch_norm)\n",
    "       10    0.000    0.000    0.004    0.000 nn_ops.py:1021(__init__)\n",
    "       10    0.000    0.000    0.001    0.000 nn_ops.py:1084(_build_op)\n",
    "      118    0.000    0.000    0.209    0.002 nn_ops.py:1093(__call__)\n",
    "      118    0.000    0.000    0.000    0.000 nn_ops.py:1531(_convert_padding)\n",
    "       10    0.000    0.000    0.001    0.000 nn_ops.py:161(__init__)\n",
    "      118    0.001    0.000    0.208    0.002 nn_ops.py:1920(conv2d)\n",
    "      118    0.000    0.000    0.209    0.002 nn_ops.py:230(__call__)\n",
    "       56    0.000    0.000    0.000    0.000 nn_ops.py:2725(_tf_deterministic_ops)\n",
    "       56    0.001    0.000    0.088    0.002 nn_ops.py:2738(bias_add)\n",
    "       10    0.000    0.000    0.002    0.000 nn_ops.py:503(__init__)\n",
    "       30    0.000    0.000    0.000    0.000 nn_ops.py:533(<genexpr>)\n",
    "       30    0.000    0.000    0.000    0.000 nn_ops.py:534(<genexpr>)\n",
    "      236    0.000    0.000    0.001    0.000 nn_ops.py:60(_get_sequence)\n",
    "      118    0.000    0.000    0.209    0.002 nn_ops.py:637(__call__)\n",
    "       10    0.000    0.000    0.001    0.000 nn_ops.py:720(_get_strides_and_dilation_rate)\n",
    "       41    0.000    0.000    0.000    0.000 node.py:117(<listcomp>)\n",
    "       76    0.000    0.000    0.001    0.000 node.py:134(iterate_inbound)\n",
    "       38    0.000    0.000    0.000    0.000 node.py:153(<listcomp>)\n",
    "       38    0.000    0.000    0.000    0.000 node.py:168(_get_all_node_dependencies)\n",
    "       41    0.000    0.000    0.003    0.000 node.py:66(__init__)\n",
    "       10    0.000    0.000    0.012    0.001 normalization.py:176(__init__)\n",
    "       10    0.000    0.000    0.000    0.000 normalization.py:249(_raise_if_fused_cannot_be_used)\n",
    "       10    0.000    0.000    0.000    0.000 normalization.py:274(_fused_can_be_used)\n",
    "      236    0.000    0.000    0.000    0.000 normalization.py:281(trainable)\n",
    "       10    0.000    0.000    0.000    0.000 normalization.py:285(trainable)\n",
    "       10    0.000    0.000    0.012    0.001 normalization.py:291(_get_trainable_var)\n",
    "       40    0.000    0.000    0.001    0.000 normalization.py:297(_param_dtype)\n",
    "      118    0.000    0.000    0.001    0.000 normalization.py:305(_support_zero_size_input)\n",
    "       10    0.001    0.000    1.764    0.176 normalization.py:310(build)\n",
    "       10    0.000    0.000    0.000    0.000 normalization.py:372(<dictcomp>)\n",
    "      118    0.001    0.000    0.837    0.007 normalization.py:523(_fused_batch_norm)\n",
    "      128    0.000    0.000    0.000    0.000 normalization.py:550(_maybe_add_or_remove_bessels_correction)\n",
    "       10    0.000    0.000    0.053    0.005 normalization.py:568(_fused_batch_norm_training)\n",
    "      118    0.000    0.000    0.565    0.005 normalization.py:584(_fused_batch_norm_inference)\n",
    "       10    0.000    0.000    0.000    0.000 normalization.py:611(<lambda>)\n",
    "       10    0.000    0.000    0.000    0.000 normalization.py:612(<lambda>)\n",
    "      118    0.000    0.000    0.022    0.000 normalization.py:710(_get_training_value)\n",
    "      118    0.001    0.000    0.860    0.007 normalization.py:724(call)\n",
    "       22    0.000    0.000    0.000    0.000 numeric.py:1786(isscalar)\n",
    "        2    0.000    0.000    0.001    0.000 numeric.py:283(full)\n",
    "      594    0.001    0.000    0.002    0.000 numerictypes.py:293(issubclass_)\n",
    "      297    0.001    0.000    0.003    0.000 numerictypes.py:365(issubdtype)\n",
    "     1312    0.000    0.000    0.000    0.000 object_identity.py:125(__init__)\n",
    "     2406    0.001    0.000    0.002    0.000 object_identity.py:128(_wrap_key)\n",
    "     1531    0.002    0.000    0.004    0.000 object_identity.py:131(__getitem__)\n",
    "      669    0.001    0.000    0.001    0.000 object_identity.py:134(__setitem__)\n",
    "      206    0.000    0.000    0.001    0.000 object_identity.py:137(__delitem__)\n",
    "     6072    0.007    0.000    0.008    0.000 object_identity.py:174(__init__)\n",
    "     6227    0.001    0.000    0.001    0.000 object_identity.py:175(<genexpr>)\n",
    "     3946    0.002    0.000    0.003    0.000 object_identity.py:183(_wrap_key)\n",
    "     1741    0.001    0.000    0.003    0.000 object_identity.py:186(__contains__)\n",
    "      200    0.000    0.000    0.000    0.000 object_identity.py:189(discard)\n",
    "     1940    0.002    0.000    0.008    0.000 object_identity.py:192(add)\n",
    "     7423    0.006    0.000    0.020    0.000 object_identity.py:195(update)\n",
    "     7423    0.003    0.000    0.011    0.000 object_identity.py:196(<listcomp>)\n",
    "      313    0.000    0.000    0.000    0.000 object_identity.py:208(__len__)\n",
    "    11767    0.006    0.000    0.006    0.000 object_identity.py:211(__iter__)\n",
    "      832    0.001    0.000    0.003    0.000 object_identity.py:220(_wrap_key)\n",
    "       58    0.000    0.000    0.001    0.000 object_identity.py:223(__len__)\n",
    "       58    0.000    0.000    0.001    0.000 object_identity.py:225(<listcomp>)\n",
    "     1568    0.001    0.000    0.001    0.000 object_identity.py:227(__iter__)\n",
    "     7184    0.002    0.000    0.002    0.000 object_identity.py:35(__init__)\n",
    "     1750    0.000    0.000    0.000    0.000 object_identity.py:38(unwrapped)\n",
    "      947    0.000    0.000    0.000    0.000 object_identity.py:42(_assert_type)\n",
    "      947    0.000    0.000    0.001    0.000 object_identity.py:54(__eq__)\n",
    "     7184    0.002    0.000    0.003    0.000 object_identity.py:63(__hash__)\n",
    "      832    0.001    0.000    0.002    0.000 object_identity.py:75(__init__)\n",
    "     1452    0.000    0.000    0.000    0.000 object_identity.py:78(unwrapped)\n",
    "     2382    0.004    0.000    0.008    0.000 op_callbacks.py:118(should_invoke_op_callbacks)\n",
    "      256    0.000    0.000    0.001    0.000 op_def_library.py:141(_MakeFloat)\n",
    "      944    0.000    0.000    0.001    0.000 op_def_library.py:148(_MakeInt)\n",
    "      420    0.000    0.000    0.001    0.000 op_def_library.py:159(_MakeStr)\n",
    "      359    0.000    0.000    0.000    0.000 op_def_library.py:166(_MakeBool)\n",
    "     1825    0.005    0.000    0.012    0.000 op_def_library.py:173(_MakeType)\n",
    "       53    0.000    0.000    0.001    0.000 op_def_library.py:184(_MakeShape)\n",
    "       40    0.000    0.000    0.000    0.000 op_def_library.py:216(_MakeFunc)\n",
    "     3124    0.001    0.000    0.001    0.000 op_def_library.py:235(_MaybeColocateWith)\n",
    "1562/1132    0.106    0.000    1.223    0.001 op_def_library.py:299(_apply_op_helper)\n",
    "     1039    0.002    0.000    0.002    0.000 op_def_library.py:38(_Attr)\n",
    "       20    0.000    0.000    0.000    0.000 op_def_library.py:447(<listcomp>)\n",
    "     2192    0.001    0.000    0.001    0.000 op_def_library.py:46(_AttrValue)\n",
    "     2393    0.002    0.000    0.006    0.000 op_def_library.py:510(<listcomp>)\n",
    "     2864    0.006    0.000    0.011    0.000 op_def_library.py:53(_SatisfiesTypeConstraint)\n",
    "     2393    0.002    0.000    0.002    0.000 op_def_library.py:64(_IsListParameter)\n",
    "      374    0.000    0.000    0.001    0.000 op_def_library.py:682(<listcomp>)\n",
    "       40    0.000    0.000    0.000    0.000 op_def_library.py:695(<listcomp>)\n",
    "       20    0.000    0.000    0.001    0.000 op_def_library.py:700(<listcomp>)\n",
    "     1562    0.002    0.000    0.002    0.000 op_def_library.py:737(<listcomp>)\n",
    "     5095    0.002    0.000    0.004    0.000 op_def_library.py:80(_IsListValue)\n",
    "     1562    0.002    0.000    0.009    0.000 op_def_library.py:84(_Flatten)\n",
    "     1562    0.002    0.000    0.006    0.000 op_def_library.py:87(<listcomp>)\n",
    "     1562    0.001    0.000    0.001    0.000 op_def_library.py:89(<listcomp>)\n",
    "     4043    0.002    0.000    0.002    0.000 op_def_registry.py:34(get)\n",
    "     1082    0.000    0.000    0.000    0.000 ops.py:102(tensor_id)\n",
    "     1170    0.002    0.000    0.010    0.000 ops.py:1058(shape)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:1071(get_shape)\n",
    "   110154    0.181    0.000    0.181    0.000 ops.py:1115(graph)\n",
    "   174602    0.205    0.000    4.781    0.000 ops.py:1220(convert_to_tensor_v2)\n",
    "1171640/1170898    4.081    0.000   17.454    0.000 ops.py:1290(convert_to_tensor)\n",
    "       20    0.000    0.000    0.015    0.001 ops.py:1365(internal_convert_n_to_tensor)\n",
    "      151    0.000    0.000    0.010    0.000 ops.py:1444(convert_to_tensor_or_composite)\n",
    "      151    0.000    0.000    0.010    0.000 ops.py:1467(internal_convert_to_tensor_or_composite)\n",
    "  1280595    0.200    0.000    0.200    0.000 ops.py:156(__init__)\n",
    "     2484    0.036    0.000    0.058    0.000 ops.py:1579(_NodeDef)\n",
    "  1280595    0.159    0.000    0.159    0.000 ops.py:159(__enter__)\n",
    "     2484    0.035    0.000    0.201    0.000 ops.py:1606(_create_c_op)\n",
    "  1280595    0.221    0.000    0.221    0.000 ops.py:162(__exit__)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:1636(<listcomp>)\n",
    "     2484    0.050    0.000    0.455    0.000 ops.py:1681(__init__)\n",
    "      102    0.000    0.000    0.000    0.000 ops.py:1764(<listcomp>)\n",
    "     4805    0.003    0.000    0.012    0.000 ops.py:1767(<genexpr>)\n",
    "     2484    0.004    0.000    0.010    0.000 ops.py:1837(_control_flow_post_processing)\n",
    "     7327    0.001    0.000    0.001    0.000 ops.py:1878(_get_control_flow_context)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:1894(name)\n",
    "     2491    0.001    0.000    0.001    0.000 ops.py:2087(_add_control_inputs)\n",
    "      357    0.000    0.000    0.001    0.000 ops.py:209(is_dense_tensor_like)\n",
    "     4987    0.001    0.000    0.001    0.000 ops.py:2149(outputs)\n",
    "     3798    0.007    0.000    0.023    0.000 ops.py:2154(inputs)\n",
    "       80    0.000    0.000    0.000    0.000 ops.py:2198(_control_outputs)\n",
    "       80    0.000    0.000    0.000    0.000 ops.py:2213(<listcomp>)\n",
    "    32643    0.011    0.000    0.032    0.000 ops.py:2218(type)\n",
    "    10406    0.002    0.000    0.002    0.000 ops.py:2223(graph)\n",
    "       58    0.000    0.000    0.002    0.000 ops.py:2264(_set_attr)\n",
    "       58    0.000    0.000    0.001    0.000 ops.py:2273(_set_attr_with_buf)\n",
    "     1759    0.009    0.000    0.137    0.000 ops.py:2313(get_attr)\n",
    "     3318    0.003    0.000    0.007    0.000 ops.py:255(uid)\n",
    "     1695    0.001    0.000    0.001    0.000 ops.py:2643(name_from_scope_name)\n",
    "       60    0.002    0.000    0.005    0.000 ops.py:2704(__init__)\n",
    "       36    0.000    0.000    0.000    0.000 ops.py:2821(_variable_creator_scope)\n",
    "       18    0.000    0.000    0.000    0.000 ops.py:2845(<lambda>)\n",
    "     4312    0.002    0.000    0.003    0.000 ops.py:2857(_variable_creator_stack)\n",
    "     4238    0.001    0.000    0.001    0.000 ops.py:2881(_variable_creator_stack)\n",
    "     5008    0.001    0.000    0.001    0.000 ops.py:2885(_check_not_finalized)\n",
    "     2484    0.010    0.000    0.012    0.000 ops.py:2894(_add_op)\n",
