From bcface6a6d630c84f0c0c44240a4073676e817bb Mon Sep 17 00:00:00 2001
From: Bo Zhou <2466956298@qq.com>
Date: Fri, 29 May 2020 13:15:18 +0800
Subject: [PATCH] replace tensorboard with summary to support VDL (#276)
* replace tensorboard with summary to support VDL in the future
* unittest
* rename keys for record
* yapf
---
benchmark/torch/a2c/train.py | 28 +-
benchmark/torch/dqn/train.py | 25 +-
benchmark/torch/td3/train.py | 9 +-
docs/tutorial/tensorboard.rst | 16 +-
examples/A2C/train.py | 6 +-
examples/DQN_variant/train.py | 12 +-
examples/ES/README.md | 2 +-
examples/ES/train.py | 4 +-
examples/GA3C/train.py | 4 +-
examples/IMPALA/train.py | 6 +-
examples/LiftSim_baseline/A2C/README.md | 47 ---
examples/LiftSim_baseline/A2C/a2c_config.py | 37 ---
examples/LiftSim_baseline/A2C/actor.py | 137 --------
examples/LiftSim_baseline/A2C/effect.gif | Bin 606895 -> 0 bytes
examples/LiftSim_baseline/A2C/env_wrapper.py | 301 ------------------
examples/LiftSim_baseline/A2C/evaluate.py | 61 ----
examples/LiftSim_baseline/A2C/lift_agent.py | 156 ---------
examples/LiftSim_baseline/A2C/lift_model.py | 75 -----
examples/LiftSim_baseline/A2C/performance.png | Bin 132419 -> 0 bytes
examples/LiftSim_baseline/A2C/train.py | 220 -------------
examples/LiftSim_baseline/A2C/utils.py | 58 ----
examples/LiftSim_baseline/DQN/README.md | 26 --
examples/LiftSim_baseline/DQN/__init__.py | 13 -
examples/LiftSim_baseline/DQN/demo.py | 48 ---
examples/LiftSim_baseline/DQN/rl_10.png | Bin 185916 -> 0 bytes
.../DQN/rl_benchmark/__init__.py | 13 -
.../DQN/rl_benchmark/agent.py | 111 -------
.../DQN/rl_benchmark/dispatcher.py | 105 ------
.../DQN/rl_benchmark/model.py | 35 --
examples/LiftSim_baseline/DQN/wrapper.py | 87 -----
.../LiftSim_baseline/DQN/wrapper_utils.py | 241 --------------
examples/MADDPG/train.py | 18 +-
.../evaluate.py | 2 +-
.../train.py | 34 +-
examples/SAC/train.py | 9 +-
examples/TD3/train.py | 9 +-
examples/offline-Q-learning/parallel_run.py | 2 +-
parl/algorithms/fluid/impala/impala.py | 2 +-
parl/utils/summary.py | 45 +++
.../{tensorboard_test.py => summary_test.py} | 10 +-
40 files changed, 141 insertions(+), 1873 deletions(-)
delete mode 100644 examples/LiftSim_baseline/A2C/README.md
delete mode 100644 examples/LiftSim_baseline/A2C/a2c_config.py
delete mode 100644 examples/LiftSim_baseline/A2C/actor.py
delete mode 100644 examples/LiftSim_baseline/A2C/effect.gif
delete mode 100644 examples/LiftSim_baseline/A2C/env_wrapper.py
delete mode 100644 examples/LiftSim_baseline/A2C/evaluate.py
delete mode 100644 examples/LiftSim_baseline/A2C/lift_agent.py
delete mode 100644 examples/LiftSim_baseline/A2C/lift_model.py
delete mode 100644 examples/LiftSim_baseline/A2C/performance.png
delete mode 100644 examples/LiftSim_baseline/A2C/train.py
delete mode 100644 examples/LiftSim_baseline/A2C/utils.py
delete mode 100644 examples/LiftSim_baseline/DQN/README.md
delete mode 100644 examples/LiftSim_baseline/DQN/__init__.py
delete mode 100644 examples/LiftSim_baseline/DQN/demo.py
delete mode 100644 examples/LiftSim_baseline/DQN/rl_10.png
delete mode 100644 examples/LiftSim_baseline/DQN/rl_benchmark/__init__.py
delete mode 100644 examples/LiftSim_baseline/DQN/rl_benchmark/agent.py
delete mode 100644 examples/LiftSim_baseline/DQN/rl_benchmark/dispatcher.py
delete mode 100644 examples/LiftSim_baseline/DQN/rl_benchmark/model.py
delete mode 100644 examples/LiftSim_baseline/DQN/wrapper.py
delete mode 100644 examples/LiftSim_baseline/DQN/wrapper_utils.py
create mode 100644 parl/utils/summary.py
rename parl/utils/tests/{tensorboard_test.py => summary_test.py} (79%)
diff --git a/benchmark/torch/a2c/train.py b/benchmark/torch/a2c/train.py
index f298536..9a49802 100644
--- a/benchmark/torch/a2c/train.py
+++ b/benchmark/torch/a2c/train.py
@@ -27,7 +27,7 @@ from parl.env.atari_wrappers import wrap_deepmind
from parl.utils.window_stat import WindowStat
from parl.utils.time_stat import TimeStat
from parl.utils import machine_info
-from parl.utils import logger, get_gpu_count, tensorboard
+from parl.utils import logger, get_gpu_count, summary
from parl.algorithms import A2C
from atari_model import ActorCritic
@@ -205,19 +205,19 @@ class Learner(object):
}
if metric['mean_episode_rewards'] is not None:
- tensorboard.add_scalar('train/mean_reward',
- metric['mean_episode_rewards'],
- self.sample_total_steps)
- tensorboard.add_scalar('train/total_loss', metric['total_loss'],
- self.sample_total_steps)
- tensorboard.add_scalar('train/pi_loss', metric['pi_loss'],
- self.sample_total_steps)
- tensorboard.add_scalar('train/vf_loss', metric['vf_loss'],
- self.sample_total_steps)
- tensorboard.add_scalar('train/entropy', metric['entropy'],
- self.sample_total_steps)
- tensorboard.add_scalar('train/learn_rate', metric['lr'],
- self.sample_total_steps)
+ summary.add_scalar('train/mean_reward',
+ metric['mean_episode_rewards'],
+ self.sample_total_steps)
+ summary.add_scalar('train/total_loss', metric['total_loss'],
+ self.sample_total_steps)
+ summary.add_scalar('train/pi_loss', metric['pi_loss'],
+ self.sample_total_steps)
+ summary.add_scalar('train/vf_loss', metric['vf_loss'],
+ self.sample_total_steps)
+ summary.add_scalar('train/entropy', metric['entropy'],
+ self.sample_total_steps)
+ summary.add_scalar('train/learn_rate', metric['lr'],
+ self.sample_total_steps)
logger.info(metric)
diff --git a/benchmark/torch/dqn/train.py b/benchmark/torch/dqn/train.py
index 26d24d8..ba64b95 100644
--- a/benchmark/torch/dqn/train.py
+++ b/benchmark/torch/dqn/train.py
@@ -22,7 +22,7 @@ import parl
import numpy as np
from tqdm import tqdm
-from parl.utils import tensorboard, logger
+from parl.utils import summary, logger
from parl.algorithms import DQN, DDQN
from agent import AtariAgent
@@ -152,18 +152,17 @@ def main():
for _ in range(3):
eval_rewards.append(run_evaluate_episode(test_env, agent))
- tensorboard.add_scalar('dqn/eval', np.mean(eval_rewards),
- total_steps)
- tensorboard.add_scalar('dqn/score', total_reward, total_steps)
- tensorboard.add_scalar('dqn/loss', loss, total_steps)
- tensorboard.add_scalar('dqn/exploration', agent.exploration,
- total_steps)
- tensorboard.add_scalar('dqn/Q value',
- evaluate_fixed_Q(agent, fixed_obs),
- total_steps)
- tensorboard.add_scalar('dqn/grad_norm',
- get_grad_norm(agent.alg.model),
- total_steps)
+ summary.add_scalar('dqn/eval', np.mean(eval_rewards),
+ total_steps)
+ summary.add_scalar('dqn/score', total_reward, total_steps)
+ summary.add_scalar('dqn/loss', loss, total_steps)
+ summary.add_scalar('dqn/exploration', agent.exploration,
+ total_steps)
+ summary.add_scalar('dqn/Q value',
+ evaluate_fixed_Q(agent, fixed_obs),
+ total_steps)
+ summary.add_scalar('dqn/grad_norm',
+ get_grad_norm(agent.alg.model), total_steps)
if __name__ == '__main__':
diff --git a/benchmark/torch/td3/train.py b/benchmark/torch/td3/train.py
index c844d8c..48bd1f7 100644
--- a/benchmark/torch/td3/train.py
+++ b/benchmark/torch/td3/train.py
@@ -15,7 +15,7 @@
import gym
import argparse
import numpy as np
-from parl.utils import logger, tensorboard, ReplayMemory
+from parl.utils import logger, summary, ReplayMemory
from mujoco_model import MujocoModel
from mujoco_agent import MujocoAgent
@@ -103,8 +103,7 @@ def main():
train_reward, steps = run_train_episode(env, agent, rpm)
total_steps += steps
logger.info('Steps: {} Reward: {}'.format(total_steps, train_reward))
- tensorboard.add_scalar('train/episode_reward', train_reward,
- total_steps)
+ summary.add_scalar('train/episode_reward', train_reward, total_steps)
if total_steps // args.test_every_steps >= test_flag:
while total_steps // args.test_every_steps >= test_flag:
@@ -112,8 +111,8 @@ def main():
evaluate_reward = run_evaluate_episode(env, agent)
logger.info('Steps {}, Evaluate reward: {}'.format(
total_steps, evaluate_reward))
- tensorboard.add_scalar('eval/episode_reward', evaluate_reward,
- total_steps)
+ summary.add_scalar('eval/episode_reward', evaluate_reward,
+ total_steps)
if __name__ == '__main__':
diff --git a/docs/tutorial/tensorboard.rst b/docs/tutorial/tensorboard.rst
index 35869af..8952a5e 100644
--- a/docs/tutorial/tensorboard.rst
+++ b/docs/tutorial/tensorboard.rst
@@ -1,14 +1,14 @@
-Tensorboard
+summary
===============
-Visualize the results with tensorboard.
+Visualize the results with tensorboard.
