ppo.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from SheepDogEnv import SheepDogEnv\n",
    "import torch\n",
    "from torch import nn\n",
    "import torch.nn.functional as F\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import rl_utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def orthogonal_init(layer, gain=1.0):\n",
    "    nn.init.orthogonal_(layer.weight, gain=gain)\n",
    "    nn.init.constant_(layer.bias, 0)\n",
    "\n",
    "\n",
    "class PolicyNetContinuous(torch.nn.Module):\n",
    "    def __init__(self, state_dim, hidden_dim, action_dim, use_orthogonal_init=False):\n",
    "        super(PolicyNetContinuous, self).__init__()\n",
    "        self.fc1 = torch.nn.Linear(state_dim, hidden_dim)\n",
    "        self.fc2 = nn.Linear(hidden_dim, hidden_dim)\n",
    "        self.fc_mu = torch.nn.Linear(hidden_dim, action_dim)\n",
    "        self.fc_std = torch.nn.Linear(hidden_dim, action_dim)\n",
    "        if use_orthogonal_init:\n",
    "            print(\"------use_orthogonal_init------\")\n",
    "            orthogonal_init(self.fc1)\n",
    "            orthogonal_init(self.fc2)\n",
    "            orthogonal_init(self.fc_mu, gain=0.01)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = torch.tanh(self.fc1(x))\n",
    "        x = torch.tanh(self.fc2(x))\n",
    "        # print(self.fc_mu(x))\n",
    "        mu = np.pi / 2 * torch.tanh(self.fc_mu(x))\n",
    "        std = F.softplus(self.fc_std(x))\n",
    "        return mu, std\n",
    "\n",
    "\n",
    "class ValueNet(torch.nn.Module):\n",
    "    def __init__(self, state_dim, hidden_dim, use_orthogonal_init=False):\n",
    "        super(ValueNet, self).__init__()\n",
    "        self.fc1 = torch.nn.Linear(state_dim, hidden_dim)\n",
    "        self.fc2 = torch.nn.Linear(hidden_dim, hidden_dim)\n",
    "        self.fc3 = torch.nn.Linear(hidden_dim, 1)\n",
    "        if use_orthogonal_init:\n",
    "            print(\"------use_orthogonal_init------\")\n",
    "            orthogonal_init(self.fc1)\n",
    "            orthogonal_init(self.fc2)\n",
    "            orthogonal_init(self.fc3)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = torch.tanh(self.fc1(x))\n",
    "        x = torch.tanh(self.fc2(x))\n",
    "        return self.fc3(x)\n",
    "\n",
    "\n",
    "class PPOContinuous:\n",
    "    ''' 处理连续动作的PPO算法 '''\n",
    "\n",
    "    def __init__(self, state_dim, hidden_dim, action_dim, actor_lr, critic_lr,\n",
    "                 lmbda, epochs, eps, gamma, device, use_orthogonal_init=False):\n",
    "        self.actor = PolicyNetContinuous(state_dim, hidden_dim,\n",
    "                                         action_dim, use_orthogonal_init).to(device)\n",
    "        self.critic = ValueNet(state_dim, hidden_dim,\n",
    "                               use_orthogonal_init).to(device)\n",
    "        self.actor_optimizer = torch.optim.Adam(self.actor.parameters(),\n",
    "                                                lr=actor_lr)\n",
    "        self.critic_optimizer = torch.optim.Adam(self.critic.parameters(),\n",
    "                                                 lr=critic_lr)\n",
    "        self.gamma = gamma\n",
    "        self.lmbda = lmbda\n",
    "        self.epochs = epochs\n",
    "        self.eps = eps\n",
    "        self.device = device\n",
    "\n",
    "    def take_action(self, state, otp=False):\n",
    "        state = torch.tensor(state, dtype=torch.float).to(self.device)\n",
    "        mu, sigma = self.actor(state)\n",
    "        if otp:\n",
    "            print(mu, sigma)\n",
    "        action_dist = torch.distributions.Normal(mu, sigma)\n",
    "        action = action_dist.sample()\n",
    "        return [action.item()]\n",
    "\n",
    "    def update(self, transition_dict):\n",
    "        states = torch.tensor(transition_dict['states'],\n",
    "                              dtype=torch.float).to(self.device)\n",
    "        actions = torch.tensor(transition_dict['actions'],\n",
    "                               dtype=torch.float).view(-1, 1).to(self.device)\n",
    "        rewards = torch.tensor(transition_dict['rewards'],\n",
    "                               dtype=torch.float).view(-1, 1).to(self.device)\n",
    "        next_states = torch.tensor(transition_dict['next_states'],\n",
    "                                   dtype=torch.float).to(self.device)\n",
    "        dones = torch.tensor(transition_dict['dones'],\n",
    "                             dtype=torch.float).view(-1, 1).to(self.device)\n",
    "        rewards = (rewards) / 4.0  # 和TRPO一样,对奖励进行修改,方便训练\n",
    "        td_target = rewards + self.gamma * self.critic(next_states) * (1 -\n",
    "                                                                       dones)\n",
    "        # print(\"td_target:\", td_target, self.critic(states))\n",
    "        td_delta = td_target - self.critic(states)\n",
    "        advantage = rl_utils.compute_advantage(self.gamma, self.lmbda,\n",
    "                                               td_delta.cpu()).to(self.device)\n",
    "        mu, std = self.actor(states)\n",
    "        action_dists = torch.distributions.Normal(mu.detach(), std.detach())\n",
    "        # 动作是正态分布\n",
    "        old_log_probs = action_dists.log_prob(actions)\n",
    "\n",
    "        for _ in range(self.epochs):\n",
    "            mu, std = self.actor(states)\n",
    "            action_dists = torch.distributions.Normal(mu, std)\n",
    "            log_probs = action_dists.log_prob(actions)\n",
    "            ratio = torch.exp(log_probs - old_log_probs)\n",
    "            surr1 = ratio * advantage\n",
    "            surr2 = torch.clamp(ratio, 1 - self.eps, 1 + self.eps) * advantage\n",
    "            actor_loss = torch.mean(-torch.min(surr1, surr2))\n",
    "            critic_loss = torch.mean(\n",
    "                F.mse_loss(self.critic(states), td_target.detach()))\n",
    "            self.actor_optimizer.zero_grad()\n",
    "            self.critic_optimizer.zero_grad()\n",
    "            actor_loss.backward()\n",
    "            critic_loss.backward()\n",
    "            self.actor_optimizer.step()\n",
    "            self.critic_optimizer.step()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "actor_lr = 1e-5\n",
    "critic_lr = 5e-4\n",
    "num_episodes = 2000\n",
    "hidden_dim = 128\n",
    "gamma = 0.9\n",
    "lmbda = 0.9\n",
    "epochs = 10\n",
    "eps = 0.2\n",
    "device = torch.device(\"cuda\") if torch.cuda.is_available() else torch.device(\n",
    "    \"cpu\")\n",
    "\n",
    "\n",
    "env = SheepDogEnv(circle_R=350, sheep_v=70, dog_v=80,\n",
    "                  sec_split_n=10, store_mode=True, render_mode=False)\n",
    "torch.manual_seed(0)\n",
    "state_dim = env._get_obs_array().shape[0]\n",
    "action_dim = env.action_space.shape[0]  # 连续动作空间"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "------use_orthogonal_init------\n",
      "------use_orthogonal_init------\n",
      "0 [ 7.          1.90212771  5.86216097 -2.32315205] -0.6608513593673706 6.3851113775346455 [13.24266047  2.23255339  5.88501811 -2.63072059]\n",
      "1 [13.24266047  2.23255339  5.88501811 -2.63072059] 0.11914443224668503 6.731460133249301 [20.21016984  2.19137242  5.90787525 -2.56668247]\n",
      "2 [20.21016984  2.19137242  5.90787525 -2.56668247] 1.6061344146728516 0.008650824393308887 [21.15311805  1.8543125   5.9307324  -2.20676541]\n",
      "3 [21.15311805  1.8543125   5.9307324  -2.20676541] 0.22038975358009338 5.619265425098869 [28.02561379  1.79968279  5.95358954 -2.12927856]\n",
      "4 [28.02561379  1.79968279  5.95358954 -2.12927856] -0.21411924064159393 6.141387738467486 [34.89747398  1.84231793  5.97644668 -2.14905655]\n",
      "5 [34.89747398  1.84231793  5.97644668 -2.14905655] 0.1340092569589615 5.6310685484296785 [41.84516636  1.81996559  5.99930382 -2.10384707]\n",
      "6 [41.84516636  1.81996559  5.99930382 -2.10384707] -0.47213056683540344 6.032810494420148 [48.18465541  1.88608237  6.02216097 -2.14710671]\n",
      "7 [48.18465541  1.88608237  6.02216097 -2.14710671] 0.14729748666286469 5.572887303415437 [55.11843004  1.86744219  6.04501811 -2.10560938]\n",
      "8 [55.11843004  1.86744219  6.04501811 -2.10560938] -0.7339905500411987 5.5386350811459355 [60.49795899  1.94502452  6.06787525 -2.16033457]\n",
      "9 [60.49795899  1.94502452  6.06787525 -2.16033457] 0.5465404987335205 4.205256398850078 [66.57773016  1.89035223  6.0907324  -2.08280515]\n",
      "10 [66.57773016  1.89035223  6.0907324  -2.08280515] -0.33121734857559204 5.961604347752598 [73.23264807  1.92144121  6.11358954 -2.09103698]\n",
      "11 [73.23264807  1.92144121  6.11358954 -2.09103698] 0.1380476951599121 5.369884238294938 [80.1718409   1.90942588  6.13644668 -2.05616451]\n",
      "12 [80.1718409   1.90942588  6.13644668 -2.05616451] 0.5733623504638672 3.599980124047984 [86.13615223  1.86532768  6.15930382 -1.98920917]\n",
      "13 [86.13615223  1.86532768  6.15930382 -1.98920917] -0.46983852982521057 5.6520376153932474 [92.43198882  1.89962119  6.18216097 -2.00064553]\n",