    "     5016    0.002    0.000    0.002    0.000 ops.py:2913(_c_graph)\n",
    "       40    0.000    0.000    0.000    0.000 ops.py:2954(seed)\n",
    "       60    0.000    0.000    0.000    0.000 ops.py:2959(seed)\n",
    "     3366    0.001    0.000    0.001    0.000 ops.py:2990(_get_control_flow_context)\n",
    "     1604    0.000    0.000    0.000    0.000 ops.py:2998(_set_control_flow_context)\n",
    "       40    0.000    0.000    0.000    0.000 ops.py:3139(_is_function)\n",
    "       40    0.000    0.000    0.002    0.000 ops.py:3161(_add_function)\n",
    "    21467    0.006    0.000    0.006    0.000 ops.py:3196(building_function)\n",
    "     2484    0.029    0.000    0.682    0.000 ops.py:3260(_create_op_internal)\n",
    "     4907    0.002    0.000    0.002    0.000 ops.py:3314(<genexpr>)\n",
    "     2484    0.022    0.000    0.070    0.000 ops.py:3363(_create_op_helper)\n",
    "      828    0.007    0.000    0.007    0.000 ops.py:3603(get_operations)\n",
    "     2583    0.003    0.000    0.003    0.000 ops.py:3642(_get_operation_by_name_unsafe)\n",
    "     2423    0.002    0.000    0.007    0.000 ops.py:3665(_get_operation_by_tf_operation)\n",
    "     2423    0.005    0.000    0.012    0.000 ops.py:3690(_get_tensor_by_tf_output)\n",
    "      922    0.002    0.000    0.005    0.000 ops.py:3710(_get_op_def)\n",
    "     2119    0.002    0.000    0.008    0.000 ops.py:3728(as_default)\n",
    "      100    0.000    0.000    0.000    0.000 ops.py:3814(get_collection_ref)\n",
    "       58    0.000    0.000    0.000    0.000 ops.py:3839(get_collection)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:3879(get_all_collection_keys)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:3882(<listcomp>)\n",
    "     3144    0.011    0.000    0.032    0.000 ops.py:389(__init__)\n",
    "     9937    0.004    0.000    0.006    0.000 ops.py:3921(_name_stack)\n",
    "     4244    0.002    0.000    0.002    0.000 ops.py:3928(_name_stack)\n",
    "     4244    0.012    0.000    0.033    0.000 ops.py:3933(name_scope)\n",
    "     2967    0.008    0.000    0.014    0.000 ops.py:4048(unique_name)\n",
    "       99    0.000    0.000    0.000    0.000 ops.py:4100(get_name_scope)\n",
    "     1484    0.002    0.000    0.012    0.000 ops.py:4117(_colocate_with_for_gradient)\n",
    "     1484    0.004    0.000    0.008    0.000 ops.py:4130(colocate_with)\n",
    "     3144    0.004    0.000    0.036    0.000 ops.py:416(_create_with_tf_output)\n",
    "     8725    0.002    0.000    0.002    0.000 ops.py:422(op)\n",
    "    10230    0.002    0.000    0.002    0.000 ops.py:427(dtype)\n",
    "     2484    0.007    0.000    0.020    0.000 ops.py:4281(_apply_device_functions)\n",
    "     4599    0.003    0.000    0.004    0.000 ops.py:432(graph)\n",
    "      822    0.001    0.000    0.001    0.000 ops.py:4364(__init__)\n",
    "      822    0.003    0.000    0.006    0.000 ops.py:4399(__enter__)\n",
    "      822    0.002    0.000    0.005    0.000 ops.py:4409(__exit__)\n",
    "      822    0.000    0.000    0.000    0.000 ops.py:4417(control_inputs)\n",
    "      822    0.001    0.000    0.001    0.000 ops.py:4421(add_op)\n",
    "       70    0.000    0.000    0.000    0.000 ops.py:4426(op_in_group)\n",
    "      822    0.001    0.000    0.001    0.000 ops.py:4431(_push_control_dependencies_controller)\n",
    "      822    0.001    0.000    0.003    0.000 ops.py:4434(_pop_control_dependencies_controller)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:4438(_current_control_dependencies)\n",
    "     2484    0.003    0.000    0.007    0.000 ops.py:4445(_control_dependencies_for_inputs)\n",
    "      822    0.000    0.000    0.000    0.000 ops.py:4476(<genexpr>)\n",
    "     2484    0.002    0.000    0.007    0.000 ops.py:4479(_record_op_seen_by_control_dependencies)\n",
    "      822    0.001    0.000    0.002    0.000 ops.py:4488(control_dependencies)\n",
    "     2178    0.001    0.000    0.013    0.000 ops.py:451(shape)\n",
    "      112    0.001    0.000    0.001    0.000 ops.py:4602(_attr_scope)\n",
    "        2    0.000    0.000    0.000    0.000 ops.py:4804(prevent_feeding)\n",
    "       20    0.000    0.000    0.000    0.000 ops.py:4812(prevent_fetching)\n",
    "       60    0.000    0.000    0.000    0.000 ops.py:4823(switch_to_thread_local)\n",
    "    13607    0.007    0.000    0.010    0.000 ops.py:4840(_device_function_stack)\n",
    "     2484    0.002    0.000    0.005    0.000 ops.py:4861(_snapshot_device_function_stack_metadata)\n",
    "     3603    0.002    0.000    0.002    0.000 ops.py:4877(_device_function_stack)\n",
    "     5750    0.003    0.000    0.004    0.000 ops.py:4886(_colocation_stack)\n",
    "     2484    0.004    0.000    0.007    0.000 ops.py:4901(_snapshot_colocation_stack_metadata)\n",
    "     2484    0.001    0.000    0.001    0.000 ops.py:4903(<dictcomp>)\n",
    "     1524    0.001    0.000    0.001    0.000 ops.py:4908(_colocation_stack)\n",
    "     8256    0.005    0.000    0.009    0.000 ops.py:4917(_control_dependencies_stack)\n",
    "     1604    0.001    0.000    0.001    0.000 ops.py:4929(_control_dependencies_stack)\n",
    "    21981    0.021    0.000    0.033    0.000 ops.py:4936(_distribution_strategy_stack)\n",
    "     4238    0.002    0.000    0.002    0.000 ops.py:4943(_distribution_strategy_stack)\n",
    "     4721    0.002    0.000    0.003    0.000 ops.py:4959(_auto_cast_variable_read_dtype)\n",
    "     5204    0.004    0.000    0.007    0.000 ops.py:4974(_auto_cast_variable_read_dtype)\n",
    "      966    0.001    0.000    0.003    0.000 ops.py:4980(_enable_auto_casting_variables)\n",
    "     2484    0.003    0.000    0.009    0.000 ops.py:5002(_mutation_lock)\n",
    "      561    0.002    0.000    0.012    0.000 ops.py:502(_c_api_shape)\n",
    "        1    0.000    0.000    0.000    0.000 ops.py:5021(device)\n",
    "      551    0.000    0.000    0.000    0.000 ops.py:510(<listcomp>)\n",
    "      742    0.002    0.000    0.007    0.000 ops.py:5110(_colocate_with_for_gradient)\n",
    "      742    0.000    0.000    0.007    0.000 ops.py:5133(colocate_with)\n",
    "      921    0.002    0.000    0.007    0.000 ops.py:5144(control_dependencies)\n",
    "    65499    0.040    0.000    0.050    0.000 ops.py:5184(get_default)\n",
    "     4238    0.004    0.000    0.006    0.000 ops.py:5201(get_controller)\n",
    "    65499    0.096    0.000    0.161    0.000 ops.py:5371(get_default)\n",
    "    45139    0.015    0.000    0.015    0.000 ops.py:5378(_GetGlobalDefaultGraph)\n",
    "     4238    0.018    0.000    0.058    0.000 ops.py:5391(get_controller)\n",
    "     1596    0.008    0.000    0.015    0.000 ops.py:5412(_get_outer_context_and_inner_device_stack)\n",
    "      640    0.002    0.000    0.010    0.000 ops.py:5450(init_scope)\n",
    "   111401    0.128    0.000    0.479    0.000 ops.py:5555(executing_eagerly_outside_functions)\n",
    "      953    0.001    0.000    0.003    0.000 ops.py:5588(inside_function)\n",
    "    65499    0.053    0.000    0.214    0.000 ops.py:5800(get_default_graph)\n",
    "        1    0.000    0.000    0.000    0.000 ops.py:5824(get_name_scope)\n",
    "     1562    0.002    0.000    0.006    0.000 ops.py:5859(_get_graph_from_inputs)\n",
    "      122    0.000    0.000    0.000    0.000 ops.py:593(get_shape)\n",
    "       75    0.001    0.000    0.001    0.000 ops.py:597(set_shape)\n",
    "      544    0.306    0.001    0.539    0.001 ops.py:6075(dismantle_graph)\n",
    "       58    0.000    0.000    0.000    0.000 ops.py:6155(get_collection)\n",
    "  1326824    2.710    0.000    4.137    0.000 ops.py:6189(name_scope)\n",
    "    44689    0.045    0.000    0.045    0.000 ops.py:6227(<genexpr>)\n",
    "     1639    0.002    0.000    0.002    0.000 ops.py:6243(__init__)\n",
    "     1639    0.003    0.000    0.034    0.000 ops.py:6262(__enter__)\n",
    "     1639    0.002    0.000    0.007    0.000 ops.py:6302(__exit__)\n",
    "    44844    0.095    0.000    0.155    0.000 ops.py:6360(enter_eager_name_scope)\n",
    "    45327    0.060    0.000    0.067    0.000 ops.py:6405(__init__)\n",
    "    45327    0.102    0.000    0.294    0.000 ops.py:6423(__enter__)\n",
    "    44844    0.027    0.000    0.083    0.000 ops.py:6437(<lambda>)\n",
    "    45327    0.062    0.000    0.173    0.000 ops.py:6444(__exit__)\n",
    "      742    0.000    0.000    0.000    0.000 ops.py:6579(_op_to_colocate_with)\n",
    "      294    0.000    0.000    0.001    0.000 ops.py:6604(_is_keras_symbolic_tensor)\n",
    "      220    0.000    0.000    0.001    0.000 ops.py:665(consumers)\n",
    "     2484    0.010    0.000    0.011    0.000 ops.py:6690(_reconstruct_sequence_inputs)\n",
    "       20    0.000    0.000    0.001    0.000 ops.py:6747(set_int_list_attr)\n",
    "      220    0.000    0.000    0.000    0.000 ops.py:675(<listcomp>)\n",
    "     5788    0.002    0.000    0.002    0.000 ops.py:699(_as_tf_output)\n",
    "   465362    1.969    0.000   58.337    0.000 ops.py:883(__bool__)\n",
    "   480366    0.521    0.000   60.015    0.000 ops.py:924(_numpy)\n",
    "  1998757    1.248    0.000    1.677    0.000 ops.py:931(dtype)\n",
    "    15004    0.053    0.000    3.760    0.000 ops.py:937(numpy)\n",
    "        8    0.000    0.000    0.000    0.000 origin_info.py:162(__init__)\n",
    "        4    0.000    0.000    0.000    0.000 origin_info.py:169(__init__)\n",
    "      307    0.000    0.000    0.000    0.000 origin_info.py:191(_absolute_lineno)\n",
    "      307    0.000    0.000    0.000    0.000 origin_info.py:194(_absolute_col_offset)\n",
    "      307    0.001    0.000    0.002    0.000 origin_info.py:197(_attach_origin_info)\n",
    "    522/4    0.001    0.000    0.005    0.001 origin_info.py:212(visit)\n",
    "        4    0.000    0.000    0.009    0.002 origin_info.py:226(resolve)\n",
    "        2    0.000    0.000    0.003    0.002 origin_info.py:263(resolve_entity)\n",
    "      118    0.000    0.000    0.000    0.000 origin_info.py:57(line_loc)\n",
    "        2    0.001    0.000    0.015    0.008 origin_info.py:89(create_source_map)\n",
    "      271    0.001    0.000    0.001    0.000 os.py:664(__getitem__)\n",
    "      271    0.000    0.000    0.000    0.000 os.py:742(encode)\n",
    "      270    0.000    0.000    0.002    0.000 os.py:760(getenv)\n",
    "        2    0.000    0.000    0.002    0.001 parser.py:119(_attempt_to_parse_normal_source)\n",
    "        2    0.000    0.000    0.010    0.005 parser.py:182(parse_entity)\n",
    "        2    0.000    0.000    0.000    0.000 parser.py:209(<genexpr>)\n",
    "       35    0.000    0.000    0.012    0.000 parser.py:220(parse)\n",
    "       13    0.000    0.000    0.001    0.000 parser.py:244(parse_expression)\n",
    "        2    0.000    0.000    0.003    0.002 parser.py:263(unparse)\n",
    "        2    0.000    0.000    0.000    0.000 parser.py:46(_unfold_continuations)\n",
    "        2    0.000    0.000    0.001    0.001 parser.py:51(dedent_block)\n",
    "       41    0.000    0.000    0.001    0.000 policy.py:302(__init__)\n",
    "       41    0.000    0.000    0.000    0.000 policy.py:351(_parse_name)\n",
    "      214    0.000    0.000    0.000    0.000 policy.py:391(variable_dtype)\n",
    "     1073    0.000    0.000    0.000    0.000 policy.py:406(compute_dtype)\n",
    "       16    0.000    0.000    0.000    0.000 policy.py:434(should_cast_variables)\n",
    "       41    0.000    0.000    0.000    0.000 policy.py:454(name)\n",