add_scalar
-------------
Common used arguments:
-* tensorboard.add_scalar(tag, scalar_value, global_step=None)
+* summary.add_scalar(tag, scalar_value, global_step=None)
* tag *(string)* – Data identifier
* scalar_value *(float or string/blobname)* – Value to save
* global_step *(int)* – Global step value to record
@@ -17,11 +17,11 @@ Example:
.. code-block:: python
- from parl.utils import tensorboard
+ from parl.utils import summary
x = range(100)
for i in x:
- tensorboard.add_scalar('y=2x', i * 2, i)
+ summary.add_scalar('y=2x', i * 2, i)
Expected result:
@@ -33,7 +33,7 @@ add_histogram
Common used arguments:
-* tensorboard.add_scalar(tag, scalar_value, global_step=None)
+* summary.add_scalar(tag, scalar_value, global_step=None)
* tag *(string)* – Data identifier
* values *(torch.Tensor, numpy.array, or string/blobname)* – Values to build histogram
* global_step *(int)* – Global step value to record
@@ -42,12 +42,12 @@ Example:
.. code-block:: python
- from parl.utils import tensorboard
+ from parl.utils import summary
import numpy as np
for i in range(10):
x = np.random.random(1000)
- tensorboard.add_histogram('distribution centers', x + i, i)
+ summary.add_histogram('distribution centers', x + i, i)
Expected result:
diff --git a/examples/A2C/train.py b/examples/A2C/train.py
index 1daba47..4050a41 100755
--- a/examples/A2C/train.py
+++ b/examples/A2C/train.py
@@ -25,7 +25,7 @@ from atari_agent import AtariAgent
from collections import defaultdict
from parl.env.atari_wrappers import wrap_deepmind
-from parl.utils import logger, get_gpu_count, tensorboard
+from parl.utils import logger, get_gpu_count, summary
from parl.utils.scheduler import PiecewiseScheduler
from parl.utils.time_stat import TimeStat
from parl.utils.window_stat import WindowStat
@@ -186,7 +186,7 @@ class Learner(object):
min_episode_steps = np.min(np.array(episode_steps).flatten())
metric = {
- 'Sample steps': self.sample_total_steps,
+ 'sample_steps': self.sample_total_steps,
'max_episode_rewards': max_episode_rewards,
'mean_episode_rewards': mean_episode_rewards,
'min_episode_rewards': min_episode_rewards,
@@ -205,7 +205,7 @@ class Learner(object):
for key, value in metric.items():
if value is not None:
- tensorboard.add_scalar(key, value, self.sample_total_steps)
+ summary.add_scalar(key, value, self.sample_total_steps)
logger.info(metric)
diff --git a/examples/DQN_variant/train.py b/examples/DQN_variant/train.py
index 0bd2279..5ca16df 100644
--- a/examples/DQN_variant/train.py
+++ b/examples/DQN_variant/train.py
@@ -22,7 +22,7 @@ from atari_agent import AtariAgent
from atari_model import AtariModel
from datetime import datetime
from replay_memory import ReplayMemory, Experience
-from parl.utils import tensorboard, logger
+from parl.utils import summary, logger
from tqdm import tqdm
from utils import get_player
@@ -120,11 +120,9 @@ def main():
total_reward, steps, loss = run_train_episode(env, agent, rpm)
total_steps += steps
pbar.set_description('[train]exploration:{}'.format(agent.exploration))
- tensorboard.add_scalar('dqn/score', total_reward, total_steps)
- tensorboard.add_scalar('dqn/loss', loss,
- total_steps) # mean of total loss
- tensorboard.add_scalar('dqn/exploration', agent.exploration,
- total_steps)
+ summary.add_scalar('dqn/score', total_reward, total_steps)
+ summary.add_scalar('dqn/loss', loss, total_steps) # mean of total loss
+ summary.add_scalar('dqn/exploration', agent.exploration, total_steps)
pbar.update(steps)
if total_steps // args.test_every_steps >= test_flag:
@@ -139,7 +137,7 @@ def main():
"eval_agent done, (steps, eval_reward): ({}, {})".format(
total_steps, np.mean(eval_rewards)))
eval_test = np.mean(eval_rewards)
- tensorboard.add_scalar('dqn/eval', eval_test, total_steps)
+ summary.add_scalar('dqn/eval', eval_test, total_steps)
pbar.close()
diff --git a/examples/ES/README.md b/examples/ES/README.md
index 207ae2d..d868202 100644
--- a/examples/ES/README.md
+++ b/examples/ES/README.md
@@ -34,7 +34,7 @@ Then we can start the distributed training by running:
python train.py
```
-Training result will be saved in `train_log` with training curve that can be visualized in tensorboard data.
+Training result will be saved in `train_log` with training curve.
### Reference
+ [Ray](https://github.com/ray-project/ray)
diff --git a/examples/ES/train.py b/examples/ES/train.py
index be2c7d7..eadf26e 100644
--- a/examples/ES/train.py
+++ b/examples/ES/train.py
@@ -23,7 +23,7 @@ from obs_filter import MeanStdFilter
from mujoco_agent import MujocoAgent
from mujoco_model import MujocoModel
from noise import SharedNoiseTable
-from parl.utils import logger, tensorboard
+from parl.utils import logger, summary
from parl.utils.window_stat import WindowStat
from six.moves import queue
from actor import Actor
@@ -202,7 +202,7 @@ class Learner(object):
logger.info(metrics)
for k, v in metrics.items():
if v is not None:
- tensorboard.add_scalar(k, v, self.sample_total_steps)
+ summary.add_scalar(k, v, self.sample_total_steps)
if __name__ == '__main__':
diff --git a/examples/GA3C/train.py b/examples/GA3C/train.py
index edc7f33..30f3a41 100755
--- a/examples/GA3C/train.py
+++ b/examples/GA3C/train.py
@@ -24,7 +24,7 @@ from atari_model import AtariModel
from atari_agent import AtariAgent
from collections import defaultdict
from parl.env.atari_wrappers import wrap_deepmind
-from parl.utils import logger, get_gpu_count, tensorboard
+from parl.utils import logger, get_gpu_count, summary
from parl.utils.scheduler import PiecewiseScheduler
from parl.utils.time_stat import TimeStat
from parl.utils.window_stat import WindowStat
@@ -313,7 +313,7 @@ class Learner(object):
for key, value in metric.items():
if value is not None:
- tensorboard.add_scalar(key, value, self.sample_total_steps)
+ summary.add_scalar(key, value, self.sample_total_steps)
logger.info(metric)
diff --git a/examples/IMPALA/train.py b/examples/IMPALA/train.py
index cf9e55c..9f2a3e6 100755
--- a/examples/IMPALA/train.py
+++ b/examples/IMPALA/train.py
@@ -22,7 +22,7 @@ import parl
from atari_model import AtariModel
from atari_agent import AtariAgent
from parl.env.atari_wrappers import wrap_deepmind
-from parl.utils import logger, tensorboard, get_gpu_count
+from parl.utils import logger, summary, get_gpu_count
from parl.utils.scheduler import PiecewiseScheduler
from parl.utils.time_stat import TimeStat
from parl.utils.window_stat import WindowStat
@@ -221,7 +221,7 @@ class Learner(object):
min_episode_steps = np.min(np.array(episode_steps).flatten())
metric = {
- 'Sample steps': self.sample_total_steps,
+ 'sample_steps': self.sample_total_steps,
'max_episode_rewards': max_episode_rewards,
'mean_episode_rewards': mean_episode_rewards,
'min_episode_rewards': min_episode_rewards,
@@ -244,7 +244,7 @@ class Learner(object):
for key, value in metric.items():
if value is not None:
- tensorboard.add_scalar(key, value, self.sample_total_steps)
+ summary.add_scalar(key, value, self.sample_total_steps)
logger.info(metric)
diff --git a/examples/LiftSim_baseline/A2C/README.md b/examples/LiftSim_baseline/A2C/README.md
deleted file mode 100644
index 235e1af..0000000
--- a/examples/LiftSim_baseline/A2C/README.md
+++ /dev/null
@@ -1,47 +0,0 @@
-# LiftSim基线
-
-## 简介
-
-基于PARL库实现A2C算法，应用于[RLSchool][rlschool]库中的电梯调度模拟环境[LiftSim][liftsim]。
-
-## 依赖库
-
-+ [paddlepaddle>=1.6.1](https://github.com/PaddlePaddle/Paddle)
-+ [parl>=1.2.3](https://github.com/PaddlePaddle/PARL)
-+ [rlschool>=0.1.1][rlschool]
-
-
-## 分布式训练
-
-首先，启动一个具有5个CPU资源的本地集群：
-
-```bash
-xparl start --port 8010 --cpu_num 5
-```
-
-> 注意，如果你已经启动了一个集群，则不需要重复运行上面命令。关于PARL集群更多信息，可以参考[文档](https://parl.readthedocs.io/en/latest/parallel_training/setup.html)。
-
-然后我们就可以通过运行下面命令进行分布式训练：
-
-```bash
-python train.py
-```
-
-
-## 评估
-可以通过下面命令来评估保存的模型
-```bash
-python evaluate.py --model_path saved_models/[FILENAME]
-```
-
-tensorboard和log文件会保存在`./train_log/train/`；可以通过运行命令`tensorboard --logdir .`查看tensorboard可视化界面。
-
-## 收敛指标
-训练30h左右，评估指标能达到-120分左右（LiftSim环境运行1天reward）
-
-
-## 可视化效果
-
-
-[rlschool]: https://github.com/PaddlePaddle/RLSchool
-[liftsim]: https://github.com/PaddlePaddle/RLSchool/tree/master/rlschool/liftsim
diff --git a/examples/LiftSim_baseline/A2C/a2c_config.py b/examples/LiftSim_baseline/A2C/a2c_config.py
deleted file mode 100644
index e3e776b..0000000
--- a/examples/LiftSim_baseline/A2C/a2c_config.py
+++ /dev/null
@@ -1,37 +0,0 @@
-# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-# http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-config = {
- #========== remote config ==========
- 'master_address': 'localhost:8010',
-
- #========== actor config ==========
- 'actor_num': 5,
- 'env_num': 5,
- 'sample_batch_steps': 5,
-
- #========== learner config ==========
- 'max_sample_steps': int(1e10),
- 'gamma': 0.998,
- 'lambda': 1.0, # GAE
-
- # start learning rate
- 'start_lr': 1.0e-4,
-
- # coefficient of policy entropy adjustment schedule: (train_step, coefficient)
- 'entropy_coeff_scheduler': [(0, -2.0e-4)],
- 'vf_loss_coeff': 0.5,
- 'get_remote_metrics_interval': 100,
- 'log_metrics_interval_s': 60,
-}
diff --git a/examples/LiftSim_baseline/A2C/actor.py b/examples/LiftSim_baseline/A2C/actor.py
deleted file mode 100644
index 4286e7f..0000000
--- a/examples/LiftSim_baseline/A2C/actor.py
+++ /dev/null
@@ -1,137 +0,0 @@
-# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-# http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import numpy as np
-import parl
-from collections import defaultdict
-from env_wrapper import ObsProcessWrapper, ActionProcessWrapper, RewardWrapper, MetricsWrapper
-from parl.utils.rl_utils import calc_gae
-from parl.env.vector_env import VectorEnv
-from rlschool import LiftSim
-from copy import deepcopy
-from lift_model import LiftModel
-from lift_agent import LiftAgent
-
-
-@parl.remote_class
-class Actor(object):
- def __init__(self, config):
- self.config = config
- self.env_num = config['env_num']
-
- self.envs = []
- for _ in range(self.env_num):
- env = LiftSim()
- env = RewardWrapper(env)
- env = ActionProcessWrapper(env)
- env = ObsProcessWrapper(env)
- env = MetricsWrapper(env)
- self.envs.append(env)
- self.vector_env = VectorEnv(self.envs)
-
- # number of elevators
- self.ele_num = self.envs[0].mansion_attr.ElevatorNumber
-
- act_dim = self.envs[0].act_dim
- self.obs_dim = self.envs[0].obs_dim
- self.config['obs_dim'] = self.obs_dim
-
- # nested list of shape (env_num, ele_num, obs_dim)
- self.obs_batch = self.vector_env.reset()
- # (env_num * ele_num, obs_dim)
- self.obs_batch = np.array(self.obs_batch).reshape(
- [self.env_num * self.ele_num, self.obs_dim])
-
- model = LiftModel(act_dim)
- algorithm = parl.algorithms.A3C(
- model, vf_loss_coeff=config['vf_loss_coeff'])
- self.agent = LiftAgent(algorithm, config)
-
- def sample(self):
- sample_data = defaultdict(list)
-
- env_sample_data = {}
- # treat each elevator in Liftsim as an independent env
- for env_id in range(self.env_num * self.ele_num):
- env_sample_data[env_id] = defaultdict(list)
-
- for i in range(self.config['sample_batch_steps']):
- actions_batch, values_batch = self.agent.sample(self.obs_batch)
-
- vector_actions = np.array_split(actions_batch, self.env_num)
- assert len(vector_actions[-1]) == self.ele_num
- next_obs_batch, reward_batch, done_batch, info_batch = \
- self.vector_env.step(vector_actions)
-
- # (env_num, ele_num, obs_dim) -> (env_num * ele_num, obs_dim)
- next_obs_batch = np.array(next_obs_batch).reshape(
- [self.env_num * self.ele_num, self.obs_dim])
- # repeat reward and done to ele_num times
- # (env_num) -> (env_num, ele_num) -> (env_num * ele_num)
- reward_batch = np.repeat(reward_batch, self.ele_num)
- done_batch = np.repeat(done_batch, self.ele_num)
-
- for env_id in range(self.env_num * self.ele_num):
- env_sample_data[env_id]['obs'].append(self.obs_batch[env_id])
- env_sample_data[env_id]['actions'].append(
- actions_batch[env_id])
- env_sample_data[env_id]['rewards'].append(reward_batch[env_id])
- env_sample_data[env_id]['dones'].append(done_batch[env_id])
- env_sample_data[env_id]['values'].append(values_batch[env_id])
-
- # Calculate advantages when the episode is done or reaches max sample steps.