      "14 [92.43198882  1.89962119  6.18216097 -2.00064553] 0.3548106551170349 4.275271781459691 [99.02584066  1.87506059  6.20501811 -1.95322779]\n",
      "15 [99.02584066  1.87506059  6.20501811 -1.95322779] -0.38235989212989807 5.557813498003666 [105.55266532   1.89980694   6.22787525  -1.955117  ]\n",
      "16 [105.55266532   1.89980694   6.22787525  -1.955117  ] 0.361921101808548 4.023467653016543 [112.12659005   1.87770066   6.2507324   -1.91015357]\n",
      "17 [112.12659005   1.87770066   6.2507324   -1.91015357] 0.4864725172519684 3.25662130208665 [118.35975346   1.85004775   6.27358954  -1.85964352]\n",
      "18 [118.35975346   1.85004775   6.27358954  -1.85964352] -0.04179267957806587 4.8360964952484435 [ 1.25353982e+02  1.85238086e+00  1.32613743e-02 -1.83911948e+00]\n",
      "19 [ 1.25353982e+02  1.85238086e+00  1.32613743e-02 -1.83911948e+00] 0.594416081905365 2.3360681057919286 [ 1.31211886e+02  1.82249973e+00  3.61185172e-02 -1.78638121e+00]\n",
      "20 [ 1.31211886e+02  1.82249973e+00  3.61185172e-02 -1.78638121e+00] -0.7185333967208862 4.786779888221817 [ 1.36559056e+02  1.85624956e+00  5.89756600e-02 -1.79727390e+00]\n",
      "21 [ 1.36559056e+02  1.85624956e+00  5.89756600e-02 -1.79727390e+00] 0.518224835395813 2.4699824955972933 [ 1.42682092e+02  1.83194578e+00  8.18328029e-02 -1.75011298e+00]\n",
      "22 [ 1.42682092e+02  1.83194578e+00  8.18328029e-02 -1.75011298e+00] -0.4657730460166931 4.845063246242413 [ 1.48969591e+02  1.85305095e+00  1.04689946e-01 -1.74836101e+00]\n",
      "23 [ 1.48969591e+02  1.85305095e+00  1.04689946e-01 -1.74836101e+00] -0.4715285301208496 4.816919727544506 [ 1.55238285e+02  1.87353535e+00  1.27547089e-01 -1.74598826e+00]\n",
      "24 [ 1.55238285e+02  1.87353535e+00  1.27547089e-01 -1.74598826e+00] -0.533141553401947 4.770990440126805 [ 1.61306026e+02  1.89559267e+00  1.50404231e-01 -1.74518843e+00]\n",
      "25 [ 1.61306026e+02  1.89559267e+00  1.50404231e-01 -1.74518843e+00] 0.8176142573356628 0.5325487578590904 [166.17225454   1.86485699   0.17326137  -1.69159561]\n",
      "26 [166.17225454   1.86485699   0.17326137  -1.69159561] 0.05331764370203018 3.734330381370168 [173.16270903   1.86270267   0.19611852  -1.66658416]\n",
      "27 [173.16270903   1.86270267   0.19611852  -1.66658416] 0.4140161871910095 2.172159190161521 [179.59337583   1.84702202   0.21897566  -1.62804636]\n",
      "28 [179.59337583   1.84702202   0.21897566  -1.62804636] -0.17896954715251923 3.984543228003964 [186.48573274   1.85370413   0.2418328   -1.61187133]\n",
      "29 [186.48573274   1.85370413   0.2418328   -1.61187133] 0.8927667140960693 -0.783225665516424 [190.95438791   1.82515065   0.26468995  -1.56046071]\n",
      "30 [190.95438791   1.82515065   0.26468995  -1.56046071] -0.7404171228408813 3.9542209016502072 [196.17853996   1.84922376   0.28754709  -1.56167667]\n",
      "31 [196.17853996   1.84922376   0.28754709  -1.56167667] -0.42259836196899414 3.988317062742845 [202.58306817   1.86339581   0.31040423  -1.55299158]\n",
      "32 [202.58306817   1.86339581   0.31040423  -1.55299158] -0.3706555664539337 3.8874672847715193 [209.12307111   1.87552096   0.33326137  -1.54225959]\n",
      "33 [209.12307111   1.87552096   0.33326137  -1.54225959] -0.11042886227369308 3.355935834894502 [216.08181065   1.87909106   0.35611852  -1.52297254]\n",
      "34 [216.08181065   1.87909106   0.35611852  -1.52297254] -0.4634140133857727 3.7996876501539907 [222.36554837   1.8931631    0.37897566  -1.51418744]\n",
      "35 [222.36554837   1.8931631    0.37897566  -1.51418744] -0.09828120470046997 3.1590297408295016 [229.33279698   1.89615814   0.4018328   -1.49432534]\n",
      "36 [229.33279698   1.89615814   0.4018328   -1.49432534] -0.17949187755584717 3.242201096016644 [236.22364445   1.90144852   0.42468995  -1.47675858]\n",
      "37 [236.22364445   1.90144852   0.42468995  -1.47675858] -0.006828202400356531 2.6717102663056114 [243.22348596   1.90164504   0.44754709  -1.45409795]\n",
      "38 [243.22348596   1.90164504   0.44754709  -1.45409795] 0.25235697627067566 1.5420711134804634 [250.00788176   1.89465397   0.47040423  -1.42424974]\n",
      "39 [250.00788176   1.89465397   0.47040423  -1.42424974] 0.8811604976654053 -2.0495748358520887 [254.51897397   1.87343453   0.49326137  -1.38017315]\n",
      "40 [254.51897397   1.87343453   0.49326137  -1.38017315] -0.15634331107139587 2.5370653407087143 [261.43586884   1.87760363   0.51611852  -1.36148512]\n",
      "41 [261.43586884   1.87760363   0.51611852  -1.36148512] 0.0694538801908493 1.6913350076633626 [268.41943177   1.87579383   0.53897566  -1.33681817]\n",
      "42 [268.41943177   1.87579383   0.53897566  -1.33681817] -1.0367653369903564 2.4605402243640606 [272.04921475   1.8979436    0.5618328   -1.3361108 ]\n",
      "43 [272.04921475   1.8979436    0.5618328   -1.3361108 ] 0.09157348424196243 1.445601652404747 [279.02061956   1.89564944   0.58468995  -1.31095949]\n",
      "44 [279.02061956   1.89564944   0.58468995  -1.31095949] 0.15665356814861298 1.0155196621688318 [285.93698944   1.89183007   0.60754709  -1.28428298]\n",
      "45 [285.93698944   1.89183007   0.60754709  -1.28428298] 0.4058758020401001 -0.354108008626437 [292.38135085   1.88237733   0.63040423  -1.25197309]\n",
      "46 [292.38135085   1.88237733   0.63040423  -1.25197309] -1.505594253540039 0.524372935777103 [292.92073918   1.90622606   0.65326137  -1.25296468]\n",
      "47 [292.92073918   1.90622606   0.65326137  -1.25296468] -0.4943358302116394 2.4443565740315214 [299.10116628   1.91732999   0.67611852  -1.24121147]\n",
      "48 [299.10116628   1.91732999   0.67611852  -1.24121147] -0.035456787794828415 1.3229341093132798 [306.09686718   1.91814066   0.69897566  -1.219165  ]\n",
      "49 [306.09686718   1.91814066   0.69897566  -1.219165  ] 1.0576999187469482 -4.622252244374015 [309.59308139   1.89844062   0.7218328   -1.17660782]\n",
      "50 [309.59308139   1.89844062   0.7218328   -1.17660782] 1.2331386804580688 -5.972688804004742 [311.98194651   1.8772688    0.74468995  -1.13257886]\n",
      "51 [311.98194651   1.8772688    0.74468995  -1.13257886] -1.2133798599243164 1.292697186065159 [314.49930648   1.89812132   0.76754709  -1.13057423]\n",
      "52 [314.49930648   1.89812132   0.76754709  -1.13057423] 0.5148224830627441 -2.0604675654307876 [320.61049435   1.8873708    0.79040423  -1.09696657]\n",
      "53 [320.61049435   1.8873708    0.79040423  -1.09696657] -0.11938190460205078 0.7154157949829507 [327.56173236   1.88991594   0.81326137  -1.07665457]\n",
      "54 [327.56173236   1.88991594   0.81326137  -1.07665457] -0.10103490203619003 0.5340125703686891 [334.52677964   1.89202651   0.83611852  -1.05590799]\n",
      "55 [334.52677964   1.89202651   0.83611852  -1.05590799] -0.8604989647865295 1.5760440330185148 [339.13272362   1.90767641   0.85897566  -1.04870075]\n",
      "56 [339.13272362   1.90767641   0.85897566  -1.04870075] 0.7778596878051758 -4.1795950624463645 [344.15470325   1.89340242   0.8818328   -1.01156962]\n",
      "57 [344.15470325   1.89340242   0.8818328   -1.01156962] -0.6687723398208618 1.3934001632161588 [349.67372346   1.90581478   0.90468995  -1.00112483]\n",
      "58 [349.67372346   1.90581478   0.90468995  -1.00112483] 0.24556167423725128 498.56427776112736 [356.46779184   1.90104095   0.92754709  -0.97349386]\n"
     ]
    }
   ],
   "source": [
    "transition_dict = {\n",
    "    'states': [],\n",
    "    'actions': [],\n",
    "    'next_states': [],\n",
    "    'rewards': [],\n",
    "    'dones': []\n",
    "}\n",
    "agent = PPOContinuous(state_dim, hidden_dim, action_dim,\n",
    "                      actor_lr, critic_lr,\n",
    "                      lmbda, epochs, eps, gamma, device,\n",
    "                      use_orthogonal_init=True)\n",
    "env.reset()\n",
    "for i in range(1000):\n",
    "    _st = env._get_obs_array()\n",
    "    act = agent.take_action(state=_st)[0]\n",
    "    st, reward, done, _, _ = env.step(action=act)\n",
    "    transition_dict[\"states\"].append(_st)\n",
    "    transition_dict[\"actions\"].append(act)\n",
    "    transition_dict[\"next_states\"].append(st)\n",
    "    transition_dict[\"rewards\"].append(reward)\n",
    "    transition_dict[\"dones\"].append(done)\n",
    "    print(i, _st, act, reward, st)\n",
    "    if done:\n",
    "        break\n",
    "\n",
    "agent.update(transition_dict=transition_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test: 538 -6.435753868106147 be catched: 10/5\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "-6.435753868106147"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def test_agent(env,agent):\n",
    "    test_dict = {\n",
    "        'states': [],\n",
    "        'actions': [],\n",
    "        'next_states': [],\n",
    "        'rewards': [],\n",
    "        'dones': []\n",
    "    }\n",
    "    catched_num=0\n",
    "    for ii in range(10):\n",
    "        for i in range(1000):\n",
    "            _st = env._get_obs_array()\n",
    "            _st[0]/=env.circle_R\n",
    "            _st[1]=(_st[2]-np.pi)/np.pi\n",
    "            _st[2]=(_st[2]-np.pi)/np.pi\n",