    "       38    0.000    0.000    0.001    0.000 policy.py:489(global_policy)\n",
    "       38    0.000    0.000    0.000    0.000 policy.py:512(policy_defaults_to_floatx)\n",
    "        1    0.000    0.000    0.000    0.000 posixpath.py:121(splitext)\n",
    "        3    0.000    0.000    0.000    0.000 posixpath.py:144(basename)\n",
    "        1    0.000    0.000    0.000    0.000 posixpath.py:154(dirname)\n",
    "        3    0.000    0.000    0.000    0.000 posixpath.py:338(normpath)\n",
    "        3    0.000    0.000    0.000    0.000 posixpath.py:376(abspath)\n",
    "       11    0.000    0.000    0.000    0.000 posixpath.py:41(_get_sep)\n",
    "        3    0.000    0.000    0.000    0.000 posixpath.py:64(isabs)\n",
    "        4    0.000    0.000    0.000    0.000 posixpath.py:75(join)\n",
    "        1    0.000    0.000    0.000    0.000 process.py:146(name)\n",
    "        1    0.000    0.000    0.000    0.000 process.py:35(current_process)\n",
    "        2    0.000    0.000    0.000    0.000 profiler.py:87(stop)\n",
    "     2056    0.000    0.000    0.000    0.000 qual_names.py:112(has_subscript)\n",
    "     2056    0.000    0.000    0.000    0.000 qual_names.py:115(has_attr)\n",
    "8615/8445    0.003    0.000    0.004    0.000 qual_names.py:163(__hash__)\n",
    " 1016/986    0.001    0.000    0.002    0.000 qual_names.py:166(__eq__)\n",
    "    24/12    0.000    0.000    0.000    0.000 qual_names.py:171(__str__)\n",
    "      773    0.001    0.000    0.010    0.000 qual_names.py:225(visit_Name)\n",
    "      246    0.001    0.000    0.007    0.000 qual_names.py:230(visit_Attribute)\n",
    "       48    0.000    0.000    0.023    0.000 qual_names.py:260(resolve)\n",
    "     1041    0.001    0.000    0.001    0.000 qual_names.py:68(__init__)\n",
    "       16    0.000    0.000    0.000    0.000 random.py:223(_randbelow)\n",
    "       16    0.000    0.000    0.000    0.000 random.py:255(choice)\n",
    "        3    0.000    0.000    0.000    0.000 random.py:680(getrandbits)\n",
    "       13    0.000    0.000    0.004    0.000 random_ops.py:204(random_uniform)\n",
    "       13    0.000    0.000    0.000    0.000 random_seed.py:41(get_seed)\n",
    "       26    0.000    0.000    0.000    0.000 re.py:169(match)\n",
    "       92    0.000    0.000    0.001    0.000 re.py:184(sub)\n",
    "      985    0.000    0.000    0.001    0.000 re.py:231(compile)\n",
    "     1103    0.000    0.000    0.000    0.000 re.py:286(_compile)\n",
    "       76    0.000    0.000    0.000    0.000 re.py:324(_subx)\n",
    "       20    0.000    0.000    0.000    0.000 re.py:330(filter)\n",
    "      173    0.000    0.000    0.002    0.000 reaching_definitions.py:101(__sub__)\n",
    "       22    0.000    0.000    0.001    0.000 reaching_definitions.py:115(__init__)\n",
    "      346    0.000    0.000    0.000    0.000 reaching_definitions.py:124(init_state)\n",
    "      173    0.002    0.000    0.012    0.000 reaching_definitions.py:127(visit_node)\n",
    "       22    0.000    0.000    0.000    0.000 reaching_definitions.py:206(__init__)\n",
    "       22    0.000    0.000    0.037    0.002 reaching_definitions.py:213(visit_FunctionDef)\n",
    "      524    0.001    0.000    0.004    0.000 reaching_definitions.py:245(visit_Name)\n",
    "       16    0.000    0.000    0.000    0.000 reaching_definitions.py:267(_aggregate_predecessors_defined_in)\n",
    "        8    0.000    0.000    0.002    0.000 reaching_definitions.py:298(visit_Try)\n",
    "        8    0.000    0.000    0.001    0.000 reaching_definitions.py:302(visit_ExceptHandler)\n",
    "  1819/22    0.002    0.000    0.038    0.002 reaching_definitions.py:308(visit)\n",
    "       22    0.000    0.000    0.038    0.002 reaching_definitions.py:320(resolve)\n",
    "      201    0.000    0.000    0.000    0.000 reaching_definitions.py:52(__init__)\n",
    "     1189    0.001    0.000    0.003    0.000 reaching_definitions.py:69(__init__)\n",
    "      497    0.001    0.000    0.001    0.000 reaching_definitions.py:72(<dictcomp>)\n",
    "      135    0.000    0.000    0.001    0.000 reaching_definitions.py:76(<dictcomp>)\n",
    "      173    0.001    0.000    0.001    0.000 reaching_definitions.py:82(__eq__)\n",
    "      173    0.000    0.000    0.001    0.000 reaching_definitions.py:88(__ne__)\n",
    "      324    0.002    0.000    0.004    0.000 reaching_definitions.py:91(__or__)\n",
    "        1    0.000    0.000    0.000    0.000 registration.py:102(spec)\n",
    "        1    0.000    0.000    0.191    0.191 registration.py:141(make)\n",
    "        1    0.000    0.000    0.000    0.000 registration.py:15(load)\n",
    "        1    0.000    0.000    0.191    0.191 registration.py:49(make)\n",
    "        1    0.000    0.000    0.191    0.191 registration.py:81(make)\n",
    "     2481    0.001    0.000    0.002    0.000 registry.py:74(list)\n",
    "     4962    0.004    0.000    0.009    0.000 registry.py:82(lookup)\n",
    "      145    0.000    0.000    0.000    0.000 regularizers.py:297(get)\n",
    "      742    0.001    0.000    0.725    0.001 resource_variable_ops.py:1233(_dense_var_to_tensor)\n",
    "       56    0.000    0.000    0.054    0.001 resource_variable_ops.py:1335(__init__)\n",
    "       56    0.003    0.000    0.054    0.001 resource_variable_ops.py:1436(_init_from_args)\n",
    "       56    0.003    0.000    0.027    0.000 resource_variable_ops.py:146(_variable_handle_from_shape_and_dtype)\n",
    "      742    0.001    0.000    0.726    0.001 resource_variable_ops.py:1824(_dense_var_to_tensor)\n",
    "       56    0.000    0.000    0.028    0.000 resource_variable_ops.py:198(eager_safe_variable_handle)\n",
    "       56    0.000    0.000    0.000    0.000 resource_variable_ops.py:268(__init__)\n",
    "      742    0.001    0.000    0.003    0.000 resource_variable_ops.py:314(_maybe_set_handle_data)\n",
    "      742    0.002    0.000    0.021    0.000 resource_variable_ops.py:326(variable_accessed)\n",
    "       56    0.001    0.000    0.003    0.000 resource_variable_ops.py:337(__init__)\n",
    "      486    0.000    0.000    0.000    0.000 resource_variable_ops.py:493(dtype)\n",
    "       10    0.000    0.000    0.000    0.000 resource_variable_ops.py:513(shape)\n",
    "      742    0.003    0.000    0.722    0.001 resource_variable_ops.py:545(value)\n",
    "      798    0.000    0.000    0.000    0.000 resource_variable_ops.py:583(trainable)\n",
    "      742    0.003    0.000    0.698    0.001 resource_variable_ops.py:631(_read_variable_op)\n",
    "      742    0.003    0.000    0.671    0.001 resource_variable_ops.py:634(read_and_set_handle)\n",
    "       56    0.000    0.000    0.000    0.000 resource_variable_ops.py:84(_set_handle_shapes_and_types)\n",
    "        2    0.000    0.000    0.002    0.001 return_statements.py:111(visit_With)\n",
    "        2    0.000    0.000    0.002    0.001 return_statements.py:151(visit_FunctionDef)\n",
    "        8    0.000    0.000    0.000    0.000 return_statements.py:159(__init__)\n",
    "        4    0.000    0.000    0.000    0.000 return_statements.py:172(__init__)\n",
    "        2    0.000    0.000    0.000    0.000 return_statements.py:221(__init__)\n",
    "        2    0.000    0.000    0.002    0.001 return_statements.py:225(visit_Return)\n",
    "        9    0.000    0.000    0.000    0.000 return_statements.py:252(_postprocess_statement)\n",
    "      4/2    0.000    0.000    0.004    0.002 return_statements.py:275(_visit_statement_block)\n",
    "        2    0.000    0.000    0.004    0.002 return_statements.py:317(visit_With)\n",
    "        2    0.000    0.000    0.011    0.005 return_statements.py:339(visit_FunctionDef)\n",
    "        6    0.000    0.000    0.000    0.000 return_statements.py:37(__init__)\n",
    "        2    0.000    0.000    0.013    0.006 return_statements.py:394(transform)\n",
    "        2    0.000    0.000    0.000    0.000 return_statements.py:68(visit_Return)\n",
    "        9    0.000    0.000    0.000    0.000 return_statements.py:72(_postprocess_statement)\n",
    "      4/2    0.000    0.000    0.002    0.001 return_statements.py:91(_visit_statement_block)\n",
    "        3    0.000    0.000    0.001    0.000 seeding.py:11(np_random)\n",
    "        4    0.000    0.000    0.000    0.000 seeding.py:21(hash_seed)\n",
    "        3    0.000    0.000    0.000    0.000 seeding.py:45(create_seed)\n",
    "        7    0.000    0.000    0.000    0.000 seeding.py:69(_bigint_from_bytes)\n",
    "        3    0.000    0.000    0.000    0.000 seeding.py:80(_int_list_from_bigint)\n",
    "      294    0.000    0.000    0.001    0.000 shape_base.py:209(_arrays_for_stack_dispatcher)\n",
    "      294    0.001    0.000    0.001    0.000 shape_base.py:348(_stack_dispatcher)\n",
    "      294    0.003    0.000    0.024    0.000 shape_base.py:357(stack)\n",
    "      294    0.001    0.000    0.002    0.000 shape_base.py:420(<listcomp>)\n",
    "      294    0.001    0.000    0.001    0.000 shape_base.py:424(<setcomp>)\n",
    "      294    0.001    0.000    0.001    0.000 shape_base.py:432(<listcomp>)\n",
    "     2522    0.003    0.000    0.004    0.000 six.py:586(iteritems)\n",
    "      128    0.000    0.000    0.825    0.006 smart_cond.py:27(smart_cond)\n",
    "      246    0.000    0.000    0.002    0.000 smart_cond.py:62(smart_constant_value)\n",
    "       36    0.003    0.000    0.003    0.000 socket.py:342(send)\n",
    "        2    0.000    0.000    0.001    0.000 space.py:21(seed)\n",
    "        2    0.000    0.000    0.001    0.000 space.py:9(__init__)\n",
    "       35    0.000    0.000    0.000    0.000 special_values.py:90(retval)\n",
    "       20    0.000    0.000    0.000    0.000 sre_parse.py:963(expand_template)\n",
    "        3    0.000    0.000    0.000    0.000 std.py:1051(__repr__)\n",
    "        2    0.000    0.000    0.000    0.000 std.py:1054(_comparable)\n",
    "        2    0.000    0.000    0.000    0.000 std.py:1058(__hash__)\n",
    "        2    0.000    0.000    0.004    0.002 std.py:1061(__iter__)\n",
    "        1    0.000    0.000    0.001    0.001 std.py:1221(close)\n",
    "        2    0.000    0.000    0.000    0.000 std.py:1238(fp_write)\n",
    "        2    0.000    0.000    0.006    0.003 std.py:1275(refresh)\n",
    "        4    0.000    0.000    0.004    0.001 std.py:1383(moveto)\n",
    "        3    0.000    0.000    0.000    0.000 std.py:1388(format_dict)\n",
    "        3    0.000    0.000    0.007    0.002 std.py:1403(display)\n",
    "        3    0.000    0.000    0.000    0.000 std.py:146(__init__)\n",
    "        3    0.000    0.000    0.000    0.000 std.py:153(__format__)\n",
    "        5    0.000    0.000    0.000    0.000 std.py:226(format_interval)\n",
    "        1    0.000    0.000    0.000    0.000 std.py:267(ema)\n",
    "        1    0.000    0.000    0.000    0.000 std.py:286(status_printer)\n",
    "        3    0.000    0.000    0.002    0.001 std.py:296(fp_write)\n",
    "        3    0.000    0.000    0.002    0.001 std.py:302(print_status)\n",
    "        3    0.000    0.000    0.000    0.000 std.py:309(format_meter)\n",
    "        1    0.000    0.000    0.000    0.000 std.py:504(__new__)\n",
    "        1    0.000    0.000    0.000    0.000 std.py:526(_get_free_pos)\n",
    "        4    0.000    0.000    0.000    0.000 std.py:529(<genexpr>)\n",
    "        1    0.000    0.000    0.000    0.000 std.py:533(_decr_instances)\n",