- if done_batch[env_id] or i + 1 == self.config[
- 'sample_batch_steps']: # reach max sample steps
- next_value = 0
- if not done_batch[env_id]:
- next_obs = np.expand_dims(next_obs_batch[env_id], 0)
- next_value = self.agent.value(next_obs)
-
- values = env_sample_data[env_id]['values']
- rewards = env_sample_data[env_id]['rewards']
- advantages = calc_gae(rewards, values, next_value,
- self.config['gamma'],
- self.config['lambda'])
- target_values = advantages + values
-
- sample_data['obs'].extend(env_sample_data[env_id]['obs'])
- sample_data['actions'].extend(
- env_sample_data[env_id]['actions'])
- sample_data['advantages'].extend(advantages)
- sample_data['target_values'].extend(target_values)
-
- env_sample_data[env_id] = defaultdict(list)
-
- self.obs_batch = deepcopy(next_obs_batch)
-
- # size of sample_data[key]: env_num * ele_num * sample_batch_steps
- for key in sample_data:
- sample_data[key] = np.stack(sample_data[key])
-
- return sample_data
-
- def get_metrics(self):
- metrics = defaultdict(list)
- for metrics_env in self.envs:
- assert isinstance(
- metrics_env,
- MetricsWrapper), "Put the MetricsWrapper in the last wrapper"
- for env_reward_1h, env_reward_24h in metrics_env.next_episode_results(
- ):
- metrics['env_reward_1h'].append(env_reward_1h)
- metrics['env_reward_24h'].append(env_reward_24h)
- return metrics
-
- def set_weights(self, params):
- self.agent.set_weights(params)
diff --git a/examples/LiftSim_baseline/A2C/effect.gif b/examples/LiftSim_baseline/A2C/effect.gif
deleted file mode 100644
index b796acbf3c872f0e960e731fb99848c774446165..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001
literal 606895
zcmbSyV{;|U^LEaOZQHi3&BnHE+s4M(*tTsa8*5{2Y}+{h`}cf=r)#QTOm%h7)Kqo#
zTz$2q^iLjM<41rDa2f1>Vub(+kBA71%iw^^Pf13B!2%THkW^Mu`&1TE*NIZnkB~JC
z+ceYaw6U^qiz)LDicc<#NUiS6EPAf5+H9`PYoFimm@VjDF6v#Y>W!N19Vs2$${#jq
z9zN;d;4|s3Hnd|-$DL=6e$Q^
zK|)PcTvSPdjfn*u@ShLTHwZ{5Fh($j|M~mhHNl`jB1lxGsAzE`#`hIJ__CGC+@$>#
zYL~5^t44gfjq1E)J#<*Gif|qX+bhrSe+%Y!4UZtGzFr@Ec?P?L1cgR~YdJ?-TF3sd
zvcif@N=Z#i#U;qh&dSZnv@6Ih^vlmH&nc-YuBa`mE^4f6ZYXUl>!@jJ?d&ZlXwM?e
z96XCSE-uC@
zAvrzC@9wMZy6E$5QzAZpFc<^}C!qCvcgPEZkj+o)92_bJhcQT`Bbux=680l_{1y-<
z5rB_Rm6B?FIGIj_7;?#zxHs*|iM{;~7M*$kl*2o=JB&@6FC?=TZ7w^Bk+D!NpUxKt
z=U$~=sa0)JI{8$!Qm^0cg9z_YvyM)`P^BMiatMb`E*_~**M6zj;Le_@RUPs}ACo%6
zc_w?+nrmSCI!Rujrop5+fqXN!2d7DIDzzDq{F%HV;O%_PN@pW&Zn;?Dou`r8MSZba
zXWBg+Bo-dEN#SJJBEQIV8-YU+OY5F3c(|hVB8~vhS5=+!
zC}159C1&muczgH?ThT?1LOUx?q2R_ui0|{Oetb~ns@~b~ulx#r$Y*=HqE7QCEl^jH
zCFn>Q)O9vi!sf_CEEdy^atH3Rhnp9GLg0vD(sE#GYTf)
zC}HjZnSn*?$65ACw^XpsLbOMvauzm~Wv9KYOIFu|zhEs&Cbg=9D;BKIt19!dV0Gj;
zpGov^FH6N>X?jwFLo$OJfUZ`ByGQi|6rR=j7JQa8FvN8QKp1aGct#BZqs*1?S%0mN
zW~{IC1*d7f*D99|G+R5^^+pMhSRF;g-^=Fs)+#Oita;n|s<_7Jo4!{{>AByRn3>4d
zxmlJ@gD}K3PD2R3DaXIoeHyZ1yK{d-!_k0=GeD(ol+hT8TX~kkN@yvSLeiI@h!Odz
zW!Gks92yfs%}zc9H%m~w4348&1l<)z>Flyf(?p&%!=W##QqIr?wl^qzliB@SM8vMy
zCi@XiXG8hkb?&}uG)2$V4_*|0u};JAssS^>nm%LTzA}n18WMLX>OvjsM(RS(Fb?=f
z(12;$m!))`(;TGnVGd*W1ZNCI@hHu}zT2fcMCIO8o3lMf1zl?}2N^FmS9@=DuvS7Q
z#dQw7fJPwz_KWJ)x76MS^pVnTn#EtlR1mkKg8}U%I)2S5<^Siyy+4t
zx`|8$Cr2Z0VG?|dtBKq~CGczDb+#?oWv?(oEMDUQ>(zR&tyC%D)S`Pe;p3~fStHNq
zCsI!4dw>|rQWx=8`I!*%mu06KoR8=vxn|nuSZX!s2_KUVr1Ye
z7UiuV%;3ljf+8yD4rx{8YfaCtncTl)ydsHbvRP}gJNqt$$TcyHq*XKEx)7bC>MDuy8V$Y3^7n4A+!?JY%L~9wJIsK6h9ESBZ4o;D
z&kv*!BUz90xjY5rioc!KeL7Y0DL$CO=EjGXz_K63a(6Yq!(en)vsSXePWVTivoMbM
z@V(vW>bA}>^!=MwDtqH=gWQWKCg>E3?d(epkNyOAM6cEHwKe`WuQB8HF8)$`uFkWA
zHcXHzQ6xO0Awsm55~f>NNP2E5;=HFQEYHRH!MV6kD;rBP{D
z$kyJ_3JJ1z6nrByqXb7nQ7{y^LrAHARSfc&%JN
zm)9gX$qzou)<$`l>f)a^r{*f1(nFNdHHpSVbnepRs=6(b04aJ=4BR>S<|ba)P@Nn_
zu$IKbl=5kdN$f5`Q>xo`8SS#itl8`2_Qm4%1{p^#Fux%1-pKGf!Zf(T|4@~Bu{=M^
zB>X_(noRo#oj`v`-_alnNqdUVzHwaia%}D>PKlu3c5he<11>2`Yiv!6H4Zgm&v947>G?R^IolS|n!C5;b((vSt!E!HHfIbiqhc+i+nCsU
zX=GRdp%VgGXQv9JPYP4^pAT!mpsbMVn4*|6#tmtBn
z#Xn_m6P(pF^7NdW5j
zF^K+K8eG(VV=P=fSUrA_z(Qa5(ZZyAgbN3#NWcOKb^+BK1br@)w=0SiiNUhux)d5JFc!MBk|Gx5o>(gQ
z)~cfZWCHxnT+s)9*ygV7tK5s!QpA>K#37RDsyxw9ew;Xd$~fV@EsP<-jQpo;*bvqU
zi#{m(KKU*5{@{i!Zn!*ZVP|A&O_Ee*O*HNIVsp1VhE$%P^JbpUW@?(!ej1*<&P=&R
z3{X#!n8_R=KRhTFJVrG_@J+pnNxUW>Jfd2H*M59cgIFUuA
z#EZpfQKSMPydqA2lBs&cN#bj8*mMutx}oqUi4k&|hIs7@nW%AC9kPPZU0MCwy#u(U
zgPy&WUU868eHYo?YErmnm#~azWNA#TTr&KZ@9AeB8QzRVK`B6r6s!PNGTJ+w^buk`
zbG}M+jkINQ2%wRwRJ`#ri8ZR?#31m`A2~MaFmh5^1`Y(0U!k*isu#mXvc{tLQom*a
zgyajV0Bf{e7`Pu1W=GU0Cz>WjML~6(VxLBe^NEb_W;92|Vsl9na1zkNg|H?k3`1f8
zCA&(KckIim0@Oi?Fdac$(#hBsVhmhx)`Q3xtpqEvdLZ&}wtF@4Wz{TFbv{fo18-FWYcddxHaEXIM;sS|0;Xzg
zD3bywPmHn)JcJLOMY`97bka}Uu~dvFHXOtzwF+*vabr|d=?oXqG}cu00kCw|R(6SE
zwI&zQ-_~ep-nbOXWa*itjCXFzsGMGr0eS7n>EplhU<`(TKtx)>@36|LV4#Qt4QyI}
z9Vcdiz*xFivg1lD1D7oW!n3P((YDP|P=99UTIKNEY4lH^3_k)QPR$Sinj(o&eT9(8
zC8=pG>Asm}>C3V+Au_k8qE*M7(@t*N+L)gEfUA3hRS`Pa`wU%9V#yQitM&M%qWIIy
ze3}Ky=eLabRFdxza?g-F-l1F%)?a&P+zmgq(RooZX+uHh=|P;C_T*mFvOwOy%ECHi
z$nz*B4nOeu+^PNidX&_c6=~+er
zSMfrF>q}SW_+HABRc0zv-tH`y!d?ztqN_b`n4#>Oe4L(bBT|H0xM7{`=an^%o0Gk!
zspCQ^{3K=~2_$j~_Ag}DLF4rtG9fming}JT#H$8Lp#xbS#M<9g{n@MA=F6(ic|BK<
zb%q$K-|6aX0vq{irj%>iqN@wV^~HWyS)lP&jS~!X)h@_XPt1o5brpJYR(ZFU{Ql$p
ztds=`El|d*KkHz!Jg?#2hUx~V|yGm0wku}u$3
z=GHFcsHQyTDi11TU)Uq9!*F`Xs__9MlQU`Ve-@EI=N+VjAYrR!j{+>0QB$Z@fp*K^
zxnxB$TLk#4&yu-@k6YYbicEA`+WBgwFIquz)h$}|E$p%-C7dP!7Kv1?sr2;qx94~t9b!S&cSfP*2
z%&>$t*Tc6LNHp8gH*@F}^-Z8oH&Log!PD{e+L(*!isk;Kz#kc4
z!vo;-oGMEOX$pV-3?ltgrCwA>IfA}YqHXQuqT8+$+y7{yr|H8y7DQmFl1G1nDcpjq
zhMHl5km<2mY2l}r)i~-ri2DZnBbb&%iF6DB5%
z-M))meIz4eGX2=d|2veLong55pzVQw9J+m~bb5qIjaP&sLrEE1|1f=6lxlsctieuC
zjF!0Rz9B|6op^uNsb|i^5BK(Rbf4fa!zOwTx08iZf_uOtbsMLLgXr=gPZ@!H9ui7
zA3^mSZ>x|{K2e9w%Vl52QfFZ(xRAnaUK4Zv9Go#E2@UpZX^}TJ*Yj^t%~Wx0y8IP8
zMU_}J9HGA`72nfxZfG2+c(Z;ZW485PLa(gcjBo}0aK-3Kq`11uXbZDwDNm1(*XLoy
zBF@jJm&yMsS3B2HdUe@ch>H@1zY(s7Q!`~GBze$PA@&~RW1DH
zddTL+Klp#l%68aYu*FxKb#c{b;nni8oB95msrp+P1Q?p1B}FF7|GwhoXKK!SoD+S@
zJQp09uPF7K2Y`V;_n=BTlx%OvvjszRt
za+nmupd}#Zzxh8?}Fx2>ZXGsy&jp6slFJ
z@;WLfn>*0RWUIITW^%I?LjRfl6QOw*X^UV
zeDH7U7SI1k6=4Pn2fxuG%D1|+Xn*EFhB@qn@1EN{3EGoCxKRz?JCEv*qEx@#%$$U0hz?YnW`z%1OfaS+s*sMqeooIL^b{aUIfnGu65|
z=5(6i^hz8pX#EzoUHMNs7u-HM#G(a39xi8>ea`nG!gRr=AOhlQG0GAtbtQ-CrxMPl
zt6(=1To_WpH5j8khkZ$d;4vM}IQov(_GMeZdTdy#BYh|}7ab7sBndLhXk;<*r}f-Q
zdOJc!?QcX~xGDejn_%2Ij+M%D0-T5S5jW=>f4j-|>a{1N(a+iWd+)KE>+@Ri5yJ1G
zO9yw-ykpN3sq{U4(w(8>`O7q(`|(aGv7r0q%1ny>@wS4i{PFw0|L$OWk9X~2Z~Go-
z%C-Kh3|jSYMUgc=s>v(Jn&7yzA&_W(qCNYGVmUOpve@EQuDco!|^@mqrL0%
zyO6-FSc1L-_wgHFmTbETp_-^1s|Sm=uH(%k8Cc}>`#Scxwaj4o~p*tVHK_7j@p96spDqD4G
z`foXo9}|HelL;SFb4V+&uUCEVRfb=cfnOE*UrPyJ5OG8bc$1QMpR>Z>Glt*Of!|d{
zA16EQq{W(xGG6Zh?ZypXf*3rYH=4lGX%nC?$
zP^|`|c5PebBU!)xt(Klb(%i9ARp&t0|prp932;%
za4>_co2)d8jt%=Fyi?DE3O{QA=BR%gt@Jf1QhUSVNIsyWKV`PJpk^_{JQy_2Jh
zvzx1jyO*brH_#U(1do!DXbALI=5Ps%D$g%Xf$*dZN_;L`Ojd)X4o32^I4r>k2;L_0
zi3CdJJT=-T6V!CcrIL^5zcS@faKyA?Db4c=`Fz382$RXDiiMIfNSs+1XDa0~pr^-f
zj@ILjNbc4KJ!=(P%{oKX|6#osTFvJ3(ekobI^0g|_zvek9N@*eq3*Vqv(s-3`uzbQ
zOoEKt-B1KnZZ&}yIt?biylJ-iR*Gvx*m@li+)>Z9C+T=Fw1PpZrH
zy7eA}_|JBmP0p*cR}yq4vSp+ew4M*H&U0hD_=@p84=yJ&h4lTWT+!UST#W||JzFfE
zH(P^&*u;6CUJoAr%z6aaA5JxMrDK*%#NU1&Pj{02JKB#^Tx_-ICoNGP-SX|=*0O&^
zAmQy$1|dn^eh)&!7M*hgLZ{+8zD^gJ`;kNsmHr@0KK}ioRY8w9{M@Dl3Ibpo`zep&
zSm(@-;yL#yj}dtPn;#t&9xL
zESR9%JLgk8zPHNz55IQq(>}?0@>wN~)h=l_NRnXcJS~n=@Hs5KfcG-ro*kSx0-%%Zu4E#PhDNL!}EJ<4@aeN@!6MTTIgQ2tv*i*bNA3{)y0|~d!O_^U1#6)
ze<2Y#dHWgEQ+0)Lja`i#L~P!*lUjb>rQ{OT9x;^(@ubx8o@}@_@I!yO)=MK3q6IL-
zzFZd&+6uU=QU?0FtXa1^I`Etydr17D#nZdp@kb_l-3w<1edi4Y3iYu1m^%tRpV)c|
z`JC2Q61-HfTFzZoN!;{3&dVbcf8OxkjkFyhOqiG7iGv4~c*^z)tmL1_hi79Rz=CsP*GET9K2G9@(zHAc{T2M)8&1Rb<
zikebe{&!&sdL_*Gu+m0Isu`H#g!e1$JuDk8$*p!&?B>=4ig`#W8mOgos+sb34yxIs
z!^J%*uv3=t>{&-{WxT5)vR(qLnH99h%saSpUbCusPkM0_>XH-a7Ld7z0w=^7GiYxf{*PHAU$kCz)KM*YrnPcbwrVSPD?$~f
zT<>d9`dcOm-Pfk|=dTuQ#XR$MNL+C%JnL*ox`M+@SdFoSz{yCk^?Hi5Cf!OqGibW?