    "            act = agent.take_action(state=_st)[0]\n",
    "            st, reward, done, _, _ = env.step(action=act)\n",
    "            test_dict[\"states\"].append(_st)\n",
    "            test_dict[\"actions\"].append(act)\n",
    "            test_dict[\"next_states\"].append(st)\n",
    "            test_dict[\"rewards\"].append(reward)\n",
    "            test_dict[\"dones\"].append(done)\n",
    "            # print(i, _st, act, reward, st)\n",
    "            # print(reward)\n",
    "            if done:\n",
    "                if reward<-900:\n",
    "                    catched_num+=1\n",
    "                break\n",
    "        env.reset()\n",
    "    print(\"test:\",len(test_dict[\"rewards\"]) ,np.mean(test_dict['rewards']),\"be catched: 10/{}\".format(catched_num))\n",
    "    return np.mean(test_dict[\"rewards\"])\n",
    "test_agent(env=env,agent=agent)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "------use_orthogonal_init------\n",
      "------use_orthogonal_init------\n",
      "0\n",
      "test: 249 14.63211608473937 be catched: 10/1\n",
      "50\n",
      "test: 253 -12.248378797146533 be catched: 10/5\n",
      "100\n",
      "test: 259 9.336241861635907 be catched: 10/2\n",
      "150\n",
      "test: 261 27.06036743032606 be catched: 10/0\n",
      "200\n",
      "test: 259 -20.027362068170245 be catched: 10/6\n",
      "250\n",
      "test: 255 2.6562048063403165 be catched: 10/3\n",
      "300\n",
      "test: 249 0.21523338888339158 be catched: 10/3\n",
      "350\n",
      "test: 242 -13.842753366558496 be catched: 10/5\n",
      "400\n",
      "test: 248 11.350910094648459 be catched: 10/2\n",
      "450\n",
      "test: 251 -15.516451904901661 be catched: 10/5\n",
      "500\n",
      "test: 250 4.386487587465094 be catched: 10/3\n",
      "550\n",
      "test: 247 -2.8296208673775776 be catched: 10/4\n",
      "600\n",
      "test: 244 0.9647551239950917 be catched: 10/3\n",
      "650\n",
      "test: 261 11.445445245122409 be catched: 10/2\n",
      "700\n",
      "test: 255 17.03489758822248 be catched: 10/1\n",
      "750\n",
      "test: 241 11.317054303224038 be catched: 10/2\n",
      "800\n",
      "test: 247 -7.676430277548676 be catched: 10/4\n",
      "850\n",
      "test: 255 -10.190845168695176 be catched: 10/5\n",
      "900\n",
      "test: 251 14.147194943305273 be catched: 10/2\n",
      "950\n",
      "test: 253 -7.492630864270313 be catched: 10/4\n",
      "1000\n",
      "test: 255 1.3916183456805182 be catched: 10/3\n",
      "1050\n",
      "test: 254 25.103977991757326 be catched: 10/0\n",
      "1100\n",
      "test: 249 -1.9726286729509626 be catched: 10/4\n",
      "1150\n",
      "test: 256 4.507428931324149 be catched: 10/3\n",
      "1200\n",
      "test: 252 18.85790576102965 be catched: 10/1\n",
      "1250\n",
      "test: 255 -5.872200196967934 be catched: 10/4\n",
      "1300\n",
      "test: 256 5.04139816104484 be catched: 10/3\n",
      "1350\n",
      "test: 251 -4.285743738436542 be catched: 10/4\n",
      "1400\n",
      "test: 257 -5.887798802930281 be catched: 10/4\n",
      "1450\n",
      "test: 252 2.1732754530320415 be catched: 10/3\n",
      "1500\n",
      "test: 260 12.768685584543453 be catched: 10/2\n",
      "1550\n",
      "test: 255 -6.816459397932406 be catched: 10/4\n",
      "1600\n",
      "test: 260 10.891847426687413 be catched: 10/2\n",
      "1650\n",
      "test: 240 -22.23519822454141 be catched: 10/6\n",
      "1700\n",
      "test: 246 -12.42486662080503 be catched: 10/5\n",
      "1750\n",
      "test: 259 -4.425248317816123 be catched: 10/4\n",
      "1800\n",
      "test: 259 10.354919715489475 be catched: 10/2\n",
      "1850\n",
      "test: 241 13.505629609116042 be catched: 10/2\n",
      "1900\n",
      "test: 249 -5.220914919758028 be catched: 10/4\n",
      "1950\n",
      "test: 255 -0.6967919438074276 be catched: 10/4\n",
      "2000\n",
      "test: 247 4.861423393423311 be catched: 10/3\n",
      "2050\n",
      "test: 258 16.507077305800625 be catched: 10/1\n",
      "2100\n",
      "test: 246 -5.05123821571976 be catched: 10/4\n",
      "2150\n",
      "test: 255 10.630892757107002 be catched: 10/2\n",
      "2200\n",
      "test: 249 28.980206986441587 be catched: 10/0\n",
      "2250\n",
      "test: 239 4.283248237783629 be catched: 10/3\n",
      "2300\n",
      "test: 253 21.181185511326664 be catched: 10/1\n",
      "2350\n",
      "test: 253 10.345757373011285 be catched: 10/2\n",
      "2400\n",
      "test: 254 -3.2162652939753804 be catched: 10/4\n",
      "2450\n",
      "test: 250 -8.14393329345407 be catched: 10/4\n",
      "2500\n",
      "test: 253 3.0453336988926 be catched: 10/3\n",
      "2550\n",
      "test: 256 11.147593540700967 be catched: 10/2\n",
      "2600\n",
      "test: 249 20.72976034983286 be catched: 10/1\n",
      "2650\n",
      "test: 267 18.30417426158572 be catched: 10/1\n",
      "2700\n",
      "test: 245 20.63510153227158 be catched: 10/1\n",
      "2750\n",
      "test: 253 -3.3395855084493635 be catched: 10/4\n",
      "2800\n",
      "test: 249 7.5343324943365175 be catched: 10/2\n",
      "2850\n",
      "test: 239 -14.81983787839462 be catched: 10/5\n",
      "2900\n",
      "test: 247 5.174182817587276 be catched: 10/3\n",
      "2950\n",
      "test: 253 19.81553286655269 be catched: 10/1\n",
      "3000\n",
      "test: 242 3.810546621619613 be catched: 10/3\n",
      "3050\n",
      "test: 244 13.028895049136255 be catched: 10/2\n",
      "3100\n",
      "test: 261 13.702969456602 be catched: 10/2\n",
      "3150\n",
      "test: 245 -4.071071605601149 be catched: 10/4\n",
      "3200\n",
      "test: 245 12.48279015767778 be catched: 10/2\n",
      "3250\n",
      "test: 242 5.265858103761123 be catched: 10/3\n",
      "3300\n",
      "test: 240 4.809438795719972 be catched: 10/3\n",
      "3350\n",
      "test: 245 28.895718143753225 be catched: 10/0\n",
      "3400\n",
      "test: 262 -3.1328621956984892 be catched: 10/4\n",
      "3450\n",
      "test: 246 -7.509748917198487 be catched: 10/4\n",
      "3500\n",
      "test: 243 5.362921360003948 be catched: 10/3\n",
      "3550\n",
      "test: 239 -4.038282569773249 be catched: 10/4\n",
      "3600\n",
      "test: 252 11.91136585371819 be catched: 10/2\n",
      "3650\n",
      "test: 250 -6.558216175996148 be catched: 10/4\n",
      "3700\n",
      "test: 257 12.278339051958957 be catched: 10/2\n",
      "3750\n",
      "test: 248 20.232090133885883 be catched: 10/1\n",
      "3800\n",
      "test: 243 10.512562870725551 be catched: 10/2\n",
      "3850\n",
      "test: 252 -13.937298186437637 be catched: 10/5\n",
      "3900\n",
      "test: 253 3.236732825585675 be catched: 10/3\n",
      "3950\n",
      "test: 239 5.986453628249596 be catched: 10/3\n",
      "4000\n",
      "test: 242 9.497826805060795 be catched: 10/2\n",
      "4050\n",
      "test: 254 14.903128974425368 be catched: 10/2\n",
      "4100\n",
      "test: 248 -3.9176210542732637 be catched: 10/4\n",
      "4150\n",
      "test: 254 22.478711889580303 be catched: 10/1\n",
      "4200\n",
      "test: 267 5.918327570822301 be catched: 10/3\n",
      "4250\n",
      "test: 251 27.719593631814277 be catched: 10/0\n",
      "4300\n",
      "test: 251 -10.543222396796125 be catched: 10/5\n",
      "4350\n",
      "test: 240 13.852100815188912 be catched: 10/2\n",
      "4400\n",
      "test: 253 16.48690825432955 be catched: 10/1\n",
      "4450\n",
      "test: 246 12.192453847180982 be catched: 10/2\n",
      "4500\n",
      "test: 250 12.928575900492238 be catched: 10/2\n",
      "4550\n",
      "test: 249 20.098041470800123 be catched: 10/1\n",
      "4600\n",
      "test: 238 12.324743562685912 be catched: 10/2\n",
      "4650\n",
      "test: 251 10.501496170766453 be catched: 10/2\n",
      "4700\n",
      "test: 251 12.378518698711 be catched: 10/2\n",
      "4750\n",
      "test: 246 -4.2374466733889795 be catched: 10/4\n",
      "4800\n",
      "test: 239 6.051056618884082 be catched: 10/2\n",
      "4850\n",
      "test: 266 13.846953251033852 be catched: 10/2\n",
      "4900\n",
      "test: 244 8.695921948778661 be catched: 10/2\n",
      "4950\n",
      "test: 240 8.66824839756838 be catched: 10/2\n",
      "5000\n",
      "test: 253 9.133577851904745 be catched: 10/2\n",
      "5050\n",
      "test: 257 -4.180816375420762 be catched: 10/4\n",
      "5100\n",
      "test: 261 -9.793395920641512 be catched: 10/5\n",
      "5150\n",
      "test: 254 -3.938517106480268 be catched: 10/4\n",
      "5200\n",
      "test: 244 13.342301521840186 be catched: 10/2\n",
      "5250\n",
      "test: 253 10.769434440480637 be catched: 10/2\n",
      "5300\n",
      "test: 262 -8.073888871549734 be catched: 10/4\n",
      "5350\n",
      "test: 241 -1.407181441446425 be catched: 10/4\n",
      "5400\n",
      "test: 256 -5.693051604404786 be catched: 10/4\n",
      "5450\n",
      "test: 252 2.8780683505211937 be catched: 10/3\n",
      "5500\n",
      "test: 243 11.852669982270376 be catched: 10/2\n",
      "5550\n",
      "test: 254 22.436363982948805 be catched: 10/1\n",
      "5600\n",