    "        1    0.000    0.000    0.000    0.000 std.py:606(get_lock)\n",
    "        1    0.000    0.000    0.003    0.003 std.py:772(__init__)\n",
    "        6    0.000    0.000    0.000    0.000 std.py:88(acquire)\n",
    "        6    0.000    0.000    0.000    0.000 std.py:92(release)\n",
    "        4    0.000    0.000    0.000    0.000 std.py:96(__enter__)\n",
    "        4    0.000    0.000    0.000    0.000 std.py:99(__exit__)\n",
    "        1    1.256    1.256    3.151    3.151 summary_ops_v2.py:1062(run_metadata_graphs)\n",
    "        1    0.000    0.000    0.004    0.004 summary_ops_v2.py:1157(trace_on)\n",
    "        1    0.470    0.470    5.869    5.869 summary_ops_v2.py:1202(trace_export)\n",
    "        2    0.000    0.000    0.000    0.000 summary_ops_v2.py:1258(trace_off)\n",
    "        2    0.000    0.000    0.000    0.000 summary_ops_v2.py:573(summary_scope)\n",
    "        1    0.000    0.000    0.000    0.000 summary_ops_v2.py:614(write)\n",
    "      640    0.001    0.000    0.001    0.000 tape.py:117(stop_recording)\n",
    "     1544    0.001    0.000    0.003    0.000 tape.py:144(record_operation)\n",
    "    14994    0.011    0.000    0.028    0.000 tape.py:187(could_possibly_record)\n",
    "      486    0.002    0.000    0.007    0.000 tape.py:73(variable_accessed)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:236(_infer_return_type)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:257(_sanitize_params)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:285(rng)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:296(__next__)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:299(<listcomp>)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:376(_get_candidate_names)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:390(_mkstemp_inner)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:430(gettempdir)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:565(__init__)\n",
    "        4    0.000    0.000    0.000    0.000 tempfile.py:580(close)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:590(__del__)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:608(__init__)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:614(__getattr__)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:622(func_wrapper)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:635(__enter__)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:641(__exit__)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:646(close)\n",
    "        2    0.000    0.000    0.000    0.000 tempfile.py:663(NamedTemporaryFile)\n",
    "       28    0.000    0.000    0.000    0.000 templates.py:111(__init__)\n",
    "       82    0.000    0.000    0.017    0.000 templates.py:129(_prepare_replacement)\n",
    "       20    0.000    0.000    0.016    0.001 templates.py:146(visit_Expr)\n",
    "        8    0.000    0.000    0.000    0.000 templates.py:154(visit_keyword)\n",
    "      4/2    0.000    0.000    0.007    0.003 templates.py:172(visit_FunctionDef)\n",
    "       22    0.000    0.000    0.001    0.000 templates.py:185(visit_Attribute)\n",
    "      108    0.000    0.000    0.019    0.000 templates.py:197(visit_Name)\n",
    "    81/70    0.000    0.000    0.000    0.000 templates.py:218(_convert_to_ast)\n",
    "        8    0.000    0.000    0.000    0.000 templates.py:228(<listcomp>)\n",
    "        4    0.000    0.000    0.000    0.000 templates.py:230(<genexpr>)\n",
    "       18    0.000    0.000    0.034    0.002 templates.py:234(replace)\n",
    "       18    0.000    0.000    0.008    0.000 templates.py:275(<listcomp>)\n",
    "        6    0.000    0.000    0.005    0.001 templates.py:279(replace_as_expression)\n",
    "       78    0.000    0.000    0.000    0.000 templates.py:43(__init__)\n",
    "   140/52    0.000    0.000    0.001    0.000 templates.py:46(visit)\n",
    "       70    0.000    0.000    0.000    0.000 templates.py:54(_apply_override)\n",
    "        6    0.000    0.000    0.000    0.000 templates.py:58(visit_Attribute)\n",
    "       10    0.000    0.000    0.000    0.000 templates.py:64(visit_Tuple)\n",
    "       54    0.000    0.000    0.001    0.000 templates.py:72(visit_Name)\n",
    "   615910    0.472    0.000    0.669    0.000 tensor_conversion_registry.py:114(get)\n",
    "   480780    0.481    0.000    6.743    0.000 tensor_conversion_registry.py:50(_default_conversion_function)\n",
    "       10    0.000    0.000    0.000    0.000 tensor_shape.py:1019(with_rank_at_least)\n",
    "       56    0.000    0.000    0.000    0.000 tensor_shape.py:1057(is_compatible_with)\n",
    "       40    0.000    0.000    0.001    0.000 tensor_shape.py:1119(most_specific_compatible_shape)\n",
    "      144    0.000    0.000    0.001    0.000 tensor_shape.py:1149(is_fully_defined)\n",
    "      448    0.000    0.000    0.000    0.000 tensor_shape.py:1152(<genexpr>)\n",
    "      533    0.000    0.000    0.001    0.000 tensor_shape.py:1163(as_list)\n",
    "      533    0.001    0.000    0.001    0.000 tensor_shape.py:1174(<listcomp>)\n",
    "     1031    0.005    0.000    0.007    0.000 tensor_shape.py:1176(as_proto)\n",
    "     1021    0.001    0.000    0.001    0.000 tensor_shape.py:1182(<listcomp>)\n",
    "        2    0.000    0.000    0.000    0.000 tensor_shape.py:1186(__eq__)\n",
    "      635    0.000    0.000    0.001    0.000 tensor_shape.py:1213(as_shape)\n",
    "       64    0.000    0.000    0.000    0.000 tensor_shape.py:1221(unknown_shape)\n",
    "       10    0.000    0.000    0.000    0.000 tensor_shape.py:130(dimension_at_index)\n",
    "     3893    0.005    0.000    0.006    0.000 tensor_shape.py:185(__init__)\n",
    "       62    0.000    0.000    0.000    0.000 tensor_shape.py:211(__eq__)\n",
    "     6497    0.001    0.000    0.001    0.000 tensor_shape.py:243(value)\n",
    "      419    0.000    0.000    0.001    0.000 tensor_shape.py:248(is_compatible_with)\n",
    "      320    0.000    0.000    0.001    0.000 tensor_shape.py:264(assert_is_compatible_with)\n",
    "      320    0.000    0.000    0.002    0.000 tensor_shape.py:278(merge_with)\n",
    "     4728    0.003    0.000    0.010    0.000 tensor_shape.py:700(as_dimension)\n",
    "     1655    0.004    0.000    0.017    0.000 tensor_shape.py:742(__init__)\n",
    "     1528    0.002    0.000    0.011    0.000 tensor_shape.py:771(<listcomp>)\n",
    "     2256    0.001    0.000    0.001    0.000 tensor_shape.py:773(_v2_behavior)\n",
    "     1714    0.001    0.000    0.001    0.000 tensor_shape.py:802(rank)\n",
    "      503    0.000    0.000    0.000    0.000 tensor_shape.py:809(dims)\n",
    "      820    0.000    0.000    0.001    0.000 tensor_shape.py:821(ndims)\n",
    "      295    0.000    0.000    0.000    0.000 tensor_shape.py:826(__len__)\n",
    "       56    0.000    0.000    0.000    0.000 tensor_shape.py:832(__bool__)\n",
    "      111    0.000    0.000    0.000    0.000 tensor_shape.py:839(__iter__)\n",
    "      444    0.000    0.000    0.000    0.000 tensor_shape.py:845(<genexpr>)\n",
    "     2219    0.003    0.000    0.004    0.000 tensor_shape.py:849(__getitem__)\n",
    "       90    0.001    0.000    0.004    0.000 tensor_shape.py:909(merge_with)\n",
    "       89    0.000    0.000    0.000    0.000 tensor_shape.py:97(dimension_value)\n",
    "       90    0.000    0.000    0.000    0.000 tensor_shape.py:971(assert_same_rank)\n",
    "       52    0.000    0.000    0.002    0.000 tensor_shape.py:999(with_rank)\n",
    "       21    0.000    0.000    0.000    0.000 tensor_spec.py:39(__init__)\n",
    "       21    0.000    0.000    0.000    0.000 tensor_spec.py:63(shape)\n",
    "       15    0.000    0.000    0.001    0.000 tensor_util.py:1003(shape_tensor)\n",
    "       93    0.000    0.000    0.002    0.000 tensor_util.py:1022(maybe_set_static_shape)\n",
    "       48    0.000    0.000    0.000    0.000 tensor_util.py:185(GetFromNumpyDTypeDict)\n",
    "       48    0.000    0.000    0.000    0.000 tensor_util.py:193(GetNumpyAppendFn)\n",
    "  173/102    0.000    0.000    0.000    0.000 tensor_util.py:218(_GetDenseDimensions)\n",
    "       72    0.000    0.000    0.001    0.000 tensor_util.py:262(inner)\n",
    "       72    0.000    0.000    0.001    0.000 tensor_util.py:263(<listcomp>)\n",
    "       30    0.000    0.000    0.000    0.000 tensor_util.py:275(_check_not_tensor)\n",
    "       30    0.000    0.000    0.000    0.000 tensor_util.py:276(<listcomp>)\n",
    "      102    0.000    0.000    0.001    0.000 tensor_util.py:304(_AssertCompatible)\n",
    "      102    0.000    0.000    0.001    0.000 tensor_util.py:334(_is_array_like)\n",
    "      102    0.003    0.000    0.010    0.000 tensor_util.py:355(make_tensor_proto)\n",
    "        2    0.000    0.000    0.000    0.000 tensor_util.py:564(MakeNdarray)\n",
    "        2    0.000    0.000    0.000    0.000 tensor_util.py:591(<listcomp>)\n",
    "       32    0.000    0.000    0.001    0.000 tensor_util.py:669(_ConstantValue)\n",
    "       42    0.000    0.000    0.001    0.000 tensor_util.py:797(constant_value)\n",
    "        2    0.000    0.000    0.001    0.000 tensor_util.py:835(constant_value_as_shape)\n",
    "        2    0.000    0.000    0.000    0.000 tensor_util.py:974(<listcomp>)\n",
    "      586    0.001    0.000    0.003    0.000 tensor_util.py:978(is_tensor)\n",
    "       18    0.000    0.000    0.000    0.000 textwrap.py:414(dedent)\n",
    "      243    0.000    0.000    0.000    0.000 tf2.py:43(enabled)\n",
    "     2119    0.004    0.000    0.042    0.000 tf_contextlib.py:25(contextmanager)\n",
    "     1100    0.001    0.000    0.001    0.000 tf_decorator.py:114(_has_tf_decorator_attr)\n",
    "       18    0.000    0.000    0.000    0.000 tf_decorator.py:128(rewrap)\n",
    "      720    0.001    0.000    0.003    0.000 tf_decorator.py:200(unwrap)\n",
    "     2149    0.005    0.000    0.006    0.000 tf_decorator.py:236(__init__)\n",
    "      742    0.000    0.000    0.000    0.000 tf_decorator.py:262(decorated_target)\n",
    "       18    0.000    0.000    0.000    0.000 tf_decorator.py:266(decorated_target)\n",
    "       12    0.000    0.000    0.000    0.000 tf_decorator.py:278(decorator_argspec)\n",
    "     2149    0.007    0.000    0.016    0.000 tf_decorator.py:67(make_decorator)\n",
    "      145    0.000    0.000    0.008    0.000 tf_inspect.py:238(getfullargspec)\n",
    "       60    0.001    0.000    0.005    0.000 tf_inspect.py:260(getcallargs)\n",
    "       60    0.000    0.000    0.000    0.000 tf_inspect.py:283(<listcomp>)\n",
    "       89    0.000    0.000    0.001    0.000 tf_inspect.py:335(getmodule)\n",
    "        2    0.000    0.000    0.000    0.000 tf_inspect.py:350(getsourcefile)\n",
    "        2    0.000    0.000    0.001    0.001 tf_inspect.py:355(getsourcelines)\n",
    "      153    0.000    0.000    0.001    0.000 tf_inspect.py:365(isclass)\n",
    "      111    0.000    0.000    0.000    0.000 tf_inspect.py:370(isfunction)\n",
    "       17    0.000    0.000    0.000    0.000 tf_inspect.py:385(isgeneratorfunction)\n",
    "      226    0.000    0.000    0.001    0.000 tf_inspect.py:390(ismethod)\n",
    "        1    0.000    0.000    0.004    0.004 tf_logging.py:171(warn)\n",
    "       36    0.000    0.000    0.000    0.000 tf_logging.py:210(vlog)\n",
    "        1    0.000    0.000    0.000    0.000 tf_logging.py:45(_get_caller)\n",
    "        1    0.000    0.000    0.000    0.000 tf_logging.py:75(_logger_find_caller)\n",