zKUs3kDXN*4u9>|d5|30`T?^Lo**Z&g8m+bFSGM~2x@)r2u40LUMUB&?gtnL}rRS6gtaK2<5(a}*QWPy5v@g`j;kWv~jK%E99qX^#>Yy|h~>
zH+wIYDwHbC9G{FiW&O=|Xf1mJnOcGEr6m2+eXJ#pyO0vkS`8YvDYakg(rn&X9*Jn>
z5cIcbipW};;d*M$y_ylEuAlqC$<>%3XlI`wu>w*2r20^HVQ`Z5uN$kqBe0iWHWavd
zwe8VXpF3){>9y_{%hht{@96)4u*@h=(gV1@_8RNn>{>Epp3%Q{rq0=eKXLE!?sUND
z)7=G>zmCXrIogD5Y`;Cf_KAKw4}Q8=fa+c+SRL*%vGiJY6kIzn1FkdXdiK~9DgL-t
z*`y%r9SKr1i0Y0yhehO^k}8uKuB
zHh3esHwx$e6>;=l4i%IE71tA<*;{_lEw0KK`QjQYw|}e+$B$O-U|d0Tz3;RmdDS8l
zRgbCp{8^k;Z#v1n^IUCPudGwI=-;|ZE$ecX7?wJRF5t8U?HS>PL;;9DEu
zJ8$88Sr7)D5L_uouVY5HWpzbnbZLfw2RFdITf~EUr~AsMj4^QDTkt2PZMw179}O#F
zlZcRsz#}IV)v9m5xukn1RBROF9V&=$rYk5VFj{A{z|&hh#r=|TR9+K!M#r&BV
zGofrFA)yK`X%Zn}1SX;q>eU??>OGODGI1&xlMSmQ7#jt+3dIwE06kHJ4UHTf?Fb7E
zh|fm0gGxoLLY5vv#;J_K!A7mFLM;qD6-XlfS!c^Zx@@7r_BG
zb~}s4a>mNq>B3%H4`NeBgJtJ{WG6vHr&ms5e?jBytzy-`XQgarqDAL2bPYKdegCO=
z!okk1s!A=0&dw^u5$wt_SkG~HKZgh3Oa#x%J%z(W&ztMY>siQKy4YBT&U;nfoRZAf
zxX4@2&flxr*w@ULuiB(z!avf?+wCgQ&n`Hp%0Ij)=(uP08(}gw;VVyt(2`x4T2`|@t;!9pcFrMt0C!Hl%)o=2V);Xv2YieLoskw+jgwLBLyX+~ygB3^hTaUs
z0www=F2V6b+Uhunyp
z+{6+rD#^&-rICdtgyk0b^+$O_28C5Mg{>5YJvW8B7KMjLg{Kz9w@1Y*4y6|k1!@Ex
ztD3tskP^)mN
z``a}dcQ~h(xw{sx%K^qCx9gJ*tm>6ULjA6t=)XbBj_{R#6=zJ3+UCka;z1aCdk7lI
zT5km$f0n0ATn*k93E2z4pp3w#0hlbi7(^;`IRi1rci#oAdROrn33_(D;0w4rf-F2|1!S$cUfQsY1VP`ty)HcOh
zWitYY3Nlr00|J9;VH&mdZOn?*L1Me$3S3~o0f3u_xj31AriOm@vvn>PZiWY2$pZ{W
z%Jwxo1FeT9`z|=cE;!48?VyI;@U!idhh2Z0?M#~8U|RfuhN&uvgDi-|L3I}bsitky;-7SMb41g(dyWhFHLkGF*x4Ro_8Zu&Ku#tdM7y*R}T-?)L6pb)^
zR@kJijJW?gB5J@uh!`hof~R=s*?Rpnt`=EjWinx6scfku>Js7Yc*hC=dR$Y{h
zY$j$kgjLJiuyhZi-5_7wIE^6q2K+e(yz^hY&f39{zyR0n0VkS<@)JN2BTx6$kN5?3
z+7z$g6^!%r0K{E?Nh1g)&A`NVurf~{v@Bw%D>WlXscLcpxKz@24HlrPY$02#%l6^7AKLbmQBslTK4Ryfnmsg~z
zS4ergPd&G9sbwhR3wT6=pTDZ^e7cEpM!+g}aC=8gxK`xDaCX
z>YYg3nMk^xh`tUXwF@ylSjCnZI)`KP@Aog*=U?3H5DL2hxwjN$-c({RfLdp&;yOeL
z&SODFf@T{nD#vM+dThN{`dk47&Mp9J7ZAXk5rvx(2g*#1;mwT8%wYA7?|X)r&TwER
ziPKyLy|=}oFF=ry04WNxd*8DA-v9&NIfHKyELs6bwxKUP4y75a0A5kU6;4F&ye$OC
zj?A11Z@?{XzA{PjuL)q0x2I|W;9V;kgy{vT0S?*X#hIMa?wvoVokNP3W4s=P`zvLN
z$KeGBhzADb5{ZVrPMfRqYfVd+->8+@&7Jp915
zYSk{K)+y500n+)zu4IN17L?HrBxqjeBG&myXXIHvCtmVaP!^PicI8V)JGkON$`lmS
z?|L3IX03L-d-#+q?iT9fRp@sWkycd1Z2*!$K6fN(j@XYSS9stmSyiPX%EGhMZ9XJs6OH1d7m1%nZ|u3PJ=e^f%+yYG$>Q{>O`Xq
zY<=A9`j#Hy7x|{IP7e|w>9X1S79!(l0seJY`V5p>+Gt}n|MA?#&;wU=n^0|;$ozyL
z)Eg!Jgv#xi%>Os(Ptnt72GiLy)%)2I@Hq_Gqg3!XHS;-Ly#zDmHnZL{gHJh1?KAp<
z3yCr?bM-lk9ggs%hw$b%_ti58$WyKlazU%mvv+QWSF
zV)lr!!vbe-Ub%n1X>QKh=BB^(*aY8(tLhSL?ouG((iGXmRl_>9#`*~1km>sxWK$OZ
z76-i6)=Scs@7C6ye}U3g@qXus_Sfc||DGw@)<^d)?2>g%@IEDx%ORoWQ?IDFY~AUX
zC)C;^>f%x(;jZb`u7sB!YWxb}_CNi+P?Bv-*??naA_$3hdPyx4;eZ2Ap`ZcI7VPX}
z>AvGc#NL@*>Ot#e!@RxpzN2o*q-ep_@a@C$?cH+Byx?se?|=*6fY@^TbSHg+sDR5~
zc|Q;kVbxN_T@CiK`>wP>+gIHR{l(jD)p?hd0a%VgM}YaY?!KGO0LuU(M*f$bi@wVn
zp@Y~w#DzYGno5$~yt5sl`{AG0((jgt{kO>Z!uQ|EW2wI1+c%2kP=XynAENoq3=xIGL9?EUhx=UKJ$tQ$6B$H?gqrt9ZTzP&o3r@6%2Ajor{e
z1N4fPqC98uOlkOTO3cM@Jj#AYQEc>*D*QrP{8E4NQfcuT-v1c|x;U!4^Kbp=41A-m
zew#=AXwH5=Bfh@aL9GkyN1oK)%K!Qzh7rR>l9b}W`ZZCEHZkZFv0#}3*J@UW%itN)
z=r^3!IARk~G
z)Hg|Dv?p`t@@ETo35k#P$x6ygtE*}&>nj><8k(K#+F00LxjuhB{@#ar03ySG;GyR$
zFlr)-@@kG4Tg&T28>@)gTDl5*N`^Z6Cj1JFNl(tN&TcMmFCMP%o*V0*T3y_^*nBv7
zIRO86;%GDw2!uxq`w6p<7P0)tZz6>jSHcV{##kV&aep)hPuKqVPiVVwR|>DAn+Xl&aL8a(T{l+v%}v>rI}`AG6c={C6iT@48yls_?6gKGo^}7a+-^2zVm#}&`}{!fhzMQ{yF+2f
z_zJOJ|HY3b=8s44ZaSPyVb-0E&AbId`6a8wP?8oe%}Rz-krvK`dYwgubwj
z=LaM2OB#ov5^0f!!O5(VgkZTul7{0Nj2lN#L|E-dk}zcy{lXi{Fp6Qn_9TlTTm$=rQ5D@3eOgu5Hd#?qKZ{3S
zD{BO+NVte&J)^23h+C88tV7GtIEuw@*EB{KQ`0;nHdWIyrSfsv;&#nyL(#hcv7D#1
z1Xta$m*iVrdlGN!&~=(t9h-iFu&z0T*
zt{Z}7k8K#f6zaKE>)ZxU)+yr~k!o3Xuwxus^&nsyClAzX8l#S}bD5;k%&ApjcD<->
z&yYfBoZ&q9cRRy(LcltQmG~y#Q#$%t-!F-u%dsF!zs0c_K!f2V=U|>g&Bgzb!?~jS
zGW|5C;nFS7{m(Ttc5&(+>x;h1IPi*dEf`#=
zV=sh!`%jWF+#5_HgzA0EVzR6O?@_vL9xs^+vVirV2C;$HepoV*&slM{K}>d67OFK?
zuA&O}cGF^j=at8SQ1?~$NkI4YFDInV>HcVg&$H3;_^$hLX2|zjto^$edp7%
zX<+a3hm8FBfWr)f`ojdf;rHE1aza22Mt?K?ve7G%@J;_BbKuRxrlIin$5`**?M;(Q
z*F;%jz?h}T113r^G=l73gN3eUTx(ynX(C8!t3f!ETZAoe64il|b~Xl<;B!?HSV=1}
z)Fx!zg;%1_*Yf_4!P{`QX<|futzn$)*}#$wQs^DB2()OHNKyhalqjnqs3s;Ki7pa!