      "test: 244 20.738129815629655 be catched: 10/1\n",
      "5650\n",
      "test: 259 2.306985500268632 be catched: 10/3\n",
      "5700\n",
      "test: 243 5.169361178988223 be catched: 10/3\n",
      "5750\n",
      "test: 255 6.191855152237593 be catched: 10/3\n",
      "5800\n",
      "test: 255 15.978557053594034 be catched: 10/1\n",
      "5850\n",
      "test: 245 29.16905986737614 be catched: 10/0\n",
      "5900\n",
      "test: 241 -3.5272859631325377 be catched: 10/4\n",
      "5950\n",
      "test: 251 5.458800219204283 be catched: 10/3\n",
      "6000\n",
      "test: 258 -3.5026972928546374 be catched: 10/4\n",
      "6050\n",
      "test: 242 21.47126861635482 be catched: 10/1\n",
      "6100\n",
      "test: 253 14.2555614496246 be catched: 10/2\n",
      "6150\n",
      "test: 249 -11.591339052758327 be catched: 10/5\n",
      "6200\n",
      "test: 249 1.942696161836599 be catched: 10/3\n",
      "6250\n",
      "test: 244 -19.38582554815792 be catched: 10/6\n",
      "6300\n",
      "test: 247 13.538403194711803 be catched: 10/2\n",
      "6350\n",
      "test: 236 1.0138724360655453 be catched: 10/3\n",
      "6400\n",
      "test: 244 15.986716984907806 be catched: 10/1\n",
      "6450\n",
      "test: 252 -3.8469498312327066 be catched: 10/4\n",
      "6500\n",
      "test: 268 1.6528124656489382 be catched: 10/3\n",
      "6550\n",
      "test: 240 5.071622864135298 be catched: 10/3\n",
      "6600\n",
      "test: 250 20.671795207289087 be catched: 10/1\n",
      "6650\n",
      "test: 258 3.29071785839986 be catched: 10/3\n",
      "6700\n",
      "test: 252 25.14683409154027 be catched: 10/0\n",
      "6750\n",
      "test: 243 -5.379045113807484 be catched: 10/4\n",
      "6800\n",
      "test: 245 10.274744455529179 be catched: 10/2\n",
      "6850\n",
      "test: 239 21.798723597708214 be catched: 10/1\n",
      "6900\n",
      "test: 235 -11.755655285405124 be catched: 10/5\n",
      "6950\n",
      "test: 250 3.7520217532524627 be catched: 10/3\n",
      "7000\n",
      "test: 252 6.103727644898833 be catched: 10/3\n",
      "7050\n",
      "test: 243 -7.381767445705008 be catched: 10/4\n",
      "7100\n",
      "test: 244 5.990680037035494 be catched: 10/3\n",
      "7150\n",
      "test: 246 23.314215092401074 be catched: 10/1\n",
      "7200\n",
      "test: 250 6.070833423439522 be catched: 10/3\n",
      "7250\n",
      "test: 244 -12.067041496874975 be catched: 10/5\n",
      "7300\n",
      "test: 258 26.398255849942736 be catched: 10/0\n",
      "7350\n",
      "test: 251 -1.933073903364577 be catched: 10/4\n",
      "7400\n",
      "test: 251 11.240799967168655 be catched: 10/2\n",
      "7450\n",
      "test: 250 9.80385504140172 be catched: 10/2\n",
      "7500\n",
      "test: 233 27.66410166650509 be catched: 10/0\n",
      "7550\n",
      "test: 249 4.058883071436982 be catched: 10/3\n",
      "7600\n",
      "test: 244 12.584814104422128 be catched: 10/2\n",
      "7650\n",
      "test: 248 -11.14219934799297 be catched: 10/5\n",
      "7700\n",
      "test: 247 13.913548938304146 be catched: 10/2\n",
      "7750\n",
      "test: 249 4.702399482020808 be catched: 10/3\n",
      "7800\n",
      "test: 248 20.65168443600172 be catched: 10/1\n",
      "7850\n",
      "test: 259 -11.07706682998625 be catched: 10/5\n",
      "7900\n",
      "test: 247 12.62224932356184 be catched: 10/2\n",
      "7950\n",
      "test: 248 15.303877101671105 be catched: 10/1\n",
      "8000\n",
      "test: 255 4.8661391751699 be catched: 10/3\n",
      "8050\n",
      "test: 253 1.9581543511658777 be catched: 10/3\n",
      "8100\n",
      "test: 249 23.193946144009306 be catched: 10/0\n",
      "8150\n",
      "test: 233 -15.365535998809586 be catched: 10/5\n",
      "8200\n",
      "test: 251 2.3931058352671966 be catched: 10/3\n",
      "8250\n",
      "test: 244 -5.258544130632896 be catched: 10/4\n",
      "8300\n",
      "test: 242 -1.7954933273943467 be catched: 10/4\n",
      "8350\n",
      "test: 238 19.43712818657081 be catched: 10/1\n",
      "8400\n",
      "test: 250 28.409168938905715 be catched: 10/0\n",
      "8450\n",
      "test: 252 0.9266833032305744 be catched: 10/3\n",
      "8500\n",
      "test: 256 4.226494411168905 be catched: 10/3\n",
      "8550\n",
      "test: 246 1.2814822692870236 be catched: 10/3\n",
      "8600\n",
      "test: 253 13.392654036079037 be catched: 10/2\n",
      "8650\n",
      "test: 244 -5.808193789391861 be catched: 10/4\n",
      "8700\n",
      "test: 249 14.457109190265605 be catched: 10/2\n",
      "8750\n",
      "test: 230 20.898241807687516 be catched: 10/1\n",
      "8800\n",
      "test: 267 -12.428532929611515 be catched: 10/5\n",
      "8850\n",
      "test: 266 9.35984312143535 be catched: 10/2\n",
      "8900\n",
      "test: 231 -5.109785941667887 be catched: 10/4\n",
      "8950\n",
      "test: 254 26.1923256291651 be catched: 10/0\n",
      "9000\n",
      "test: 257 -6.430083164408677 be catched: 10/4\n",
      "9050\n",
      "test: 251 -1.1933423842587891 be catched: 10/4\n",
      "9100\n",
      "test: 242 -14.490230936725146 be catched: 10/5\n",
      "9150\n",
      "test: 240 3.130452880890295 be catched: 10/3\n",
      "9200\n",
      "test: 250 12.100965951205987 be catched: 10/2\n",
      "9250\n",
      "test: 254 4.727830935814225 be catched: 10/3\n",
      "9300\n",
      "test: 236 10.31609994706809 be catched: 10/2\n",
      "9350\n",
      "test: 242 -5.237916111306686 be catched: 10/4\n",
      "9400\n",
      "test: 257 3.3483088488905812 be catched: 10/3\n",
      "9450\n",
      "test: 245 4.643701892571803 be catched: 10/3\n",
      "9500\n",
      "test: 249 11.250185082643293 be catched: 10/2\n",
      "9550\n",
      "test: 246 4.082420916876441 be catched: 10/3\n",
      "9600\n",
      "test: 234 13.216776650947843 be catched: 10/2\n",
      "9650\n",
      "test: 233 4.791077900033818 be catched: 10/3\n",
      "9700\n",
      "test: 249 11.702006654649598 be catched: 10/2\n",
      "9750\n",
      "test: 232 -4.299461029831405 be catched: 10/4\n",
      "9800\n",
      "test: 240 7.139113610082676 be catched: 10/3\n",
      "9850\n",
      "test: 246 -0.24758057960712068 be catched: 10/3\n",
      "9900\n",
      "test: 242 28.144449702431633 be catched: 10/0\n",
      "9950\n",
      "test: 252 -9.691671169744572 be catched: 10/5\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x2d8dad2d450>]"
      ]
     },
     "execution_count": 132,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "actor_lr = 7e-6\n",
    "critic_lr = 1e-4\n",
    "num_episodes = 2000\n",
    "hidden_dim = 128\n",
    "gamma = 0.5\n",
    "lmbda = 0.9\n",
    "epochs = 5\n",
    "eps = 0.2\n",
    "device = torch.device(\"cuda\") if torch.cuda.is_available() else torch.device(\n",
    "    \"cpu\")\n",
    "\n",
    "agent = PPOContinuous(state_dim, hidden_dim, action_dim,\n",
    "                      actor_lr, critic_lr,\n",
    "                      lmbda, epochs, eps, gamma, device,\n",
    "                      use_orthogonal_init=True)\n",
    "\n",
    "env = SheepDogEnv(circle_R=350, sheep_v=80, dog_v=80,\n",
    "                  sec_split_n=5, store_mode=True, render_mode=False)\n",
    "r_list=[]\n",
    "\n",
    "for ii in range(10000):\n",
    "    transition_dict = {\n",
    "        'states': [],\n",
    "        'actions': [],\n",
    "        'next_states': [],\n",
    "        'rewards': [],\n",
    "        'dones': []\n",
    "    }\n",
    "    for i in range(1000):\n",
    "        _st = env._get_obs_array()\n",
    "\n",
    "        L1 = _st[0]\n",
    "        L2 = env.circle_R\n",
    "        theta1 = _st[1]\n",
    "        theta2 = _st[2]\n",
    "        L3 = np.sqrt(L1**2 + L2**2 - 2*L1*L2*np.cos(theta2-theta1))\n",
    "        theta3 = np.arcsin(L2/L3*np.sin(theta2-theta1))\n",
    "        if(np.abs(theta2-theta1) < np.arccos(L1/L2)):\n",
    "            theta3 = np.pi/2 if theta3 >= 0 else -np.pi/2\n",
    "        \n",
    "        _st[0]/=env.circle_R\n",
    "        # _st[1]=(_st[1]-np.pi)/np.pi\n",
    "        # _st[2]=(_st[2]-np.pi)/np.pi\n",
    "        act = agent.take_action(state=_st)[0] if np.random.random()>(0.4)*(1-i/10000) else theta3\n",
    "\n",
    "        st, reward, done, _, _ = env.step(action=act)\n",
    "        st[0]/=env.circle_R\n",
    "        # st[1]=(st[1]-np.pi)/np.pi\n",
    "        # st[2]=(st[2]-np.pi)/np.pi\n",
    "        transition_dict[\"states\"].append(_st)\n",
    "        transition_dict[\"actions\"].append(act)\n",
    "        transition_dict[\"next_states\"].append(st)\n",
    "        transition_dict[\"rewards\"].append(reward)\n",
    "        transition_dict[\"dones\"].append(done)\n",
    "        # print(i, _st, act, reward, st)\n",
    "        # print(reward)\n",
    "        if done:\n",
    "            break\n",
    "    env.reset()\n",
    "    if(i<2000):\n",
    "        env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "        env.dog_theta=np.array([(env.sheep_polar_coor[1]-np.pi)%(np.pi*2)])\n",
    "    elif(i<4000):\n",
    "        env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "        env.dog_theta=np.array([env.sheep_polar_coor[1]])\n",