    "       37    0.000    0.000    0.000    0.000 tf_logging.py:93(get_logger)\n",
    "       35    0.000    0.000    0.002    0.000 tf_stack.py:105(__init__)\n",
    "     1835    0.001    0.000    0.001    0.000 tf_stack.py:123(get_filtered_filenames)\n",
    "     2514    0.004    0.000    0.079    0.000 tf_stack.py:131(extract_stack)\n",
    "       70    0.000    0.000    0.000    0.000 tf_stack.py:53(__enter__)\n",
    "       70    0.000    0.000    0.000    0.000 tf_stack.py:71(__exit__)\n",
    "       35    0.000    0.000    0.000    0.000 tf_stack.py:83(reset)\n",
    "       35    0.000    0.000    0.000    0.000 tf_stack.py:95(reset)\n",
    "       20    0.001    0.000    0.003    0.000 tf_utils.py:102(get_reachable_from_inputs)\n",
    "    72/36    0.000    0.000    0.001    0.000 tf_utils.py:157(map_structure_with_atomic)\n",
    "       36    0.000    0.000    0.000    0.000 tf_utils.py:185(<listcomp>)\n",
    "       36    0.000    0.000    0.001    0.000 tf_utils.py:193(convert_shapes)\n",
    "      108    0.000    0.000    0.000    0.000 tf_utils.py:220(_is_shape_component)\n",
    "       72    0.000    0.000    0.000    0.000 tf_utils.py:223(_is_atomic_shape)\n",
    "       72    0.000    0.000    0.000    0.000 tf_utils.py:230(<genexpr>)\n",
    "       36    0.000    0.000    0.000    0.000 tf_utils.py:234(_convert_shape)\n",
    "      966    0.001    0.000    0.011    0.000 tf_utils.py:325(are_all_symbolic_tensors)\n",
    "     1967    0.001    0.000    0.010    0.000 tf_utils.py:326(<genexpr>)\n",
    "     1001    0.003    0.000    0.009    0.000 tf_utils.py:332(is_symbolic_tensor)\n",
    "        3    0.000    0.000    0.000    0.000 tf_utils.py:411(assert_no_legacy_layers)\n",
    "      128    0.000    0.000    0.826    0.006 tf_utils.py:42(smart_cond)\n",
    "        3    0.000    0.000    0.000    0.000 tf_utils.py:425(<listcomp>)\n",
    "       46    0.000    0.000    0.002    0.000 tf_utils.py:436(maybe_init_scope)\n",
    "        3    0.000    0.000    0.000    0.000 tf_utils.py:475(get_tensor_spec)\n",
    "      118    0.000    0.000    0.001    0.000 tf_utils.py:68(constant_value)\n",
    "       44    0.000    0.000    0.000    0.000 threading.py:1062(_wait_for_tstate_lock)\n",
    "        1    0.000    0.000    0.000    0.000 threading.py:1076(name)\n",
    "       44    0.000    0.000    0.000    0.000 threading.py:1104(is_alive)\n",
    "        1    0.000    0.000    0.000    0.000 threading.py:1230(current_thread)\n",
    "       68    0.001    0.000    0.001    0.000 threading.py:215(__init__)\n",
    "        8    0.000    0.000    0.000    0.000 threading.py:239(__enter__)\n",
    "        8    0.000    0.000    0.000    0.000 threading.py:242(__exit__)\n",
    "        6    0.000    0.000    0.000    0.000 threading.py:248(_release_save)\n",
    "        6    0.000    0.000    0.000    0.000 threading.py:251(_acquire_restore)\n",
    "     2490    0.002    0.000    0.003    0.000 threading.py:254(_is_owned)\n",
    "        6    0.000    0.000    0.007    0.001 threading.py:263(wait)\n",
    "     2484    0.006    0.000    0.009    0.000 threading.py:334(notify)\n",
    "     2484    0.004    0.000    0.013    0.000 threading.py:357(notify_all)\n",
    "        8    0.000    0.000    0.000    0.000 threading.py:498(__init__)\n",
    "       45    0.000    0.000    0.000    0.000 threading.py:506(is_set)\n",
    "        8    0.000    0.000    0.007    0.001 threading.py:533(wait)\n",
    "       60    0.000    0.000    0.000    0.000 threading.py:74(RLock)\n",
    "      294    0.002    0.000    0.216    0.001 time_limit.py:14(step)\n",
    "        1    0.000    0.000    0.010    0.010 time_limit.py:23(reset)\n",
    "        1    0.000    0.000    0.000    0.000 time_limit.py:5(__init__)\n",
    "        1    0.000    0.000    0.000    0.000 tmp555ck6zk.py:2(<module>)\n",
    "        1    0.000    0.000    0.000    0.000 tmp555ck6zk.py:2(create_converted_entity_factory)\n",
    "        1    0.000    0.000    0.000    0.000 tmp555ck6zk.py:4(create_converted_entity)\n",
    "        1    0.000    0.000    0.067    0.067 tmp555ck6zk.py:6(tf__initial_inference_compiled)\n",
    "       17    0.000    0.000    0.000    0.000 tmpqiwkscxv.py:2(create_converted_entity_factory)\n",
    "       17    0.000    0.000    0.000    0.000 tmpqiwkscxv.py:4(create_converted_entity)\n",
    "       17    0.000    0.000    0.000    0.000 tmpqiwkscxv.py:6(tf____hash__)\n",
    "        1    0.000    0.000    0.000    0.000 tmpwx_3xg_e.py:2(<module>)\n",
    "       17    0.000    0.000    0.000    0.000 tmpwx_3xg_e.py:2(create_converted_entity_factory)\n",
    "       17    0.000    0.000    0.000    0.000 tmpwx_3xg_e.py:4(create_converted_entity)\n",
    "       17    0.001    0.000    1.224    0.072 tmpwx_3xg_e.py:6(tf__recurrent_inference_compiled)\n",
    "      981    0.000    0.000    0.001    0.000 tokenize.py:152(_compile)\n",
    "        2    0.000    0.000    0.000    0.000 tokenize.py:224(__init__)\n",
    "        2    0.000    0.000    0.000    0.000 tokenize.py:243(untokenize)\n",
    "        2    0.000    0.000    0.000    0.000 tokenize.py:280(compat)\n",
    "        2    0.000    0.000    0.000    0.000 tokenize.py:317(untokenize)\n",
    "     1059    0.003    0.000    0.006    0.000 tokenize.py:492(_tokenize)\n",
    "       10    0.000    0.000    0.000    0.000 tokenize.py:739(generate_tokens)\n",
    "     2481    0.003    0.000    0.021    0.000 tpu.py:210(tpu_replicated_input_resolver)\n",
    "     4962    0.007    0.000    0.016    0.000 tpu.py:226(replace_with_unreplicated_resources)\n",
    "     4547    0.007    0.000    0.007    0.000 traceable_stack.py:119(peek_objs)\n",
    "     4547    0.001    0.000    0.001    0.000 traceable_stack.py:121(<genexpr>)\n",
    "     4968    0.002    0.000    0.002    0.000 traceable_stack.py:123(peek_traceable_objs)\n",
    "     7808    0.003    0.000    0.004    0.000 traceable_stack.py:127(__len__)\n",
    "      116    0.000    0.000    0.000    0.000 traceable_stack.py:131(copy)\n",
    "     1720    0.001    0.000    0.001    0.000 traceable_stack.py:83(__init__)\n",
    "2210/2090    0.002    0.000    0.008    0.000 tracking.py:417(wrapped)\n",
    "      164    0.000    0.000    0.000    0.000 tracking.py:77(__setattr__)\n",
    "        3    0.000    0.000    0.010    0.003 training.py:166(__init__)\n",
    "        3    0.000    0.000    0.001    0.000 training.py:352(_reset_compile_cache)\n",
    "        3    0.000    0.000    0.000    0.000 training_utils.py:1901(get_input_shape_and_dtype)\n",
    "        3    0.000    0.000    0.000    0.000 training_utils.py:1915(_is_graph_model)\n",
    "        3    0.000    0.000    0.000    0.000 training_utils.py:1935(get_static_batch_size)\n",
    "       64    0.000    0.000    0.000    0.000 transformer.py:100(__init__)\n",
    "       24    0.000    0.000    0.000    0.000 transformer.py:108(__enter__)\n",
    "       24    0.000    0.000    0.000    0.000 transformer.py:112(__exit__)\n",
    "       84    0.000    0.000    0.000    0.000 transformer.py:115(enter)\n",
    "       44    0.000    0.000    0.000    0.000 transformer.py:118(exit)\n",
    "        2    0.000    0.000    0.000    0.000 transformer.py:121(stack)\n",
    "      322    0.000    0.000    0.000    0.000 transformer.py:125(level)\n",
    "      710    0.000    0.000    0.000    0.000 transformer.py:133(__iter__)\n",
    "      108    0.000    0.000    0.000    0.000 transformer.py:136(__getattr__)\n",
    "       51    0.000    0.000    0.000    0.000 transformer.py:139(__setattr__)\n",
    "      110    0.000    0.000    0.000    0.000 transformer.py:175(__init__)\n",
    "     1216    0.000    0.000    0.001    0.000 transformer.py:178(__getitem__)\n",
    "      110    0.000    0.000    0.000    0.000 transformer.py:236(__init__)\n",
    "   372/76    0.001    0.000    0.074    0.001 transformer.py:273(visit_block)\n",
    "11164/110    0.025    0.000    0.185    0.002 transformer.py:408(visit)\n",
    "        2    0.000    0.000    0.000    0.000 transformer.py:43(__init__)\n",
    "       21    0.000    0.000    0.001    0.000 unparser.py:115(_Assign)\n",
    "     15/9    0.000    0.000    0.000    0.000 unparser.py:14(interleave)\n",
    "        6    0.000    0.000    0.000    0.000 unparser.py:141(_Return)\n",
    "        2    0.000    0.000    0.000    0.000 unparser.py:223(_Raise)\n",
    "        2    0.000    0.000    0.000    0.000 unparser.py:245(_Try)\n",
    "        2    0.000    0.000    0.000    0.000 unparser.py:292(_ExceptHandler)\n",
    "        2    0.000    0.000    0.001    0.001 unparser.py:32(__init__)\n",
    "      6/2    0.000    0.000    0.001    0.000 unparser.py:346(_FunctionDef)\n",
    "      6/2    0.000    0.000    0.001    0.000 unparser.py:352(__FunctionDef_helper)\n",
    "       43    0.000    0.000    0.000    0.000 unparser.py:42(fill)\n",
    "        2    0.000    0.000    0.001    0.000 unparser.py:422(_generic_With)\n",
    "        2    0.000    0.000    0.001    0.000 unparser.py:435(_With)\n",
    "        4    0.000    0.000    0.000    0.000 unparser.py:445(_Str)\n",
    "      388    0.000    0.000    0.000    0.000 unparser.py:46(write)\n",
    "       12    0.000    0.000    0.000    0.000 unparser.py:50(enter)\n",
    "       83    0.000    0.000    0.000    0.000 unparser.py:520(_Name)\n",
    "       17    0.000    0.000    0.000    0.000 unparser.py:523(_NameConstant)\n",
    "       12    0.000    0.000    0.000    0.000 unparser.py:55(leave)\n",
    "        8    0.000    0.000    0.000    0.000 unparser.py:555(_Num)\n",
    "      5/1    0.000    0.000    0.000    0.000 unparser.py:570(_List)\n",
    "        7    0.000    0.000    0.000    0.000 unparser.py:572(<lambda>)\n",
    "    263/2    0.000    0.000    0.001    0.000 unparser.py:59(dispatch)\n",
    "       15    0.000    0.000    0.000    0.000 unparser.py:651(_Tuple)\n",
    "        8    0.000    0.000    0.000    0.000 unparser.py:658(<lambda>)\n",
    "       28    0.000    0.000    0.000    0.000 unparser.py:706(_Attribute)\n",
    "    19/15    0.000    0.000    0.000    0.000 unparser.py:716(_Call)\n",
    "       11    0.000    0.000    0.000    0.000 unparser.py:772(_arg)\n",
    "        6    0.000    0.000    0.000    0.000 unparser.py:779(_arguments)\n",
    "       10    0.000    0.000    0.000    0.000 unparser.py:837(_keyword)\n",
    "        2    0.000    0.000    0.000    0.000 unparser.py:859(_withitem)\n",
    "        2    0.000    0.000    0.000    0.000 unparser.py:88(_Expr)\n",
    "        8    0.000    0.000    0.000    0.000 unsupported_features_checker.py:32(visit_Attribute)\n",
    "        2    0.000    0.000    0.001    0.000 unsupported_features_checker.py:60(verify)\n",
    "        3    0.000    0.000    0.000    0.000 util.py:1070(__init__)\n",
    "        3    0.000    0.000    0.000    0.000 util.py:1338(saver_with_op_caching)\n",
    "        3    0.000    0.000    0.000    0.000 utils.py:134(__init__)\n",
    "        3    0.000    0.000    0.000    0.000 utils.py:138(__format__)\n",
    "        1    0.000    0.000    0.000    0.000 utils.py:151(__eq__)\n",
    "        1    0.000    0.000    0.000    0.000 utils.py:188(_is_utf)\n",
    "        1    0.000    0.000    0.000    0.000 utils.py:202(_supports_unicode)\n",
    "        6    0.000    0.000    0.000    0.000 utils.py:209(_is_ascii)\n",
    "        1    0.000    0.000    0.000    0.000 utils.py:218(_environ_cols_wrapper)\n",