z_R~L9Vs}yhKJv*UE$Av>8Y6WC3Q?QKN1SWx%;{ru5vL->*=JGWxaAITH^+Z^A(DCB
zeC%OgoR08&uqNp0EkA6$h>%IJ{#L=DAceLb=e4`_cZJv^da{y~+;2<-CB;hMe07L1
z*03hq{42rdT$fhpn@@(GH>U9F7>^KeNlKKXq}M!?@f3uMuL~6;we=p?-C#?{NhyA$
zyOU8DW6KyII7Zz|sVNLcODWPTW6OSJ`Wy|Rx
zDCe2-uK!Ua@3oklY|>JM&Q2w|02b#L;+9W(MJ@eiy^u%;k%*^T2~}w|
zpKkG3gqKTO^Y4x(!NrI@YrRszcXBcRoufj8uu3ULW~msDvl8JHgrJ&!zEtM%vr_46
z#wv(Br`m+GT4#$+v-^CxK7_MIt(8u@(no#&H+8LzKE3Yd`AYi`XPrwfy?$zPUbhBJ
zz0VB2;mi4IAMm+81d+iQ+IDS-?71OYp1~CRVr^9HxiKk@!JOK5ebVH)DQkwo68eH1
zX9QZRNMcS}!e%3>h@-Wfkc7f6d9=J-%G$(n(J}nk(M1Ps!vR$ZznFh8+)f
zaqY1=wf(}T>1Zta1ED
zFzDF|LikR1&^WON#2;&eyOAu3Cwu!)Y6A!a1}=%*{|rsOO{&EA}loxrlWck?-hy{9%Xn$Xy$4~-f#XHYwuldnDxg@tWN
z^@s?A(2B3)St=Ua=5MyUI!VvvWk?r`FT>6eZ0b3_d
zO1l79tAQ%?T`<)L-j#;DCxLYqn|b@)N!YcgI@-UWQ(X|>VES=9@62nkrU;r45
zj-6@`k)#QK2UlN#LwDOB!eoroIKhcC}Pelpt_9kt1ZOCU9sW
z*L>NRy)Za5REBtG;J@|(j1I72(`&avQ;fqOvjZc9P7ernL154wV9+QGE|sBQ2PWxh
zgY%S*NCi1y*bO&;$84*L;oeKQPjAJR;3ptRau6I`iGM42+Y@#)SJ~dscl$-*aX?uA
z+5>^Vll`}0oh;xcHvRRGe#d?UUuIj3M+9(5ABKX$gZV>BU>enNpdWpt4;%$8Soy}M
zUzJ6wi5vc^RO!QOQ|P5*=j+ma$9H?@`(+4O=+58p)1t)anOLe%JIncV4`29wAYJIO
z-0|z&6!f-H{{3A4pS`Q_=Qg(R>zL#BeKF{xeJAWF{SyE|LNZ!BHgHcX?C}xkf0^N!
z+OxA`3K9Uo^yIf54fXbB0?2j&bVfk7T_B$k_+4C|9vqw_6W-&9G8}f?v07x_lSWeByjsARKMc<
z(nv=m<0Bz|VOvuH^$Yu7)RQAagiYPTqC@l6VZOEv6OsGT1sH69(lJ=?Ni)hnK
zXjDuBanN&H_vyP#7>4#4CQTR@_8B)!nD+OX&P|x__nF^KSbzug@kXpk1*~KTtaPSq
zYzJ(7rtD$|>@ubtDhC`orXWreQ(%_Sw{tP&(GR3a0MAM=Pu?!VeF5(?fOj;QuL{6F
zM8dZb%-=v*<3uM9
zzvn6Z0EaSAW(xa<3PRfQ$Hj`XX2rdSN<(JKIyy?~b0Z5)J-lN*GAh#bBO|slqq=0nJae=ABSWz=L6l>2u45CT
zSz{Lq6Q43jYYQurGE+DO^DHX!qB66*GJo@9t7r>bC_-u!LDmx@Z&+~~;r)AJ_
zxxWVx53d4`>@=9pDunGcgwHBe?9^m;FHFTMT<0|0#42Kvl+?xQm(S_15Ua>&8gDy|
zD36t>EUW0EQ&4o3RZP=qOjiZ8NLp-{Rou{N+(w02A$8z&MY8@{U>0>S+^IjBwdQR_
z;ImZ{StSi(vUjclF!+=*_kMxbz+xAYFjZF
z!T^}nbV?U(R+=?TWJZ<=7{=&X#=cg@e{UE8mH#G3Le|tT->YoyxXmw77Yp
z+D)IfEd)5POy9w_(`8W|WKTc!=c4gmuQQ&$R-?RMgT6*Xq;w`E_uckSfN~0)E!3at
zYOG71m5b4&#oi`*sgjGKGLhl$>OVho2eaneBj}r4^aqMAr*x?Mib7B>tLw@JHZ$!e
zOwdQTZ26EFn##bE?yPyA8OHch#`=!;qH5~IM8K8sD$cgYju}*kYCb>hI_mHsFc^l2
z>{dhc3L^#}%437A-CX@a*Taa2E{97w)U;nzX0R7&dHd
zLCgD!P&l^ZKzneGAPBXrDVN&)W8^iLpq8wGA&sEaxXae0v!y&n(Mrax$yzQBI}9}u
zxSE0LC%vYbt22v)Eei*9t$~;c>;K+i@#*UF8uh?O4yt1f{jhG;n{oG@DL*l1CnPL~
zEGuX9`kc(M!G($RkO?XU45DH7zNk)|x6-f0q11dJPulE$fk|%NVa!hC)gmbW`Fj7`
zAsMnh*~eyt7i@=&;>CoCzq1-V3jf)pYCcj~>5maVn=OyaG~dv{H=L~OwGrQQ{V3LL
z4;4!}(atUbQIVRV?KPDtX*&X-O|AGgThcb;Ksu|a7uhvO7!FmXmEXvD0v
zY^?NE`{et}v`vw*_n5wN&U7v&Sjdf5g=EoojaV_PIPGKDnvMUFEby>dF#|QQ+4uFO
zr3w1?v)y0`>!S$aO;bpMz-}9}p6|l$SV=ydNkGn|&`l73p>rnV8C}}e`q@LV5XjC28rAPh!Zk6
z+qVEDB3z62Z1_Q7ks{E3uG|SY+0WwSrmkGJ>?e#5k^Zhc2@h=9>;TUJ8ymFv
zj~iE3%|{8#768EmDD^LfR|{Wgi-c(l;P>OtY>v;oKQC0OLNARnFL#XW%>cI`nAaAV
z(-y|;W+5$C1$Z!p)u#U%MT+2@?CH&;e{__wIaQFG8Q}+1KyE7VZf`|upG4v!32R)$
zU|O3l(0Y%G{VfV(9P~j?%I$>O*g@Lt?mD`hJ(R!MULGRN+4ZJ4K|1Lzn%D)8z;HgP
zR+fi5E$nAb$sp~Q7H!)gRYy)yQ!u{b2Q9;9W5ZS`77w0;0aMQh0ow%jwMVt~R)Ezr
z42Os1lxwzvBjFEM`WANs-)7lG5BrP9Rv{;vKhF-U9**MAI{5SdHJJGg-5qToBLUn-
zF+lm%ztALH=Hi}ilAfQ4ZUQ{*Zl<0d@L<6IQXc%OyRD~JF}vsD!yB+d4tW?OO9ZyT
z)3^J@ci7W!*wZC@$Y<)s|KPZ@0>B5w+}cRDd|s;)N&yC-g;2!zx?X0dlx^&5DFcXqXR{DpXC_gfD0K+YWR
z|55jrL2>t8x^6c#?h@SHAwX~l0fIXO4-i6tAc5dcV6?(Xgc*Wm7MA$>Z}^UgbS
z-kPbstM<&UbNEaZ-}v8aUF*Jn$U_EY!_DcXF6r0JDW-gymGfZhbL5GuK?ZmK--%))RsJ%odPP3pLzx=KWJ9zF-A=qf4%48KfCf
zoGFmD(p!}IRK_fjA(d3GE6@OGZisY21@iO-^dM0@6<|Eqtn^f0yVP9s7bsa`RkfzK
zwx;*CUR20==Vpo7xaG>V<|DL}#6D%Z31%X+0NBI1LTRh8ZK=@^1dPAj9Wk0Y(kW2w+B
zg@CyGE?$|>D@xDXUap+ZFT|q4-A?UY2G4sI9{Wnq
zvx34((n1vnt~=Bn`KcZ(937yd_FU$ULlTjiUXMfS?%~nM^<4gq!pH6v-?L2q^IZP*
zLi`Ik@I@y7<)m<|Mu*IpXX}C|Tix&DT!CHusT)DhP4e#DnRk0!ZTs8am(-mg<95p6
z&gszKyDJ^1pS-qU#h*((Pk1Mwk_kxZk(O4ZddA|ON3Hz8*rVSZxo=UvC$z!4q>2K&+%((5<13|@CS
zn)h$Mk~?)aPAd7_$9oAE2P$QW_`ss!NjP8pE3XzkRlyJGEu5`ebR26bj*&2((v6_zi`gI~b-PQdLltRy-l>I(fZ1aAEoEi5t
zA|rI6=1?hDJ+ee5MUg2cTch0(PWR2pT(!AvM!PnSsZL#t*p2ryqzFeqnNq49x4|~<
zqjFbJI~<9o%}AF&!oCh%O};^+FDZ_vyK~N`5k6+M24`)~RloFCuJA9~(6xpAJ62K$
zI)izYqGxy(V%-zm`9l2O2ua6=Z$Ay3p3f8%gO{afF$gPz)1qvp$JqJl(QB;`+Z}Wr
zg*4hx*EyZT<3Z$Wc{i6)b+Ma@18$c|y9)vIkPhJ+!;3Zg%*yM=la1$>dxVR^~ewwrp52`UQ>I!YjmkFuF+i*~4yn3{fTU
ztz@vmUUi#&@5QikukIz`CuZff>XJc}CU3>!Or`R|J&^WMSCN^*2bV;fh;ky+Y|GHP
zp`sBp_GWRSx>kX>MmYv?xl&0d!y<)clX9)`p1-kR2+=DBR|gS>c1Q~nmGKz%9yl%*8({4u3K
zH7Ugd8&FdU)RFQZ1u6gHMS+7*&_@`5%`hveusrUtVVShIT6eBEcRz=8W!Xrx5Je8<
z*$!I@WKI>?%a-d@E|xitp1O8!R5{DDrA{8yxsHEpUB75@SFBt)e`xcZJb#9Oy?uQB
z`~w1mfX!XqN1qGMvAc9X=UYCcR`i91)
z=9X4YthUas?w(#1g1*6_;gQj?@rlW)AJa3lbMp&}OUo;(Yd_aFHn+BScK7xV4v&sc
zPS4ISF0Za{Ztw0N9-p3nz5IqKl0CpkS7mkkA_l%ENUF;24aB6DO_Q$P;x#~JH7~1<
z5+4Ys5D&rK0{09?GpIpI^oDElM-w?dkEF@e7L2Fz*&VG8*A`A@iuxjxLYM15a;0OL
zK91BC&lD=>%BIWKm&}!F)tax3)R!()8uo^e$~BZN)tb$ed>m~kUum@29!ZyLtXOMx
zzCKzTZLD1H1jC|`$v0JP_6FfF>yI^6Zx2RM%Vo$n*X)iaaasHvYp&g&%n}bJQ)sC>
zoGDT()gN!EKVGQ%Jer}<+HktkWOw{?ytVOsy~`JcT(Pa`a(gJ2*n|`<&W}E&5CPkY8B#sB0ffNA@TR}7lW=8mEX+>Lz
z^yLRzp&SDTIw16^AeC^kO|#Dtyvs$o5eTXj${;Z|b(&~#vSQ;Hd7?)!mz6n!#pEyXuvCXTiTVM}G)Yoo+S|z@1;KkM<_(7c@6RDid1=OUGFb`E8qK@$E>|*!
z>ELNbvt*c>uSKarM976wz7vG|xk?R5!