    "    elif(i<7000):\n",
    "        env.sheep_polar_coor=np.array([np.random.randint(env.sheep_v,env.circle_R-50),np.random.random()*np.pi*2])\n",
    "    # elif(i<4000):\n",
    "    #     env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "    #     env.dog_theta=np.array([(env.sheep_polar_coor[1]-np.pi)%(np.pi*2)])\n",
    "    # elif(i<5000):\n",
    "    #     env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "    #     env.dog_theta=np.array([env.sheep_polar_coor[1]])\n",
    "    # elif(i<6000):\n",
    "    #     env.sheep_polar_coor=np.array([np.random.randint(env.sheep_v,env.circle_R-50),np.random.random()*np.pi*2])\n",
    "    # elif(i<7000):\n",
    "    #     env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "    #     env.dog_theta=np.array([(env.sheep_polar_coor[1]-np.pi)%(np.pi*2)])\n",
    "    # elif(i<8000):\n",
    "    #     env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "    #     env.dog_theta=np.array([env.sheep_polar_coor[1]])\n",
    "    # elif(i<9000):\n",
    "    #     env.sheep_polar_coor=np.array([np.random.randint(env.sheep_v,env.circle_R-50),np.random.random()*np.pi*2])\n",
    "    agent.update(transition_dict)\n",
    "    if(ii % 50 == 0):\n",
    "        print(ii)\n",
    "        r_list.append(test_agent(env, agent))\n",
    "plt.plot(r_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-0.776055097579956 4.276175585202766\n",
      "{'distance': array([353.14461242])}\n",
      "critic:_st_score: tensor([3.7420], device='cuda:0', grad_fn=<AddBackward0>)\n",
      "critic:st_score: tensor([-1.1441], device='cuda:0', grad_fn=<AddBackward0>)\n",
      "critic:st_score: tensor([15.8134], device='cuda:0', grad_fn=<AddBackward0>)\n"
     ]
    }
   ],
   "source": [
    "env.reset()\n",
    "_st = env._get_obs_array()\n",
    "_st[0]/=env.circle_R\n",
    "# _st[2]=(_st[2]-np.pi)/np.pi\n",
    "act = agent.take_action(state=_st)[0]\n",
    "st, reward, done, _, info = env.step(action=act)\n",
    "# print(_st)\n",
    "print(act, reward)\n",
    "# print(st)\n",
    "print(info)\n",
    "print(\"critic:_st_score:\",agent.critic(torch.tensor(_st,dtype=torch.float).to(device)))\n",
    "print(\"critic:st_score:\",agent.critic(torch.tensor(st,dtype=torch.float).to(device)))\n",
    "print(\"critic:st_score:\",agent.critic(torch.tensor([0.2,0,0.3,0.3],dtype=torch.float).to(device)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "torch.save(agent.actor.state_dict(),\"actor-4\")\n",
    "# torch.save(agent.critic.state_dict(),\"critic-1\")\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([-0.3276], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.5475], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.00857143  0.85596093 -0.14403907 -3.14159265] 1.935356855392456 0.4032782272187774\n",
      "tensor([-0.2578], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.3012], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.00972365  0.54793932 -0.13676341 -2.15105708] -0.21455302834510803 2.6758121405991213\n",
      "tensor([-0.2660], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.3171], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.01819033  0.57992766 -0.12948776 -2.22869425] -0.08088591694831848 2.6737848580752486\n",
      "tensor([-0.2680], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.3129], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.0267427   0.58817181 -0.1222121  -2.23173687] 0.2411925196647644 2.3915335760616516\n",
      "tensor([-0.2628], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2862], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.03512574  0.56960792 -0.11493645 -2.15055955] 0.7381347417831421 1.4021157704474663\n",
      "tensor([-0.2503], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2375], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.04186546  0.52561451 -0.10766079 -1.98949305] -0.9912803173065186 2.139856096654104\n",
      "tensor([-0.2631], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2690], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.04710847  0.57426426 -0.10038514 -2.1194736 ] -0.4681514501571655 2.5753252512808107\n",
      "tensor([-0.2688], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2771], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.05489408  0.59671047 -0.09310948 -2.16713329] 2.0718719959259033 -1.894698622335667\n",
      "tensor([-0.2547], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2282], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.05133013  0.54992312 -0.08583383 -1.99728937] -1.0146762132644653 2.064621731448523\n",
      "tensor([-0.2655], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2534], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.05632735  0.59117717 -0.07855818 -2.10403565] -1.8220829963684082 0.20894128411559887\n",
      "tensor([-0.2771], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2854], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.05482827  0.63956254 -0.07128252 -2.23318561] 0.41956138610839844 2.1026881885805473\n",
      "tensor([-0.2723], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2595], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.0627535   0.62184236 -0.06400687 -2.15465889] -0.39109355211257935 2.5881933629879663\n",
      "tensor([-0.2759], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2610], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.07075321  0.63654733 -0.05673121 -2.17799877] -0.37117981910705566 2.596479429793183\n",
      "tensor([-0.2789], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2606], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.07880227  0.64910887 -0.04945556 -2.19460489] -0.43403875827789307 2.576743406368264\n",
      "tensor([-0.2821], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2609], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.08665392  0.6623537  -0.0421799  -2.21335759] -1.391900658607483 1.0937409983588169\n",
      "tensor([-0.2890], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2766], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.08858163  0.69270877 -0.03490425 -2.28586371] 0.3898153007030487 2.1901272017536453\n",
      "tensor([-0.2861], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2568], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.09656497  0.68196962 -0.02762859 -2.22926854] 0.5343418121337891 1.8408867125339958\n",
      "tensor([-0.2824], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2348], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.1040332   0.66860948 -0.02035294 -2.16443928] 0.7918357253074646 1.1639907595779317\n",
      "tensor([-0.2774], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2092], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.11022388  0.65098516 -0.01307728 -2.0862137 ] -1.29338800907135 1.331399427157081\n",
      "tensor([-0.2828], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2192], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.11287274  0.67425384 -0.00580163 -2.13645725] 0.556124210357666 1.6004671093640908\n",
      "tensor([-0.2793], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1981], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 1.20237685e-01  6.62272212e-01  1.47402439e-03 -2.07595873e+00] -1.214385747909546 1.4735376122981267\n",
      "tensor([-0.2841], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2058], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.1234899   0.68299223  0.00874968 -2.11819545] -0.8998430967330933 2.0161061136128136\n",
      "tensor([-0.2879], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2095], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.12899385  0.69956593  0.01602533 -2.14740613] -0.290619432926178 2.4849908942774155\n",
      "tensor([-0.2892], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2033], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.13722783  0.70526336  0.02330099 -2.14244796] -0.9517178535461426 1.9074136079888984\n",
      "tensor([-0.2927], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2062], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.14237294  0.72087667  0.03057664 -2.16864149] -0.6949721574783325 2.2567483763381895\n",
      "tensor([-0.2954], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2055], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.14905751  0.73260064  0.0378523  -2.18261628] -0.628262996673584 2.3205846332459004\n",
      "tensor([-0.2977], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.2035], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.15607354  0.74287689  0.04512795 -2.19204295] 0.11190846562385559 2.3165136701879234\n",
      "tensor([-0.2971], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1910], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.16459414  0.74102572  0.05240361 -2.16337017] -1.012857437133789 1.7448565702224599\n",
      "tensor([-0.3000], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1923], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.16928843  0.7547025   0.05967926 -2.18347989] -1.2464754581451416 1.2483158756738617\n",