    "        1    0.000    0.000    0.000    0.000 utils.py:270(_environ_cols_linux)\n",
    "        4    0.000    0.000    0.000    0.000 utils.py:288(_term_move_up)\n",
    "       58    0.000    0.000    0.000    0.000 variable_scope.py:1167(use_resource)\n",
    "      116    0.000    0.000    0.000    0.000 variable_scope.py:1203(set_use_resource)\n",
    "       58    0.000    0.000    0.000    0.000 variable_scope.py:1421(get_variable_scope_store)\n",
    "       58    0.000    0.000    0.001    0.000 variable_scope.py:1434(get_variable_scope)\n",
    "       56    0.000    0.000    0.055    0.001 variable_scope.py:2558(default_variable_creator)\n",
    "      112    0.000    0.000    0.000    0.000 variables.py:150(validate_synchronization_aggregation_trainable)\n",
    "       56    0.000    0.000    0.055    0.001 variables.py:181(_variable_v1_call)\n",
    "       56    0.000    0.000    0.055    0.001 variables.py:198(<lambda>)\n",
    "   112/56    0.000    0.000    0.056    0.001 variables.py:257(__call__)\n",
    "        3    0.000    0.000    0.000    0.000 version_utils.py:46(__new__)\n",
    "       41    0.000    0.000    0.019    0.000 version_utils.py:55(__new__)\n",
    "   114/44    0.018    0.000    0.018    0.000 version_utils.py:61(swap_class)\n",
    "   137/85    0.000    0.000    0.014    0.000 version_utils.py:73(<genexpr>)\n",
    "      426    0.001    0.000    0.002    0.000 weakref.py:102(__init__)\n",
    "       11    0.000    0.000    0.000    0.000 weakref.py:109(remove)\n",
    "      173    0.000    0.000    0.000    0.000 weakref.py:125(_commit_removals)\n",
    "       55    0.000    0.000    0.000    0.000 weakref.py:165(__setitem__)\n",
    "      228    0.000    0.000    0.001    0.000 weakref.py:208(items)\n",
    "      426    0.000    0.000    0.000    0.000 weakref.py:288(update)\n",
    "       55    0.000    0.000    0.000    0.000 weakref.py:334(__new__)\n",
    "       55    0.000    0.000    0.000    0.000 weakref.py:339(__init__)\n",
    "       47    0.000    0.000    0.000    0.000 weakref.py:354(__init__)\n",
    "    46/40    0.001    0.000    0.053    0.001 weakref.py:356(remove)\n",
    "       38    0.000    0.000    0.000    0.000 weakref.py:371(_commit_removals)\n",
    "    15639    0.013    0.000    0.013    0.000 weakref.py:393(__getitem__)\n",
    "      291    0.000    0.000    0.000    0.000 weakref.py:406(__setitem__)\n",
    "     2329    0.002    0.000    0.003    0.000 weakref.py:428(get)\n",
    "    15709    0.051    0.000    0.051    0.000 weakref.py:431(__contains__)\n",
    "       76    0.000    0.000    0.000    0.000 weakref.py:438(items)\n",
    "      123    0.000    0.000    0.000    0.000 weakref.py:480(pop)\n",
    "       38    0.000    0.000    0.000    0.000 weakref.py:487(update)\n",
    "    36163    0.026    0.000    0.026    0.000 {built-in method __new__ of type object at 0x9d12c0}\n",
    "        4    0.000    0.000    0.000    0.000 {built-in method _hashlib.openssl_sha512}\n",
    "        4    0.000    0.000    0.000    0.000 {built-in method _imp.acquire_lock}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method _imp.extension_suffixes}\n",
    "        8    0.000    0.000    0.000    0.000 {built-in method _imp.lock_held}\n",
    "        4    0.000    0.000    0.000    0.000 {built-in method _imp.release_lock}\n",
    "        7    0.000    0.000    0.000    0.000 {built-in method _struct.unpack}\n",
    "      125    0.000    0.000    0.000    0.000 {built-in method _thread.allocate_lock}\n",
    "     2590    0.001    0.000    0.001    0.000 {built-in method _thread.get_ident}\n",
    "        5    0.000    0.000    0.000    0.000 {built-in method _weakref._remove_dead_weakref}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method atexit.register}\n",
    "       11    0.000    0.000    0.000    0.000 {built-in method builtins.abs}\n",
    "   139469    0.087    0.000    0.197    0.000 {built-in method builtins.all}\n",
    "     5753    0.005    0.000    0.015    0.000 {built-in method builtins.any}\n",
    "     3009    0.000    0.000    0.000    0.000 {built-in method builtins.callable}\n",
    "       37    0.002    0.000    0.002    0.000 {built-in method builtins.compile}\n",
    "       66    0.001    0.000    0.001    0.000 {built-in method builtins.dir}\n",
    "       19    0.000    0.000    0.000    0.000 {built-in method builtins.divmod}\n",
    "      3/1    0.000    0.000  190.543  190.543 {built-in method builtins.exec}\n",
    "   262157    0.248    0.000    0.432    0.000 {built-in method builtins.getattr}\n",
    "   155572    0.074    0.000    0.076    0.000 {built-in method builtins.hasattr}\n",
    "79498/79328    0.039    0.000    0.045    0.000 {built-in method builtins.hash}\n",
    "    12533    0.002    0.000    0.002    0.000 {built-in method builtins.id}\n",
    "  6914761    1.723    0.000    2.721    0.000 {built-in method builtins.isinstance}\n",
    "   616835    0.172    0.000    0.172    0.000 {built-in method builtins.issubclass}\n",
    "     4952    0.001    0.000    0.001    0.000 {built-in method builtins.iter}\n",
    "213609/213471    0.057    0.000    0.057    0.000 {built-in method builtins.len}\n",
    "   131443    2.668    0.000  119.248    0.001 {built-in method builtins.max}\n",
    "    71630    0.625    0.000   12.027    0.000 {built-in method builtins.min}\n",
    "80698/65569    0.052    0.000    0.292    0.000 {built-in method builtins.next}\n",
    "       63    0.000    0.000    0.000    0.000 {built-in method builtins.ord}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method builtins.print}\n",
    "       29    0.000    0.000    0.000    0.000 {built-in method builtins.repr}\n",
    "    74904    0.039    0.000    0.063    0.000 {built-in method builtins.setattr}\n",
    "    16448    0.030    0.000    0.031    0.000 {built-in method builtins.sorted}\n",
    "    15288    0.019    0.000    0.020    0.000 {built-in method builtins.sum}\n",
    "        1    0.000    0.000    0.000    0.000 {built-in method fcntl.ioctl}\n",
    "        3    0.000    0.000    0.000    0.000 {built-in method from_bytes}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method io.open}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method marshal.dumps}\n",
    "    59976    0.037    0.000    0.037    0.000 {built-in method math.exp}\n",
    "   227712    0.208    0.000    0.208    0.000 {built-in method math.log}\n",
    "   227725    0.078    0.000    0.078    0.000 {built-in method math.sqrt}\n",
    "      735    0.004    0.000    0.004    0.000 {built-in method numpy.array}\n",
    "1642/1348    0.020    0.000    0.050    0.000 {built-in method numpy.core._multiarray_umath.implement_array_function}\n",
    "      294    0.000    0.000    0.000    0.000 {built-in method numpy.core._multiarray_umath.normalize_axis_index}\n",
    "      297    0.001    0.000    0.001    0.000 {built-in method numpy.empty}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method numpy.frombuffer}\n",
    "       12    0.000    0.000    0.000    0.000 {built-in method numpy.geterrobj}\n",
    "        6    0.000    0.000    0.000    0.000 {built-in method numpy.seterrobj}\n",
    "        1    0.000    0.000    0.000    0.000 {built-in method numpy.zeros}\n",
    "       24    0.000    0.000    0.000    0.000 {built-in method posix.fspath}\n",
    "       14    0.000    0.000    0.000    0.000 {built-in method posix.getpid}\n",
    "        4    0.000    0.000    0.000    0.000 {built-in method posix.open}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method posix.replace}\n",
    "       52    0.001    0.000    0.001    0.000 {built-in method posix.stat}\n",
    "        6    0.000    0.000    0.000    0.000 {built-in method posix.urandom}\n",
    "       36    0.000    0.000    0.000    0.000 {built-in method sys._getframe}\n",
    "     1719    0.001    0.000    0.001    0.000 {built-in method sys.exc_info}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method sys.getrecursionlimit}\n",
    "       10    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tf_session.AddControlInput}\n",
    "     2481    0.004    0.000    0.004    0.000 {built-in method tensorflow.python._pywrap_tf_session.GetOperationInputs}\n",
    "       58    0.001    0.000    0.001    0.000 {built-in method tensorflow.python._pywrap_tf_session.SetAttr}\n",
    "      782    0.007    0.000    0.007    0.000 {built-in method tensorflow.python._pywrap_tf_session.SetHandleShapeAndType}\n",
    "       60    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tf_session.SetRequireShapeInferenceFns}\n",
    "       20    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_AddInputList}\n",
    "     2373    0.009    0.000    0.009    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_AddInput}\n",
    "     1952    0.003    0.000    0.003    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_DeleteBuffer}\n",
    "      538    0.088    0.000    0.088    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_DeleteFunction}\n",
    "      544    0.228    0.000    0.228    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_DeleteGraph}\n",
    "     2484    0.056    0.000    0.056    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_FinishOperation}\n",
    "       58    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_FunctionToFunctionDef}\n",
    "      175    0.038    0.000    0.038    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GetBuffer}\n",
    "       40    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GraphCopyFunction}\n",
    "       76    0.001    0.000    0.001    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GraphGetOpDef}\n",
    "      561    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GraphGetTensorShapeHelper}\n",
    "      782    0.003    0.000    0.003    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GraphSetOutputHandleShapesAndTypes_wrapper}\n",
    "       75    0.001    0.000    0.001    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GraphSetTensorShape_wrapper}\n",
    "       58    0.011    0.000    0.011    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_GraphToFunction_wrapper}\n",
    "       58    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_NewBufferFromString}\n",
    "     1894    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_NewBuffer}\n",
    "       60    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_NewGraph}\n",
    "     2484    0.025    0.000    0.025    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_NewOperation}\n",
    "     1759    0.104    0.000    0.109    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationGetAttrValueProto}\n",
    "       80    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationGetControlOutputs_wrapper}\n",
    "     2443    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationName}\n",
    "     2484    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationNumOutputs}\n",
    "    32643    0.021    0.000    0.021    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationOpType}\n",
    "      220    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationOutputConsumers_wrapper}\n",
    "     3144    0.005    0.000    0.005    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_OperationOutputType}\n",
    "     5101    0.021    0.000    0.021    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_SetAttrValueProto}\n",