z22{=B>S5{?x}Bsddm@J+kyi$^7zTz{0C8+=z+glQa5tP~Su}R&bI-w8E
zkgyzLK7m7i<0_X~`lp<37}P#CC>o*`eLk%-1>AbUe5nGQG_6;NUUYxvc()G_umN85
zO@4E`uCu6m_QEy15WK0uh-^O@sGN8{UK#xD{xIEpD0n?(LEdpW^sKpiwaLZt`;N$v
z&G2@?y7Kqk06wDU!*#Fz3$lI3#VLl{2{q;~0Kr>$)C=`_af54g#P;7<2VN^E6gOnf
zoiIn0-X|z)UU*5}@Cg_`kG*7YsSEAcb?k2m`ZC~CoqC?%Pk||fG7w+H#gKk}3+OA&
z27i0#e)!Vm2@^tj111ongdg%{8rwv9z6PxKvjfp)*K#8E*(o1OVK5$y2>1b#kk={E
zVV_y3i(qJqacmC7vX?iBY74R%9lwPXjpmZw4-JYuT}FaY^C%ExhQx@jqJre}sPKn}
zBn7Ubqr&oN=w*gwb*^HP#`5TRhldqiuVS-M^BJUMMpRR-;)>++nKXw-G+M6WtHSbG
zOl3xOmah_;#`4*mhDY_Et`fUY3pfI0#*B!rlZNCAI1`7*Oa!izr@{)j3uVU5b*@vA
zSJQZ;HiYRd)j;cE;(WbG{@-owkoQ3J0+YoPj*^t=BnNbuAnZEtsaV+iu|nZ{?MctA
z*v!{cg>T`WCcR#qmaR7&+8I7~x_P)X4t9JnG1J{tP_4BR`gnA+yw64`#$4
zD5j(Iar1~y`(+rPd}2zt^BKY!QT2#mQzq>5>Dw9QwOvtTjjbp|;O3-NFlTdM9E#{U
zj+7FKVDpjVi#cpc)N)ILTTr9YHvd6Z~X9(}0ri6`LB$C7xHKTmRs
zTWAqps=J?}(XjlX!lD28<7cMn_J-R^o83}_7otUfbi7LcFF!sVrf9VbBvkr;WHDSv
zo-r;9`Tnl8%!F%H+x^>oZF)b8nH2N#wC8+XwlMV1jp|HS%-5H{XSFnCURmy#Z>WY|
z2HTD5E}zXew)eAgZOG_34$L(5JDA!=mg?;YG&WCtsj%Oa@!8e6Z@!u7KXb@!T|2SW
zY~6UNaBi2=Co(&2qZVZ+T@W0+D_Lki3}<&+mos?maq75>&qbeJ)*4cs?*#0fd%=-x
z{O06V2Psy0;|m$W<2Q9c^KfEsE%+N+a`N{7!*aXi3*h8p(1`9l2)vVNgUgu6MP(2
zj;M(%l;362eH_u6s7YuX+hucm95qC*O&XNngL1PmGsW7Jg)x)Wjp}imh}yJ$`F(+v
z#|h_&+Kl_LeUazKNicd{7J|Zo81XU&<-Y|Dwto*A{{IXb|54bcBbNPt;%po2*!-uq
zt)zeBZ*$xKkhTq~V@v$MA6VMjJN^nReffXjS!s5)0F>}P=CkRebf5S859+h%i|I4EjLzeO~;hK>xClw+}Ncltdt
z1fM0R+}^v#KPbIX*;6zdrynV0ECLy6vyNt3oBN&gN1)LGaZ-u|$*XR$qYpatf96y4sxR;nGW;fG|Uh4lT1nu3(_2q4hypan2w6_63mZ^
zi=eiMlCp-QqtdDYrsFbO(J#lC%0EhuE86akn2AbBfeDH%Sfi!-%_Aw@@bIXHU?Al$
zfjM|s_11TF?XGIwpgE1_V*^U{sZ#*4^yo>`v=1hvr*dG2gKe$)3%&*%`#&x>;Z+@5d^gm3r00#(XbiH9;
zS#}YZk^~A_!=WixSu_z_g26Bd3^v}d=-kTPqvTpPU6TkKX`(PooPe82e)^58R&oNg
z8vr_LdnqsZl)L@3TxOX$JP8S!JuKSIQQZQ%a#d}Gh}(-Dg6X+={Q@~+_H806d5rh+
z+t#V%WIG%fn4buJtSOkoEyztU-9Af=6xO}ze2D||AEjr&`{fo0;iaU!<#C!B`I_0c
zG`w;%
z+o%}Q&$tk#q#{p`ayHdHbmlufU+$KjtOY5(95C$=J4|g}$MhFQwq1^tu6Iyn_mYSq
zQn;^vk1Cs~@H%~-sX&y+ryiE!5d{D_=%kPB$e;0wL~oHczq>z#Vq%Jdk^K5#KuMXm
z0ecH_&hHI{-}B-lsI}kFYpHJ$IzmSVt{*asb$O#Ee~V$cCcmPDQ;k0;p&hS-_xaEn$9LET=1a6<$2VL_|E-N
z(BQkhAKq@iwux65WR$S3Vj@>_sc?p6(%eU)vn=z-U>S!+RBip--1C?anv19Wx?m`~
zx|l!9jH1rQB-Hrjvwlh*)gsz+umk~NVK4zeTs6Ru{NC%>KY>LJfQ>j_fF~($gjWRs
zGJtt%wZSm1YJhcAdcHwM040m6DitCCI9sfU)^weIc4C6G&#wehj>@=3?IL3}95*si
zgQd9;6@4?M;8<~;1urTphBl%IdlsF2%mNmb5%sYm!_B;_D+FFNE8?edss1dY6Q`A(
za=4pG0V$hF^6H2=;8oE;$UW4hF$7dWhOq_QAHGVS1}k|au&L2d7QB8hs~*Iup2x?!
z2jr@+$Pu_LRz?HBh$&ag$LE3Q8N88XyXMMWKa}dc@z(4o(x^^(Qzrj73p+nD-_UXk
zM076ITYBp1f5SwB>f8l$GqTY3^q~?p=vW_NRI>+t9&Tndj(#wzjED-EmJZwls~4Zh
zxmd3<9MjdvCg_jjtge&w{(~abvmc!3CVhs$JLqY&_k7D;eWeG$(iCL{r*FDC>`D?@
zOHOy0EvvCzp<9)G4v1TZ*HjhHYHM+vsF)0H9JM;N&OO%IbVLS~_@CLv%Y8&zTswwv=uo?nE302-o7oYYk@SBMw*#d=M
zZ5=2QG{0hlaVPiRJSZS&f*FX2R1`tf8!y{vJ&y|af?(wuf{;vYd(^QaRT@L>E
zfu!KmkEn>cJbHygS>31Uq=~u$-tj|4x2Ksb^!g%*l){l}>eFnIVtt9`_>o5I(_B?V
zeVM7kvChiVJQLBXyjae${`J#>AM9O)ufoYE;+2Jd$-8R*@)HxT=cVC*yP8~u)33TK
zOLGHv(AwnbH@oMRCHVV>-j6377~Nd^1wARIIok|b&)$a_JuM3_@y@a6Kd-*=wjwQ{
zxpnFyU(es;{C+uS0Hv*;-8HtORamDzjogf
z*|(hfK7La`kG%EJ(Kwqe;@~fpex%jtI#*c%xh=DMIo8E!UFiRLTjT3-V#X`DG+A+1
z=POLg`Pj6oCw||&^6|_$vi0Y2hEvD+%bC}?;KqyMLpQPhMG!{Y7D}e`0Q>KYNNb@T
zQl-aXUHz-1$Y1=5SE4C48MGM~9{W-luG3z>Z*th%k5sImW@|lfDA`iZz1b$+CeiB-KlFok8
zNq(|Te)5ZciVuFu1pcag{_0x(8qWS&N&Y%b{(6i4`Vanw1OY~T0mfPZCe8t7Nde|f
z0hWsa-yQ;N2m)>S0`0W|9i0Q6lLB3v0^Jt_Js$$cy#Z8j)KxwN`4(vpB{E|6tAs5E
zg%JR#qD;8p8R_#CCu5WolawC)6sR|Vyt}H$Gyu{8MoEX@>{NUD)~%>zr9mKaQYDD_kb>ztlp1H;p-3Y#cE+z=qL
z8-|q+gxLkAOT!RdgwgImOF^ue)nP$7N;?NIyUr2IaS=ya07+#N>3fD_F_7|?h#TjK
zKK01GAiy~6t7vgXdP5BWVH6C16vzceYV%8I42(;?hFgD>f4`>BEx@H-6NfR{vLC?6
z5&WA^)2=>>%w|8QVF$0g0Y9XW4fIb~#OoCWU*PL;kzfK19$bv^d=$4WGE(Yz6
zkr7Fc9bV6pDgw+v|28@(+dnFVDpvfFW}`k%@i9)BFkY2EUR^t0!zEse5S4Pt3ytqh
z54QfNC=J_yxWJlt1rBWmr44Pd$1l+YMA3^0URa6jSi$t~62H4(XU77ggRqT*8Pa~p
z@L$U1*9j-oeJ}i;6!104nNTD+I4SI_%tEb9%;8&wLxv<=S+rT`*ZUvBXNv!9Li*od
z#r|{p`R6wFznXsjWi1Px#{S!Xpc7Km{|6@|tba^M#Q!G~(qQCwD!vQ&n!MpyY-U`_
zV9pG<1lG?DU%P_y;V99K_fjqjd@&*MnPk6aljRslaWrnJb6
z|JQ_6H*bDi-n3W3{F=pIwm?1ajOnDR56|MHdWf#{q^2@xhXrxc73WlXN=$INep;2;
zs$RzR8ld&cT`mE73N8oT=oZL9=3}Nk1!G#Vx|l0#uRG)_r77~5@>+HyM%i0_!>^vl
zEaIJ%{Sj7Q)XU!;b97-qo-1Qce_EXPU+Sd29yr@e%~OUEXQ}CfaZ0@yMkx%l>7@>o
zx9vN|Zo3@2oi4j|3
z4z8BCepC;CJa5`?frY2Tz}m`j6c=K7yUbK8bi0I8D|k0;&z#P+`Z0Iybkgo~IY(PJ
zI^j3CZEUhj*j=jMtXtoo4@|e6*zJI&Git*9nI$=2ZtR9+{v71zn&~416)W8WHs*~%SQ;`uI`_^-=*?m
zviBcSr|;NEiEb93dB_O>5E+!gkNg+jKn?(0l05*hpY?+DiZYIrf~RC|H{I|tZGEdQ}*K}*+jEj6eHc@@E0MWM2HZrCwakO5}Qhl
z9O9JVK&@6#_z@S8*9323WmQJlH%#8jVj_SMO+XO%Ayf^50Rf6Hj0ID&bvw764#j;8|
zs`R9n`4H`}40-mHF-id3cWplUTE~%@47|dy&oCuQfwHqXM7KrK3MHzEBeMm6CZzBZ
z^+MUX5}n%;t??2Ks7koP^|lnM_0$@aov%r`Ei+Rn)miv!LMpcjFV)+ZU1(Xpt#BSM
z)xRHE=y#=M-(ep9SZ@(CGFx#f-VxU3kRtL
zh|tqqy%P0W)y+7huo2m9KF{oKT?PCcU5CfBu?K;>;Ml;x-CguHO+D48C%27REDTuK
z4zP#Y{on%YLpyAp0bojM15pkdL@(Ble=`gGgj04TqqDhUu#|zb;OphU+t`7cCsEfG
zTaZkQcPNIoM`djC8_&rrGcm>S!!}Mo9NJ$?km;D5ymo_Sb+^O
zk{--JYLZRlP7)Ii-{xEuGWz!)_$O)JsAoM`PD@}dEB&HlOh;EZpqe3>5*mn&w33`o
zeQZ+UbL^UsPN`*KUZxT1)sD|FK99s5AAEa{)!Eq|os0CMKu>D_<0Wr10rH$SC`PY}
z$SFTv$h{a+ZwsE%(|d}IE$yO~bRN2jjL!*>
zs=3CL>}RPRnE0ik9_!v)-F5!j#`Q&whTL7&k0I9s3WwT_uDq2`#qA|T~m4~Q#j9A|`amBq6SNCz9JAp$y
znpK*dT)U?=L1H!H`8^lL%1|`m8n3<03`fePKu4Z$>X2mHA-AzU_2+ARKVEi#|Eoj@ea7T}A{4?yx?=t%1q#F|S1ww^~4Q
zz8m-h_9ska&#@bqtX4LApFGIxV;;T0P_S?H#Z%tZ-#aw?{i`6kS6^sj#S-G<6rv;k
zfPIDUGP&6*`R|JgN+r|EQpHm$YwOe~#IT9#bt+nYUAtU*5_?ZqCD+LoS^AFSIGZ`ml_Q
z=NpvkODI3i)Wk;lL8AD=0Upy_ML%4k;D=W=^QgxQsN}+$g3TDQOQmWfn33T>WBY}+
z;ayT|ltRIN|3zq5yzDTG_zV>6^Qz2JTU8nt`N~pq`!7v>DwkVC$zzj2!Jk*J583(~
z>_2|}1NLi|OqBkB{ap1>;y++t>ncZoT*q)}{vTk!tvlUj+FPreRLJmtHclkq_-My5vWh4eGMO-%v5vP~c@@
z;!E2rk?M_Y=GnYYC|Gv+py)DeCbk;$#ZJr!1RodQS6vnyo`l`_Y#zE{X3j_u5{bH0;)*j~
zW#=OCY9a(dkh(wnyi?7^=@1ma(iHi}(`EkerJdbW@-}CIK4b=s;LRAb-nh+aVz&FI6}Y0Y;t@JHRfODEzvC3V+gUt4jL{
z+&)10f|SlRz-w~RjTjhnx|S(!1~uzRSE(tJeha0QQ<15eqXChD1~0P)=wR#+c|(eL
zT`2H@e7)LcU1MRC&1E1Yt&4m39XY7UFr2!GY-{&Q{{v9m(;f@~E(V#EPgw-T1K-4U
zzmPp%kiCA#++I2u7RXN^K|qP)=%oz5d-FBwUZU#S#BZH|1NQ-o)PY9B|3lNSjRK9w
z(;IbV)C3@+n!n8p)`=-bYQtLw4X>(-Qso0<|Q|CjpoAGr#C{s-PoY`!Aarc3&DUwH^X8-qJ4fRIK9Zghp#YJ
zshIOF8eo7R>=|rgq0=H3k|rrG0pu{)+e|v}jKiUy^A@w}QpHNn&1?RYttlL~3`zt=
zav)lUj7DSC`e=!$R2dVKt!7opYjJV0eU~$ErXtAX>aFa9gyTDLF9eBv>gE4$1N-E@
zVC|MG0Bu)3tJAQY8kLxLJL)c)FBI&v*(MGA8`$R(9CH6B*f$NP<7pfo|K@s~x?$*nXqa5w0nKVzzoYrN0^ybS>Yle440IDOEaC6Fho>w8V{w1K?