      "tensor([-0.3032], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1952], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.1722116   0.76972526  0.06695492 -2.20781814] 1.8002686500549316 -1.5048997674086693\n",
      "tensor([-0.2982], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1732], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.17046638  0.75413325  0.07423057 -2.13597726] 0.9252908229827881 0.6874309190563136\n",
      "tensor([-0.2945], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1523], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.17575639  0.74172993  0.08150622 -2.07415394] 0.25643035769462585 1.9034238949793993\n",
      "tensor([-0.2933], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1381], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.18406038  0.73797024  0.08878188 -2.03948537] -1.758068561553955 0.0037643505558065393\n",
      "tensor([-0.2961], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1417], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.1826588   0.75265126  0.09605753 -2.06275002] 0.46932968497276306 1.517062012030719\n",
      "tensor([-0.2941], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1253], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.1903429   0.74616772  0.10333319 -2.01952423] -0.7702617645263672 2.0012391274091916\n",
      "tensor([-0.2962], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1231], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.19658548  0.75583333  0.11060884 -2.02703251] -0.9348678588867188 1.7814993523669962\n",
      "tensor([-0.2986], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1221], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.20179412  0.76671302  0.1178845  -2.03835492] 1.740407943725586 -1.7010908936298859\n",
      "tensor([-0.2940], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1016], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.20052532  0.75329818  0.12516015 -1.97335381] -1.0789973735809326 1.529247072206109\n",
      "tensor([-0.2965], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.1015], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.20471234  0.76504912  0.13243581 -1.98741335] 1.0950872898101807 -0.1559854685459523\n",
      "tensor([-0.2927], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0814], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.20877688  0.75342919  0.13971146 -1.92805111] -0.8483833074569702 1.8151722478098309\n",
      "tensor([-0.2948], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0793], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.21454067  0.76297127  0.14698712 -1.93517129] 0.4458240270614624 1.2070160518464643\n",
      "tensor([-0.2931], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0639], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.22230502  0.75767885  0.15426277 -1.89568752] -1.0585417747497559 1.4870032543791951\n",
      "tensor([-0.2953], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0628], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.22662943  0.76817436  0.16153842 -1.90580301] 0.30713951587677 1.3628564520484243\n",
      "tensor([-0.2940], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0489], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.23481404  0.7646614   0.16881408 -1.87190955] -0.3828110694885254 1.9241511440498136\n",
      "tensor([-0.2949], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0417], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.24278616  0.76885915  0.17608973 -1.86224004] 1.3498585224151611 -1.2041839307209516\n",
      "tensor([-0.2910], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0226], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.24480744  0.75798298  0.18336539 -1.80521439] -0.02241721749305725 1.5398486932641475\n",
      "tensor([-0.2908], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0119], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.25337679  0.75822434  0.19064104 -1.78311553] -0.39307159185409546 1.7284630614045149\n",
      "tensor([-0.2915], device='cuda:0', grad_fn=<MulBackward0>) tensor([1.0046], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.26131515  0.76222361  0.1979167  -1.77282245] 0.6614248752593994 0.21088640800055547\n",
      "tensor([-0.2893], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9886], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.26813072  0.75597298  0.20519235 -1.73032836] -0.5805873870849609 1.6129796023380933\n",
      "tensor([-0.2904], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9828], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.27533778  0.76140857  0.21246801 -1.72454765] 0.15392208099365234 1.0320869384786624\n",
      "tensor([-0.2896], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9707], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.28381091  0.7599347   0.21974366 -1.69706019] -1.2198175191879272 0.9887249293178257\n",
      "tensor([-0.2913], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9686], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.28687086  0.76886686  0.22701932 -1.70226426] 0.005120396614074707 1.1585818146934528\n",
      "tensor([-0.2910], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9578], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.29544218  0.76881957  0.23429497 -1.67925856] -0.8353633880615234 1.3674239358643552\n",
      "tensor([-0.2923], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9533], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.30125988  0.77553586  0.24157062 -1.67750125] 0.985308051109314 -0.9474119080937438\n",
      "tensor([-0.2894], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9369], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.30607988  0.76810594  0.24884628 -1.63130235] -1.1105289459228516 1.030458505739354\n",
      "tensor([-0.2910], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9338], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.30998234  0.77599249  0.25612193 -1.63322152] 0.36874672770500183 0.29716249507101944\n",
      "tensor([-0.2896], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9206], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.31799261  0.77289981  0.26339759 -1.60064845] 1.6370724439620972 -2.689440254332058\n",
      "tensor([-0.2858], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.9042], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.31754014  0.76432543  0.27067324 -1.5508541 ] -0.42349323630332947 1.0896607592372274\n",
      "tensor([-0.2864], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8966], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.32537343  0.76777144  0.2779489  -1.53882291] -2.7865757942199707 -2.5090189775923655\n",
      "tensor([-0.2853], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8916], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.31735049  0.77075997  0.28522455 -1.52535452] 1.2126688957214355 -2.0054078175923653\n",
      "tensor([-0.2820], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8751], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.32045553  0.76278528  0.29250021 -1.47744412] -1.9772303104400635 -0.5635629634987748\n",
      "tensor([-0.2828], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8735], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.31716466  0.77068769  0.29977586 -1.47941313] 0.7316368818283081 -1.0500047791777813\n",
      "tensor([-0.2804], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8586], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.32359318  0.76505442  0.30705152 -1.43885855] -0.24653640389442444 0.6166600491243127\n",
      "tensor([-0.2806], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8498], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.33191203  0.76706053  0.31432717 -1.4223038 ] 0.10562580823898315 0.019532693360239772\n",
      "tensor([-0.2798], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8387], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.34043689  0.76621558  0.32160282 -1.39679218] 0.4213124215602875 -0.7191392764651814\n",
      "tensor([-0.2782], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8257], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.34827641  0.76301178  0.32887848 -1.36386999] -0.7075169682502747 0.6908012229766008\n",
      "tensor([-0.2791], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8200], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.35483425  0.76800953  0.33615413 -1.35671375] -0.765110969543457 0.6688679945883015\n",
      "tensor([-0.2800], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8146], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.36106567  0.77324349  0.34342979 -1.35029956] -0.6089102029800415 0.5778418744316336\n",
      "tensor([-0.2807], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.8081], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.36812922  0.77748277  0.35070544 -1.34076051] 0.4558708369731903 -1.0790905675596545\n",
      "tensor([-0.2791], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7954], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.37584426  0.77428684  0.3579811  -1.30786307] -0.40145573019981384 0.25110973958756677\n",