    "       30    0.001    0.000    0.001    0.000 {built-in method tensorflow.python._pywrap_tf_session.TF_TryEvaluateConstant_wrapper}\n",
    "       58    0.005    0.000    0.005    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_ContextAddFunction}\n",
    "        2    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_ContextDisableRunMetadata}\n",
    "        1    1.149    1.149    1.149    1.149 {built-in method tensorflow.python._pywrap_tfe.TFE_ContextExportRunMetadata}\n",
    "      538    0.541    0.001    0.541    0.001 {built-in method tensorflow.python._pywrap_tfe.TFE_ContextRemoveFunction}\n",
    "        3    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_MonitoringBoolGaugeCellSet}\n",
    "        3    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_MonitoringGetCellBoolGauge1}\n",
    "    14994    0.125    0.000    0.125    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_EncodeArg}\n",
    "    15288   12.214    0.001   12.214    0.001 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_Execute}\n",
    "  1514521   48.281    0.000   48.281    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_FastPathExecute}\n",
    "    16850    0.020    0.000    0.020    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeSetIsEmpty}\n",
    "      320    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeSetIsStopped}\n",
    "    14994    0.015    0.000    0.015    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeSetPossibleGradientTypes}\n",
    "     1544    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeSetRecordOperation}\n",
    "      152    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeSetRestartOnThread}\n",
    "      152    0.000    0.000    0.000    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeSetStopOnThread}\n",
    "      486    0.001    0.000    0.001    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_TapeVariableAccessed}\n",
    "     3318    0.004    0.000    0.004    0.000 {built-in method tensorflow.python._pywrap_tfe.TFE_Py_UID}\n",
    "      136    0.001    0.000    0.001    0.000 {built-in method tensorflow.python._pywrap_utils.AssertSameStructure}\n",
    "   203612    0.470    0.000    0.471    0.000 {built-in method tensorflow.python._pywrap_utils.Flatten}\n",
    "    30816    0.013    0.000    0.013    0.000 {built-in method tensorflow.python._pywrap_utils.IsAttrs}\n",
    "    30816    0.010    0.000    0.010    0.000 {built-in method tensorflow.python._pywrap_utils.IsCompositeTensor}\n",
    "    15368    0.005    0.000    0.005    0.000 {built-in method tensorflow.python._pywrap_utils.IsMappingView}\n",
    "    15404    0.005    0.000    0.005    0.000 {built-in method tensorflow.python._pywrap_utils.IsMapping}\n",
    "    15954    0.009    0.000    0.009    0.000 {built-in method tensorflow.python._pywrap_utils.IsMutableMapping}\n",
    "    30816    0.066    0.000    0.066    0.000 {built-in method tensorflow.python._pywrap_utils.IsNamedtuple}\n",
    "    90561    0.049    0.000    0.049    0.000 {built-in method tensorflow.python._pywrap_utils.IsSequenceOrComposite}\n",
    "     5181    0.002    0.000    0.002    0.000 {built-in method tensorflow.python._pywrap_utils.IsSequence}\n",
    "    30816    0.009    0.000    0.009    0.000 {built-in method tensorflow.python._pywrap_utils.IsTypeSpec}\n",
    "     2514    0.072    0.000    0.074    0.000 {built-in method tensorflow.python._tf_stack.extract_stack}\n",
    "        6    0.000    0.000    0.000    0.000 {built-in method time.time}\n",
    "     4968    0.017    0.000    0.017    0.000 {method 'ByteSize' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "     5256    0.009    0.000    0.009    0.000 {method 'CopyFrom' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "     8276    0.006    0.000    0.006    0.000 {method 'HasField' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "      175    1.070    0.006    1.070    0.006 {method 'ParseFromString' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "     5942    1.839    0.000    1.839    0.000 {method 'SerializeToString' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "      434    0.000    0.000    0.000    0.000 {method 'SetInParent' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "       40    0.000    0.000    0.000    0.000 {method 'WhichOneof' of 'google.protobuf.pyext._message.CMessage' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method '__enter__' of '_io._IOBase' objects}\n",
    "        8    0.000    0.000    0.000    0.000 {method '__enter__' of '_thread.lock' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method '__exit__' of '_io._IOBase' objects}\n",
    "        8    0.000    0.000    0.000    0.000 {method '__exit__' of '_thread.lock' objects}\n",
    "  1999051    0.430    0.000    0.430    0.000 {method '_datatype_enum' of 'tensorflow.python.framework.ops.EagerTensor' objects}\n",
    "   480366   59.494    0.000   59.494    0.000 {method '_numpy_internal' of 'tensorflow.python.framework.ops.EagerTensor' objects}\n",
    "      474    0.001    0.000    0.001    0.000 {method '_shape_tuple' of 'tensorflow.python.framework.ops.EagerTensor' objects}\n",
    "        6    0.000    0.000    0.000    0.000 {method 'acquire' of '_multiprocessing.SemLock' objects}\n",
    "        8    0.000    0.000    0.000    0.000 {method 'acquire' of '_thread.RLock' objects}\n",
    "    22514    0.022    0.000    0.022    0.000 {method 'acquire' of '_thread.lock' objects}\n",
    "     7818    0.003    0.000    0.005    0.000 {method 'add' of 'set' objects}\n",
    "       42    0.000    0.000    0.000    0.000 {method 'append' of 'collections.deque' objects}\n",
    "       56    0.000    0.000    0.000    0.000 {method 'append' of 'google.protobuf.pyext._message.RepeatedCompositeContainer' objects}\n",
    "   398801    0.068    0.000    0.068    0.000 {method 'append' of 'list' objects}\n",
    "       22    0.000    0.000    0.000    0.000 {method 'astype' of 'numpy.ndarray' objects}\n",
    "       16    0.000    0.000    0.000    0.000 {method 'bit_length' of 'int' objects}\n",
    "      294    0.028    0.000    0.039    0.000 {method 'choice' of 'numpy.random.mtrand.RandomState' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'close' of '_io.TextIOWrapper' objects}\n",
    "       61    0.000    0.000    0.000    0.000 {method 'copy' of 'dict' objects}\n",
    "    15006    0.052    0.000    0.052    0.000 {method 'copy' of 'numpy.ndarray' objects}\n",
    "       38    0.000    0.000    0.000    0.000 {method 'count' of 'list' objects}\n",
    "       98    0.000    0.000    0.000    0.000 {method 'decode' of 'bytes' objects}\n",
    "        1    0.000    0.000    0.000    0.000 {method 'difference' of 'set' objects}\n",
    "        4    0.000    0.000    0.000    0.000 {method 'digest' of '_hashlib.HASH' objects}\n",
    "      294    0.009    0.000    0.016    0.000 {method 'dirichlet' of 'numpy.random.mtrand.RandomState' objects}\n",
    "        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}\n",
    "      427    0.000    0.000    0.000    0.000 {method 'discard' of 'set' objects}\n",
    "     5874    0.003    0.000    0.003    0.000 {method 'encode' of 'str' objects}\n",
    "    45065    0.024    0.000    0.024    0.000 {method 'endswith' of 'str' objects}\n",
    "        6    0.000    0.000    0.000    0.000 {method 'extend' of 'bytearray' objects}\n",
    "      129    0.000    0.000    0.000    0.000 {method 'extend' of 'collections.deque' objects}\n",
    "       20    0.000    0.000    0.000    0.000 {method 'extend' of 'google.protobuf.pyext._message.RepeatedCompositeContainer' objects}\n",
    "      414    0.000    0.000    0.000    0.000 {method 'extend' of 'google.protobuf.pyext._message.RepeatedScalarContainer' objects}\n",
    "     4803    0.002    0.000    0.002    0.000 {method 'extend' of 'list' objects}\n",
    "      931    0.000    0.000    0.000    0.000 {method 'find' of 'str' objects}\n",
    "       18    0.000    0.000    0.000    0.000 {method 'findall' of '_sre.SRE_Pattern' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'flush' of '_io._IOBase' objects}\n",
    "      165    0.000    0.000    0.000    0.000 {method 'format' of 'str' objects}\n",
    "      296    0.000    0.000    0.000    0.000 {method 'from_address' of '_ctypes.PyCArrayType' objects}\n",
    "   164547    0.074    0.000    0.083    0.000 {method 'get' of 'dict' objects}\n",
    "       23    0.000    0.000    0.000    0.000 {method 'getrandbits' of '_random.Random' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'getvalue' of '_io.StringIO' objects}\n",
    "       66    0.000    0.000    0.000    0.000 {method 'group' of '_sre.SRE_Match' objects}\n",
    "      200    0.000    0.000    0.000    0.000 {method 'insert' of 'list' objects}\n",
    "       20    0.000    0.000    0.000    0.000 {method 'intersection' of 'set' objects}\n",
    "       10    0.000    0.000    0.000    0.000 {method 'isdigit' of 'str' objects}\n",
    "     1337    0.000    0.000    0.000    0.000 {method 'isidentifier' of 'str' objects}\n",
    "       35    0.000    0.000    0.000    0.000 {method 'issubset' of 'set' objects}\n",
    "    81879    0.023    0.000    0.023    0.000 {method 'items' of 'dict' objects}\n",
    "       77    0.000    0.000    0.000    0.000 {method 'join' of 'str' objects}\n",
    "    19534    0.006    0.000    0.006    0.000 {method 'keys' of 'dict' objects}\n",
    "     3040    0.001    0.000    0.001    0.000 {method 'lower' of 'str' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'lstrip' of 'str' objects}\n",
    "     5617    0.009    0.000    0.009    0.000 {method 'match' of '_sre.SRE_Pattern' objects}\n",
    "      888    0.001    0.000    0.001    0.000 {method 'newbyteorder' of 'numpy.dtype' objects}\n",
    "     9619    0.002    0.000    0.002    0.000 {method 'pop' of 'dict' objects}\n",
    "    54795    0.029    0.000    0.029    0.000 {method 'pop' of 'list' objects}\n",
    "    19900    0.042    0.000    0.042    0.000 {method 'popitem' of 'collections.OrderedDict' objects}\n",
    "      129    0.000    0.000    0.000    0.000 {method 'popleft' of 'collections.deque' objects}\n",
    "      294    0.005    0.000    0.005    0.000 {method 'randint' of 'numpy.random.mtrand.RandomState' objects}\n",
    "       48    0.000    0.000    0.000    0.000 {method 'ravel' of 'numpy.ndarray' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'read' of '_io.FileIO' objects}\n",
    "       77    0.000    0.000    0.000    0.000 {method 'readline' of '_io.StringIO' objects}\n",
    "     1346    0.014    0.000    0.015    0.000 {method 'reduce' of 'numpy.ufunc' objects}\n",
    "        6    0.000    0.000    0.000    0.000 {method 'release' of '_multiprocessing.SemLock' objects}\n",
    "        8    0.000    0.000    0.000    0.000 {method 'release' of '_thread.RLock' objects}\n",
    "    19968    0.006    0.000    0.006    0.000 {method 'release' of '_thread.lock' objects}\n",
    "      964    0.000    0.000    0.000    0.000 {method 'remove' of 'set' objects}\n",
    "       50    0.000    0.000    0.000    0.000 {method 'replace' of 'str' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'reshape' of 'numpy.ndarray' objects}\n",
    "        6    0.000    0.000    0.000    0.000 {method 'rfind' of 'str' objects}\n",
    "       16    0.000    0.000    0.000    0.000 {method 'rpartition' of 'str' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'rsplit' of 'str' objects}\n",