z5Ooj24<*Wcm>R*hAIcRiS@XUZNhzs~)+A&le73@5b8rFGro#lHo
z52JtRs@bu3i|D3stjDR<96e@9AO0!8=)I~|pMXiV4d2P96?)Dg?1Y-=@-h?t#=$Yy
zyPDBOmQQc-mnU`R-}*$;j7L%Ai(p@}8F-Xid>Zvzn!rPDa0tS+eb!k$n%8glfPvcV
z{5L}!usy4kaM2m&QT;6?C&Sq~@}9*G6z^N`Yri?VUFjKoXl&JcVH@cw{Wy`L(l+N0
zLHUk?JbOuS-*_I+>b}ps{#23Jk@4x}w+JGl7i2UFCy6wgE3l
zfQkqO`}Fdg$Z~u=coS6tykndGMooW#eeIc_Sb+h0bbr9UUJ`aK0T+YkUtmA3sc9fO
z`1N04pL%;hq#`uJ_df^wmT#F=wV4ImI9(P?zKr$Nj~bHgvOf|-!M=76oX8)rzsD0w
zIsWZ$u)jY453t|*7uYuq;Q9;fJ84Y?{MnPUYDsYY*^_gNld}Hnp4=IG0SjU#$)(~8
z-IK#_E2=4k@}7lEmrcOv1acgkwA_|8d0MbJkuK=Z2rP8Xzp3jWIX0T?8_9;N3H$+$
zazI`rLXfXY$!_@|W_)2aFwxM^IBumziDueoPQaF@K(21;(}^4wsVUMA;a`&(4gG@`
zc#zUA&aEPquLM4WVbIO=ko2ywJi))r)*`5B^(cKX0bEBuTG?0X6i*XwC&=X7dsS>=_vJpHuor<9h`n_jC^>EPN{-jyt
zc@sVh^oD6(}Pe=+tEH`Z=JaAI)kgKnpSG8=vPv
zw9W)~KoCJd$b)~L?I3X!GsVO*RF*t3}69F?Mhc6^gE2O|V
z#7WcLAs*o7uUR$};t3a8(G=od@7_1eM;sC&i#z|_&NxQ^JF#_nnlB7k&cr?YT@&j-T
zfp{E1t>AcZet_Omyq-1yej{F=&?X)b-+=WQ!iOX)%oq>i0FoF2t@(lC&9QniiLihK
z!>@_10vqsn;yE+A&46FQnUKRw6mTX;
zC!^Gq*Ph;kulT!mjd_ksfJK%F1R467y~$Fc&(;>RCVJy0V3!fy=eZ)DnLC)RGlhkR
zY^DK#ftC-^rgGSxa@bn{90Iuy7yaOoeJh1qY@D1|fxN$S|nU1gnLd0^A&5f?tO&
z7ez6tMTQh@zAFw(DYEA%j&xPBUM@66F8FL~t@g#5?ls)KWXX$i$wNa)u4@U~c1ek=
zlA{2$q5?!sq0E_&n4K+#XdD&FA{ROPlmyHb%+VBS$QH>G!N6C2X`3(W7$}xpF0Drn
z9Xqm^41q6Rl&f_u%Pc9Ax$q1aESz>N=+c2{cd8HvS9DRs9hp}g3c#H%S14mvo{v=A
z9#wRSn}4S+C@?Nl3n`OoDR>sBdQ9K}p`>&aR^m8U!RuD_A-kXSm8u<;N}1={P
zjoLNIy5Xj8LUauf>Ootb_=b=54eI?30tXG|D-E{K4R*wh_HK=isf{+RjSi!Y&c}_e
zD~&$SjlQXNESGFmNQm58SfMLO;m3%PD~Qpph_TPuL2eN!w$1av9DiLKZC$?b^o$wW
zdf0yTaFFzJ3iWd9_42s)@}>0(wDk(D_KN)K6(#8t6Y7)D>yvcvlTPcCZR`7aqk-7f
zQajM1gW2Cn2T~P6RJZK^u!^J|W}$O}^l`Pnl%`S&942Wvpdea#Oax~hR@qw$vJ$Gi
zmmN@?8ZeW0vvaSsZ5yyp11ZEX2id0$xulM24>b$}+9`S{+J-__hhDN@lb;Mxk_^WQ
z4JV8Z#fJ?imJJ2_r8@E__!EwV2G{ca=pl4Wj?o5`q$QQC#=Cw^CN>1VO-^=T1T?zJ
z;@L?Ga5
z#fAxmcQvwW0v
z%d=50W?90yIj3{Z3nqnuRY@L&f-hzP?V$4Ty!sH+S_PDXmz*Mxe0yQnR?^%?57SOz
zp9suVHi1=E^Q=;*Rko*PxfZihQm*9jc~f(irM1r$TSz~UYd(;!fal{@(ibvEid^Rx
zETdMEC3AGTmgLJ+m=s(^QPu_V=E5n~(+1XwhCF=Qra%?3XaM$G+BNXYfWVv@BxiGj
zDfYD`88o_P`->#~D
zaAd=%B>EO`BcVrV^m8&m6IHW_Z_ijiq&2T)0Bg^5ZeksEBIz-4MsJ+zTk7bOruS`#
zbH=`F$G-b-=c
zKTPP^s+cuI%0@_lZv(3gBg6<=w>&a-Iils5wEjA2t9_*K?fW}K9J{742(h?d^Y1YG
z2TI9nHDNZp)6eKY{IplV9qz{Ep@@^DTM{8YTp<1_Eyd}q$P}ysh$@vIEiDq~F_3n`
zw;<%>+pA_S5zIFk4*L~NhY>M-AyZ++X)|Q!vlX#dBB$_%G1H5xcTMN(e0n#>=MTD|
zJdscQ*gx<+E)WfV_1o{3LBO*F|+KFxBt8A;I@@g8dW
zX8iQAYx;^XlRW*3-b!E|;oDkA
z8I&KP0=@2^f61{wQfI=w1|6ObE8i%H0*Kn9#s9TKV
zC(m;}+gmD5_S?uHRP1%7=J=n59xKgy=)#@%!4Dm9fruy^Aa8b%uK|qzB*=~yWaqG6
z>4$8tG-o-9Z3DS@1DR24>&N->9*mA*Kz^4_jx$W*WTw1&lV_75`e+l0hpeddJb(DU
z>hY-OzGgDW)I0m7Z!$0R=`qQ)1mQUl|0<(0!c7CM#PC`C>aJxna!_vxO@q1|0_uQl
zh`fDf8h_Zk!CMl(9b&&8GW=zUXOFmYauG;8s{dwe@>viAk=OsQ!WyeIQfYA#xlhWf
zZsR4%KYUj;-{R4wT!y}_(;}erp(=%I{
zLL&0COS3C7YtrdrHI`aBD;i@53bPYidVeg%+au_%&+H~1ti3uusRc3LpbY-X)~&1f
z_|8{vb`Gc#ND=E-D^K{fNWtiRkd#oBchrPYU#KdXD=t5+v2+Znf{(<~rD2{7Hr-yY
z=9eFxOzd*sNBCzJ%T%GNuHm
zK)~%(4T${IDETZ&Fgc}!cAqD@<1I=o*P53KcHbB!o_%b0lFiw9P8ibf^?#^WaTbBB_%v?)0YN8M!B?1~V>2@85<5?`dK@(wJ@kzRNn@8E6
z$onztG$1^0{ei}%cYEOdJ2suV?`w_b!h@v3P7Vh_dB@)#Ez{}0{-g%uV{i#QUak*&
zt)jO`1aVVx*!U&cdU!lbq2BrycfvqCH(%tKx?VUBBJ-b*hClP~VA}Vy)Jlh-AR*Gm
zTreZebSonv@5f5Vo%f(%^$PiIy~Dc1yOVl4piKc4yRbaz
z(g9m%@$q4EB8P?Viz~}(bL$&(l?vOfyW%aI`upl9N9PxjOPBJDH=id@@1fO3_kAiJ
ztao{RN|nwq;i$0#9MPSDn5Fs5aZ~+BbSy!6wuxm%c>-jjAvAc^dBg9;IK}j6_cvr@
z4f#6K4z2Y>*k?uj^9Xkfa3RS;z9q5TL$oP=cS&)zreFHyebvH{)1Sr*lY(W0h^CLI
zNHt2J)yDop8lseR?XrkJ)y7!16R^o82mLEX$JC15u$_i;QW=||r`uK}ri$*rtBtQ^
z{}2!puE~=fC;@@&l>VtUTFm;8DLid1O#h?WC{XP6adciP2JvKRCd`Jmj@ZfkN5|V_
zM8!6~@%DkY#HZ=*(yA)9snBZUn#0Y}Vr^yf9|B?>w&#IAAaumiE7R0jv6Zsn%yKq^
zD&pAA=oOrYnQa14$a|vMs{ObY%MG8J4q1yojj~DkDLFI8^(!ys?Ie&NLTso9Cw;}P2-GH_%xA^x0jh;M6BaEKRb-ms(@)0%a-mnNOG
zIgnt#H;j?ydZ+!-$CC)5$nz8YQw*6uQHdGzD>i6Tz)KX`6tLftIn2*O4f$U+1^(?0
z=C7u}ab*`g^MAa98RLQe|D2LsS%4!lV>+#0G$~~@Qu#JoT>mqG`Rwmz<3T|wFt-QWSv|
z`eYKm)y)rS`tqA;dEV2T8D%Nf+gWu@tJ^s()AHMSJ*UaqryG7$ftGgvL{9!PP
zMd7Jq;H=N{-Kzb9)%}|De);`R_xscPbua?k!-gOJw};Ij`ih6GFy1q0wNZ-gaVJjm
z+n+m_ipRY)r?bcXtU$J>gS^CV|G0xGYdm{8t{U_#Ssk`yfVkl(C%6mFw>*oxO|;*O
zpPu#MTmQNoqObgQHOBk)J*j8XVjHm?-)2745i+|j~yc1)o@})q{MwF87!A`mGW022A)*R}=Yq{{}2+Kw_mF^{4z6jtO%SLw^
z>Lq)+2o$~XKb?a0$DlhCD}9K0RyOYY@PUeh)g^~6OS+#?MlE&bK
zilF^ZQy^N-efonTUY}%F-DUW!>W>3aoSrXuLy@_
z#anEnD{;e;B)&4KDPqURpX9{TdCM4;|A_6{*&%jfEY~ICN^Gc8-zFg()#$lOuDKy$
zi<_m~xzIDR8>3MA1hh+C
zirZ$%{xV__#FdyMN6$64B(LgomX0|0S(@dL-`)H=1NMV)WSa{pMBq61^Np!cYmtiA
zQH<2yYyvzs1vn<8YoYJ4#~n11sXHWz%1Sr@QoKMIx&H<@?wI$$M8sG
z1!iKPihU{JBLp5BHE8#-0EtgjEmZ*U21`Y2!KOqW?ZO+|d9DAB5SVD_7?V2yM^AkxkfrNg?&6liOeRlcfWuJ
z03pco<6+5MTZebXs=Ka1eTh~#^r3X3Za+>Wgg{IZ(R}C~j4V!+$W$-|9DPu^rK#M?