      "tensor([-0.2794], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7877], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.38374882  0.77706509  0.36525675 -1.29373404] 0.1475260853767395 -0.663581116352661\n",
      "tensor([-0.2785], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7768], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.39222917  0.7760426   0.37253241 -1.26766467] -0.86885666847229 0.37744791569625685\n",
      "tensor([-0.2794], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7716], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.39781759  0.78127981  0.37980806 -1.26126071] 0.2692648470401764 -1.0736598696326327\n",
      "tensor([-0.2782], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7601], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.40608656  0.77949248  0.38708372 -1.23278848] -0.8322248458862305 0.24828525424443493\n",
      "tensor([-0.2790], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7547], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.41190589  0.78439049  0.39435937 -1.22531891] 0.04224476218223572 -0.8037555548543589\n",
      "tensor([-0.2783], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7446], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.42046982  0.78411645  0.40163502 -1.20160085] -0.36204126477241516 -0.2650460264828558\n",
      "tensor([-0.2784], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7367], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.42849637  0.78637167  0.40891068 -1.18582868] -0.08182714879512787 -0.7773283315173354\n",
      "tensor([-0.2779], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7273], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.43703968  0.78688194  0.41618633 -1.16457458] -0.22318251430988312 -0.6452598362563213\n",
      "tensor([-0.2777], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7187], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.44540256  0.78823775  0.42346199 -1.14597686] -0.9847611784934998 -0.04539238740109175\n",
      "tensor([-0.2784], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7137], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.45019973  0.79328709  0.43073764 -1.13898268] -0.11326970160007477 -0.96591106879291\n",
      "tensor([-0.2780], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.7047], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.45871726  0.79395936  0.4380133  -1.11823753] 0.19134020805358887 -1.6708675761795835\n",
      "tensor([-0.2770], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6943], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.4671351   0.79284861  0.44528895 -1.09189088] -0.45560339093208313 -0.6588168697432046\n",
      "tensor([-0.2771], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6869], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.47484719  0.79537681  0.45256461 -1.0769763 ] -0.937877893447876 -0.3148286406590869\n",
      "tensor([-0.2777], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6816], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.47996695  0.7999604   0.45984026 -1.06851894] 0.035166800022125244 -1.6223883063184614\n",
      "tensor([-0.2769], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6721], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.48853317  0.79976404  0.46711592 -1.04504492] -0.7207703590393066 -0.5854209904720081\n",
      "tensor([-0.2773], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6660], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.49500519  0.80340171  0.47439157 -1.03361585] -0.2342427372932434 -1.3023093197545312\n",
      "tensor([-0.2770], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6578], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.50334647  0.80465984  0.48166722 -1.01471123] -0.25979045033454895 -1.3584720384049818\n",
      "tensor([-0.2768], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6497], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.51163501  0.80602968  0.48894288 -0.99615758] -0.1950024515390396 -1.5751474632783689\n",
      "tensor([-0.2764], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6414], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.52004664  0.80704627  0.49621853 -0.97649415] -0.027212947607040405 -2.011235850983699\n",
      "tensor([-0.2757], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6325], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.52861495  0.80718671  0.50349419 -0.95407821] -0.2230874001979828 -1.7491455099072288\n",
      "tensor([-0.2753], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6244], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.53697731  0.80831084  0.51076984 -0.93475261] -0.7503316402435303 -1.059498521105751\n",
      "tensor([-0.2756], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6184], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.54327844  0.81173535  0.5180455  -0.92265387] -0.20941919088363647 -1.9527685558203411\n",
      "tensor([-0.2752], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6104], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.55166547  0.81276352  0.52532115 -0.90302683] -0.2839145362377167 -1.9204989589222874\n",
      "tensor([-0.2749], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.6027], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.5598989   0.81412852  0.53259681 -0.88445796] -0.02386099100112915 -2.5495789614393787\n",
      "tensor([-0.2741], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5941], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.56846793  0.81424303  0.53987246 -0.86196056] -0.5941380858421326 -1.613662066345468\n",
      "tensor([-0.2741], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5877], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.57559048  0.81689656  0.54714811 -0.84743972] -0.5968369245529175 -1.6871153909440875\n",
      "tensor([-0.2741], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5813], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.58269997  0.81952817  0.55442377 -0.83285003] 0.6154910922050476 -4.258891883459503\n",
      "tensor([-0.2722], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5711], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.58971923  0.81685695  0.56169942 -0.80160101] 0.5896134376525879 -4.366633348241649\n",
      "tensor([-0.2703], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5612], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.59686245  0.81431516  0.56897508 -0.7707586 ] 0.5291664600372314 -4.4021621009255645\n",
      "tensor([-0.2686], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5515], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.60427704  0.81203586  0.57625073 -0.74074081] 0.8168501853942871 -5.099808725727299\n",
      "tensor([-0.2663], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5414], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.61017635  0.80877614  0.58352639 -0.70764298] -0.6916682720184326 -2.1537872478280606\n",
      "tensor([-0.2662], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5356], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.61680215  0.81159754  0.59080204 -0.6936495 ] -0.5579995512962341 -2.4677311289434285\n",
      "tensor([-0.2661], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5293], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.62408993  0.81391236  0.5980777  -0.67806459] -0.710464358329773 -2.2793487742652814\n",
      "tensor([-0.2660], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5236], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.63061236  0.81673411  0.60535335 -0.66407222] 0.07429921627044678 -4.010624222106781\n",
      "tensor([-0.2650], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5155], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.63916046  0.81641724  0.61262901 -0.64021961] 0.14108744263648987 -4.302511393013905\n",
      "tensor([-0.2638], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5074], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.64764784  0.81582484  0.61990466 -0.61550141] -0.7911177277565002 -2.445217323356904\n",
      "tensor([-0.2638], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.5020], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.65370241  0.81879298  0.62718031 -0.60196895] -0.6941823959350586 -2.6905033248586157\n",
      "tensor([-0.2636], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4964], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.66031298  0.82143645  0.63445597 -0.5874165 ] 0.05322316288948059 -4.4219248006689105\n",
      "tensor([-0.2626], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4888], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.66887243  0.82121945  0.64173162 -0.56387764] 0.27772057056427 -5.062959168614842\n",
      "tensor([-0.2612], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4807], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.6771195   0.82011474  0.64900728 -0.53754993] 0.4317026436328888 -5.523273437554451\n",