    "       16    0.000    0.000    0.000    0.000 {method 'rstrip' of 'str' objects}\n",
    "        1    0.000    0.000    0.000    0.000 {method 'search' of '_sre.SRE_Pattern' objects}\n",
    "        3    0.000    0.000    0.000    0.000 {method 'seed' of 'numpy.random.mtrand.RandomState' objects}\n",
    "       38    0.000    0.000    0.000    0.000 {method 'setdefault' of 'dict' objects}\n",
    "      100    0.000    0.000    0.000    0.000 {method 'sort' of 'list' objects}\n",
    "      981    0.000    0.000    0.000    0.000 {method 'span' of '_sre.SRE_Match' objects}\n",
    "       24    0.000    0.000    0.000    0.000 {method 'split' of 'str' objects}\n",
    "     8159    0.002    0.000    0.002    0.000 {method 'startswith' of 'str' objects}\n",
    "       19    0.000    0.000    0.000    0.000 {method 'strip' of 'str' objects}\n",
    "      111    0.001    0.000    0.001    0.000 {method 'sub' of '_sre.SRE_Pattern' objects}\n",
    "      294    0.001    0.000    0.003    0.000 {method 'sum' of 'numpy.ndarray' objects}\n",
    "     1719    0.001    0.000    0.003    0.000 {method 'throw' of 'generator' objects}\n",
    "        4    0.000    0.000    0.000    0.000 {method 'to_bytes' of 'int' objects}\n",
    "       54    0.000    0.000    0.000    0.000 {method 'tostring' of 'numpy.ndarray' objects}\n",
    "     2227    0.001    0.000    0.001    0.000 {method 'update' of 'dict' objects}\n",
    "     9163    0.002    0.000    0.003    0.000 {method 'update' of 'set' objects}\n",
    "      321    0.000    0.000    0.000    0.000 {method 'upper' of 'str' objects}\n",
    "      272    0.000    0.000    0.000    0.000 {method 'values' of 'collections.OrderedDict' objects}\n",
    "    18050    0.006    0.000    0.006    0.000 {method 'values' of 'dict' objects}\n",
    "      145    0.000    0.000    0.000    0.000 {method 'values' of 'mappingproxy' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'write' of '_io.FileIO' objects}\n",
    "      431    0.000    0.000    0.000    0.000 {method 'write' of '_io.StringIO' objects}\n",
    "        2    0.000    0.000    0.000    0.000 {method 'write' of '_io.TextIOWrapper' objects}\n",
    "      295    0.010    0.000    0.010    0.000 {resize}\n",
    "       20    0.000    0.000    0.000    0.000 {tensorflow.python.framework.fast_tensor_util.AppendBoolArrayToTensorProto}\n",
    "       28    0.000    0.000    0.000    0.000 {tensorflow.python.framework.fast_tensor_util.AppendFloat32ArrayToTensorProto}\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "eaxsO4xQr7pA"
   },
   "outputs": [
    {
     "ename": "FileNotFoundError",
     "evalue": "[Errno 2] File trace.txt does not exist: 'trace.txt'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mFileNotFoundError\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-49-3c4c33d3fe83>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;31m# df_trace = pd.DataFrame([x.split()[:5]+[''.join(x.split()[5:])] for x in trace.split('\\n')[1:]], columns=['ncalls','tottime','percall','cumtime','percall_cum','filename'])\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mdf_trace\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread_csv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'trace.txt'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0mdf_trace\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdf_trace\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'tottime'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m'float'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'percall'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m'float'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'cumtime'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m'float'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'percall_cum'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m'float'\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/robotic_learning/rl_venv/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36mparser_f\u001b[0;34m(filepath_or_buffer, sep, delimiter, header, names, index_col, usecols, squeeze, prefix, mangle_dupe_cols, dtype, engine, converters, true_values, false_values, skipinitialspace, skiprows, skipfooter, nrows, na_values, keep_default_na, na_filter, verbose, skip_blank_lines, parse_dates, infer_datetime_format, keep_date_col, date_parser, dayfirst, cache_dates, iterator, chunksize, compression, thousands, decimal, lineterminator, quotechar, quoting, doublequote, escapechar, comment, encoding, dialect, error_bad_lines, warn_bad_lines, delim_whitespace, low_memory, memory_map, float_precision)\u001b[0m\n\u001b[1;32m    674\u001b[0m         )\n\u001b[1;32m    675\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 676\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0m_read\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfilepath_or_buffer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    677\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    678\u001b[0m     \u001b[0mparser_f\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/robotic_learning/rl_venv/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36m_read\u001b[0;34m(filepath_or_buffer, kwds)\u001b[0m\n\u001b[1;32m    446\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    447\u001b[0m     \u001b[0;31m# Create the parser.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 448\u001b[0;31m     \u001b[0mparser\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mTextFileReader\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfp_or_buf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    449\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    450\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mchunksize\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0miterator\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/robotic_learning/rl_venv/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, f, engine, **kwds)\u001b[0m\n\u001b[1;32m    878\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moptions\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"has_index_names\"\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mkwds\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"has_index_names\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    879\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 880\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_engine\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mengine\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    881\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    882\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mclose\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/robotic_learning/rl_venv/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36m_make_engine\u001b[0;34m(self, engine)\u001b[0m\n\u001b[1;32m   1112\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_make_engine\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mengine\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"c\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1113\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mengine\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"c\"\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1114\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mCParserWrapper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moptions\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1115\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1116\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mengine\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"python\"\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/robotic_learning/rl_venv/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, src, **kwds)\u001b[0m\n\u001b[1;32m   1889\u001b[0m         \u001b[0mkwds\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"usecols\"\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0musecols\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1890\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1891\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_reader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mparsers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTextReader\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msrc\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1892\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munnamed_cols\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_reader\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munnamed_cols\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1893\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.TextReader.__cinit__\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.TextReader._setup_parser_source\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mFileNotFoundError\u001b[0m: [Errno 2] File trace.txt does not exist: 'trace.txt'"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "# df_trace = pd.DataFrame([x.split()[:5]+[''.join(x.split()[5:])] for x in trace.split('\\n')[1:]], columns=['ncalls','tottime','percall','cumtime','percall_cum','filename'])\n",
    "df_trace = pd.read_csv('trace.txt')\n",
    "df_trace = df_trace.astype({'tottime':'float','percall':'float','cumtime':'float','percall_cum':'float'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vrLVXzXNyRAx"
   },
   "outputs": [],
   "source": [
    "pd.set_option('max_colwidth', 800)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Ongm5L7KuhIH"
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'df_trace' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-51-8cf5eb3da8e0>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mdf_trace\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msort_values\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mby\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'tottime'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mascending\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhead\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m25\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'df_trace' is not defined"
     ]
    }
   ],
   "source": [
    "df_trace.sort_values(by=['tottime'],ascending=False).head(25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "jEzfMX92xRJJ"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "> <ipython-input-51-8cf5eb3da8e0>(1)<module>()\n",
      "-> df_trace.sort_values(by=['tottime'],ascending=False).head(25)\n"
     ]
    },
    {
     "name": "stdin",
     "output_type": "stream",
     "text": [
      "(Pdb)  \n",
      "(Pdb)  q\n"
     ]
    }
   ],
   "source": [
    "pdb.pm()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "b_jnSLpCxRJL"
   },
   "source": [
    "#### Status\n",
    "Self play runs! Now to check if it's actually doing what we want it to..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tSMjtIwExRJM"
   },
   "outputs": [],
   "source": [
    "# Load the TensorBoard notebook extension.\n",
    "%load_ext tensorboard"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9LuM6HgQLE4O"
   },
   "outputs": [],
   "source": [
    "# Launch TensorBoard and navigate to the Profile tab to view performance profile\n",
    "%tensorboard --logdir=logs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "OznNSvZnLITu"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "MuZero_tf.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}