      "tensor([-0.2595], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4724], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.68491393  0.81844795  0.65628293 -0.50945643] 0.47486287355422974 -5.745023941557273\n",
      "tensor([-0.2577], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4642], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.69254807  0.81664669  0.66355859 -0.48094045] -0.48697662353515625 -3.7277866110875353\n",
      "tensor([-0.2573], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4583], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.70013457  0.81847029  0.67083424 -0.46381231] 0.04723694920539856 -5.09446411149572\n",
      "tensor([-0.2562], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4512], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.70869656  0.8182885   0.6781099  -0.44038407] -0.5833311676979065 -3.719622184102914\n",
      "tensor([-0.2559], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4457], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.71586612  0.8203878   0.68538555 -0.42412207] 0.6810920238494873 -6.495368965345434\n",
      "tensor([-0.2537], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4377], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.72254529  0.81801021  0.69266121 -0.39379552] 0.006139695644378662 -5.359727781702802\n",
      "tensor([-0.2527], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4309], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.73111656  0.8179873   0.69993686 -0.3708664 ] -0.23713679611682892 -4.899681326784623\n",
      "tensor([-0.2519], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4248], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.73945086  0.8188541   0.70721251 -0.35073237] 0.6617108583450317 -6.729349338744923\n",
      "tensor([-0.2498], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4171], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.7462318   0.81660746  0.71448817 -0.32081722] -0.49769285321235657 -4.464994145406446\n",
      "tensor([-0.2493], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4118], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.75377451  0.81833547  0.72176382 -0.30338878] -0.11562694609165192 -5.493629152817712\n",
      "tensor([-0.2483], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.4056], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.76228934  0.8187484   0.72903948 -0.28182889] 0.25031906366348267 -6.336649648306564\n",
      "tensor([-0.2468], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3989], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.77059656  0.81787135  0.73631513 -0.2562164 ] -0.6094794273376465 -4.430616111699886\n",
      "tensor([-0.2463], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3939], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.77764015  0.81987978  0.74359079 -0.23966895] 0.15935279428958893 -6.276770753428853\n",
      "tensor([-0.2450], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3875], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.78610415  0.81932905  0.75086644 -0.21508162] -0.024821847677230835 -5.955334341842724\n",
      "tensor([-0.2438], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3816], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.79467297  0.81941426  0.7581421  -0.19249218] -0.12595167756080627 -5.764141572377598\n",
      "tensor([-0.2428], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3759], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.80317722  0.81984098  0.76541775 -0.17097563] -0.008576631546020508 -5.98915165495479\n",
      "tensor([-0.2417], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3700], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.81174834  0.81986981  0.77269341 -0.14820905] 0.2592378258705139 -6.2926380155363875\n",
      "tensor([-0.2401], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3639], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.8200363   0.81901692  0.77996906 -0.12267247] 0.29200249910354614 -6.084808163134912\n",
      "tensor([-0.2385], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3578], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.82824857  0.81806863  0.78724471 -0.09683619] -0.8655918836593628 -3.6189214886273744\n",
      "tensor([-0.2380], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3537], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.83383003  0.82056029  0.79452037 -0.08180681] -0.47383880615234375 -4.4493008805387175\n",
      "tensor([-0.2373], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3490], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.84146617  0.82203982  0.80179602 -0.06359775] -0.37766969203948975 -4.36275998945279\n",
      "tensor([-0.2365], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3442], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.84943942  0.82322425  0.80907168 -0.04446162] -0.20362679660320282 -4.167314417440247\n",
      "tensor([-0.2355], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3392], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.85783551  0.82386743  0.81634733 -0.02362507] -0.45559024810791016 -3.2084590812317995\n",
      "tensor([-0.2348], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3346], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[ 0.86554089  0.82525438  0.82362299 -0.00512517] -0.24076873064041138 -2.9343255458809283\n",
      "tensor([-0.2352], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3328], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.87386746 0.82599886 0.82599886 0.        ] 0.1760702133178711 -2.954066025856058\n",
      "tensor([-0.2359], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3323], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.88230765 0.82545721 0.82545721 0.        ] -0.23677195608615875 -2.917094957645368\n",
      "tensor([-0.2367], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3315], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.89064221 0.82617577 0.82617577 0.        ] 0.5664237141609192 -2.5355997997748148\n",
      "tensor([-0.2372], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3312], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.89788678 0.82454517 0.82454517 0.        ] -0.2704036235809326 -2.8920018932403053\n",
      "tensor([-0.2381], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3304], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.90614964 0.82534945 0.82534945 0.        ] 0.4141331911087036 -2.748674150209986\n",
      "tensor([-0.2386], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3300], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.91400299 0.82414826 0.82414826 0.        ] -0.20815648138523102 -2.9358359460225074\n",
      "tensor([-0.2395], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3293], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.9223911  0.82475954 0.82475954 0.        ] -0.3386363685131073 -2.8311505251977813\n",
      "tensor([-0.2404], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3285], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.9304801  0.82573363 0.82573363 0.        ] -0.10095623135566711 -2.9848638198729027\n",
      "tensor([-0.2412], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3278], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.93900828 0.82602647 0.82602647 0.        ] -0.3749633729457855 -2.7933841877883197\n",
      "tensor([-0.2421], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3270], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.94698938 0.82708164 0.82708164 0.        ] -0.3031131625175476 -2.8644348848408097\n",
      "tensor([-0.2429], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3262], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.95517348 0.82793426 0.82793426 0.        ] -0.5472517013549805 -2.5654901888765878\n",
      "tensor([-0.2438], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3254], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.96250345 0.82940926 0.82940926 0.        ] -0.5065767765045166 -2.6263518418374936\n",
      "tensor([-0.2446], device='cuda:0', grad_fn=<MulBackward0>) tensor([0.3247], device='cuda:0', grad_fn=<SoftplusBackward0>)\n",
      "[0.97000731 0.83077396 0.83077396 0.        ] 0.2448594570159912 -1002.9112865371017\n"
     ]
    }
   ],
   "source": [
    "\n",
    "env = SheepDogEnv(circle_R=350, sheep_v=30, dog_v=80,\n",
    "                  sec_split_n=10, store_mode=False, render_mode=False)\n",
    "\n",
    "# env.reset()\n",
    "env.sheep_polar_coor=np.array([env.sheep_v,np.random.random()*np.pi*2])\n",
    "env.dog_theta=np.array([(env.sheep_polar_coor[1]-np.pi)%(np.pi*2)])\n",
    "for i in range(2000):\n",
    "    _st = env._get_obs_array()\n",
    "    _st[0]/=env.circle_R\n",
    "    _st[1]=(_st[1]-np.pi)/np.pi\n",
    "    _st[2]=(_st[2]-np.pi)/np.pi\n",
    "    action = agent.take_action(state=_st, otp=True)[0]\n",
    "    observation, reward, done, _, info = env.step(action)  # 和环境交互\n",
    "    print(_st, action, reward)\n",
    "    if done:\n",
    "        break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "rl",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}