z@^V9i$g7XA6kULyPYsQ&YpfuOC;%d8p=n5%%|<5CXH($5d1|-Z#!XP~=;^Nc&W!kX
z-3oYTNnCM6&YNv{sl{UU+IpWB;v-yC*ZiX=q7C4#=qjFw`#^SvcO&QxqU&1!{dC`n
z`cMgml)~5;p~GP>o`c~~z2Vpi;96`dx74!?5uSOw@aU?%*%N(0rcIq_ziqrIE!TSw
zHT!<`bA+J$s>W7=>wYz16QqLr316D;JFesmd{V|b2S`(wX8vFSDqxSp=2oUfWl7~g%f5vh%6|HUt8N&tQN?|Scnpc84A2eNbHy%eIBcz7cTsc-}wm{LLSSr0m5^i~}0zK#Nra;TLz+-V4
zf&rGU4DqUMiW4K8=fym3$C_y2lP`j;OH&$-b;+!!=4#K&lMRjy)xxJ%ZmlaD^Nvl;
ztY@~d&#UViJXQo$G~z6IYxmBLp%j$oSuJkR8zhN#;w@VD$!8i@L}xM)N*hYtUr4tk
zyfnCM7!-TE4OoLnJ`5FN+9p)G
zbI~d_CrcUF*&*}FvdcHgEwe;CsByh^MT+OzveBiA*H
z-3tGaKF58W5_1DkX#G=Bq6-MqlmQE`Nk+kh30NSfAa$D5JGPqiwvW=c2Nk^>@>9r7
zu%9UwISB(@AY4KZOCKK-MmP!O?SZe@qOSx2j3}R<h#ids
z@Xw}zHXkr=%HvtoS24&}VDU8!2SCX`z+4Mpu^3=E6yUuZpb>;ffrtTp%Ybde4-xtf
z942&m?|)yT9rIq({*4A_JrLZlX153CUI4fT0WiQo!lBoU9{|w~!G#W*JhV{VDv(AL
zhL%F(y8~1o^Z&4SmSIuw>$;yAhOVI-hb}?7W9SY+kd$turIDe#JER-w2GOBY8Yz`7
zNht}HgYUcI-TSPy&aQR#zOHk=&G(tP{?Bti_x-zxi~yjx-4IMT&}a=*a>bc#8JWKD
zRIp#0;UtnCsw2n%IGQKEbJhs}4?Q-!l9_WYpVlbEAvxnhYOw9kec*i
zF6kZ7y#Q#yIvM?(D+#R*z|;OHrNuI>loSe2>B&gGEwRKTHqmJ}K}LyKfuznw0MOxp
z;g*!E=@gu8>pL7dGZO{eMQ98O96&V*ghJA(CdH_DY~nvy$qf_p%99WyqM&sFE8n=R
z+D&YRZGDA^%{DHjzDDdl1we)Y
z2_6HI+qk@JO|$t;WHU))ARrx^m*;C-p+Y9%LMH1aq&_4d#2&^hPiEE!D|BZUCzPzz
z78~qIBRjcFq;1fEH_G@P=>7td2Y)4%|0)6T>f<%qXCd$$3u`2CmN4TpmsVo5_FQRr
zLY_+|^K=%@WftgzZDDdYR+v>0U0!O1xh#5~M{wR{vZnfv|G6riQEcBuJt>0-SZ<_A{<&t)_;mlR?)
zo}pi;ClLysl*;VSmVlbof_U)d0Qj!0k99E=l!SReG$BS)!Otr(9mvb7(aOl7W$Ibr
zYkHM#3vXj{3~c8xL(=k~sL*F<6i@9+EiL$^`!%^4!kw7IPe^rK(PAT^1bAB|S|bYP
z1kOPU5icwwUS&jt$;OGhl!kW@*xGR)G$Lo@>+s4x$spDHG^CwJ8g;4=8QF}
z>?<=Opr)>iZ7h!zkga;3q4mb5d?=)Rz@4uu%hkSFcM$EVb4g78P)wk6-if738AROdjV^mhY|c<*kZgrf9fdb}
zox^68kvYcqt9Z0*O>}1|l=w;JHr25MzKPbDQL4I`<{#(usOQMME+3XhH~b(}fH<@2g*
zcp>yMMG4r5Yp?p$+!Y#Cw4)dGIjdq1%s+9WFXEF{7%4I~(xf)ZHzfm(+s?Xsq}2Q-
zs|8WaN5h%3paE7GHD>){3~y;t&DPMPBYEs=YK`eiW8WOn$7hGvO2FAlH&A2wwfU}(
z?-fMVQY6nNtOZ0-ckZkmkBY?sBu&3i&aT4B2u07^Z_lY}&%#6wa_K0r@1T(CsG`vP
zCfOD?YJ9bY)noh2Gl;lNS**OHvvs+%pcFE}<4h^6Ujc1IQ9}3shT>zL-_E8#&ek~D
z>GN*+%?QPtIp41L6kTsS-^`9Sl&rp47%i@-Z~ehg_=B%Di>}D9(`E==Xu}(febV-$
z-ArooZDb-q5&rI8a~xXazh&bLTm#kJ7E;DoIT_Y!b~4M2et4>;I|>gimb
z-}iJZU+;CFf#@4=MBlV&JHKy{pGdR3$$)ucPq-|&EEP<#3R=scPJJqq&6`nTqyPD
zDS)5&al&Ck9k#>fUr2o91v7x)TkHbtQG2F7?*WZQ6
zPTu8w92zhj8;a47eFy*wC8Q?v8oGSxiP9i#%B{lbM&__i7i>1;34~
z=U;j;{ylqqdyM}8D=FE0}*Jdh=k*y5M
zpZcT;uwXx?e1%c`{PgmD=X9vc`ao7YMz~DR?cYl<>9o)H5OwGUle36?65;w*x6>J(+N-;UEP^*
zSno6{$QUll-wn2%7g{o$$i@f%$t_ItZ*~j+#VY(qeAm56_;-R=WAE0z4)%{t!v741
z@E?O`_io|8pBamf)7Vh{4FlG}j4t-E=hEkY!+_bkhDAo%I7Q!Mz?34SBk)ZVf5(8y
z66ew6OA#mM{)Pe5ry_a%8wO0JsQDKL>|awgg7_A4DyU;m-IgGsYZCrSdN+VDqr`
zk1U6QYInJ?^lB`nCT9F3xn%O#A&p-cFqbac%S!3ldkk2O9XVFThfIkVp+A)V=@>BT
zO&DnY(?16T_FtA6Tk?XT$rD#th<5#G`7fYqc=K_m<;P;l+TeAu|DtL{>7g^E;Ijgm
zOH6z!n6uOHS%}~Eo9>q9J*y}uNgjAyk^bz(R|In{23{3Kouju+#eHV%&stfS^L~DP
z+&ZuZN6^ioZcOS%vN2CQyDE87$No31?6BqIXD_=}xF#F3VsFsCRn_)yTA2>-@PA$_
z8=_DDvsO0IMN?B0tF~-EHum=GVG-}~&1s>~gq33LZo@-&7m4GJy-DR8!W{F2CJnT
zEB4!GpKtF4K0DN~Uo8iZ8eM%fsM6zKhy&Puq#8^cr`!hb?rG#fR-PHJ!(;$4A@1KWk-vN%oAZ
z-_IuHAN@F=B}oKi;cF!PqH0Y1`1Uy*_hn^ITBh)RRbIeLG}LBw#)G0kaj$FJ&#^y$
zd=dFvx_k01UNAcw(_}|@hw*Os(Mw4>tsClai64i!fhZyQ9~LZH-fKI&c(BvKnWg>4~9&M`?BepR_V8c%5811F;nVBrm!S
zhSv2`f9AZ$-XU<@BBX)&hml%iaN<@40Eo9f5(TXY7*t+)6N(5W7=n>M7VPIIhQ(<1
ztpi?@3Syev;YEWp5MVUt!ADV#VxML8wS
zsYpv1tz}vbxXe8UOrv-$J*htsJL5MD7z
zxzkZ5Ejj$#l}*54YL9g;cFhlkf{?lfc}kU}OL*t%4R~Z3Ef3
zD|dK=uKov-+2K;0hbBO*h%(W8ss{4pS?>8J40&j{1RFt+cO?UW0lxqpYa)@3DVfC%
zD<=JP5USKy*3ug+laV8h`Hrq-G87{k(cG}S?;n#S#75vzK7>WjrFP|rmL5lcAt=0K
z(=JzJkDB&ioGoO3#jG$)z)I)LQyd3IJ$RkIykCB%n=F4|M_cr5-boYx#s}7f_J~OGYB*ROEoQ2QyWK{vgXK^u;k^C8o@d
z6dP$}{q7(gU8h&&+C)*>ucwI6vxx@vjSNB;>^xeYzdEQ48v!ph5LNy(q2_z*Gu!5a
z0H9^WJC_tpRSu*gq22!klA?3>X@Waw6gWJRO`b$kd*lzkpZb)BFi`6@#kW+wW0&7p
z3Jjm#4T)=E&tH?}O-8gZa
zmJ_t{Q2nbOQNMNZWoMlVev%scPb?Z@#Zr_g`2c1)7s7X5(!diuB;Vq70^~Z~BVywG
z6$TrmN%TU+Qo(-l6fS(Ogg_1y@4T0x)joPRAHzLJ=7#>fcW*dH(p
zKHQZexS4z*eV$M7epgmxX)-qKyol)DEv$7joj7n_%=3Ov)pu$7RcT$yo%|<70tviO
zFdRnC1lXV(sUKDG>NUG%MXs47$I*Y!SL5lw|BOj*-iV>+S-^XHvs?Z7K@R?g*!TBZ
znZe?u+GRvw(6-%Kee}YBM_A%mDTX_K%IdQMpx5gX&5qNLyJ`%mSFagyv1T&f@mg_i
zs5*JY3zdu)m-3d7y|=sZT@gX{$h;AmqTFzMZL5C@>&2Q(e|t$1(xUE(IsTcvNO28d
zDJYKgwlq>;cJud2`>E=XsCFLzC~QJxo#Nfq5H59g+TizXrg)|0_l|^x
z%3WU!-uW;lUE(clWa_pf8M8!CL-#h$*p$qU3<1BPR|jSn7_4j=@9vPNGHOVX5>ZVso-m
z={gU+m{iH@Mi^~`FxNVZxK1ERnjIO;LT=ltzYq+$f;(v0u$eJiyW6Rsn-d7iK25Vs
zoei7~l9XbGac$E2TUjxOS=psoK5u!h1`XEm5@I*QARx6Y;kPRdZGoSW+sZ<(>c&-cy#^;rI6om6xF5t`>P{O0bib+5;Bf5E8Bpowd?lf?MFlubpsSwRM
zFf43&P@`$V8=ftG0J7@4hFs3fqU)97@
z!rBnP7)GZMHog#59uDpr0=y$t2a@o*lS+n@&`Oem>f7ACzG-E7MYk*nv@6i!%|D&d
zjdi^U>1&I6VMBXT1|DLLUfb1eUWmzwkYK^*-S=|WOb0Kb#XB~}vaM6B=)QQ!nQt%BPHMi3zr-FFx`nKH+9_(nzExN+Kl#
zkyj*G&BLuU(I1q$(G2L)vhdLhsfUuNSBEEQakFVPdy(zDs4Al`g(Nc!CwfSO1+LxM
z^(5$oycznR;jyIDG0-cJfFAB?y0#&AGto(yr%LR3-=QcXAEnbKo24*5^}cRJ?j)th
zL8OSIr{h&1>)#i1Km*P#jeUlsK=8D&#FWpRY$=2aXc=%f6xibyX%=Ex=9WMQ1pud&
z1vhcn4p{AnbI=njQ`Vs%RkU!D3YfG;C=-f>CQ7*AY_RTSFtI#kgh|FSJj340O1o5A
zIl{6iQrdKwl42DX2V&XhZKgXEraxq73kGJAW%o^40}8^@Hp3iuLwdeO$(;bGU>OE*
zm1#JdH)%$GU-*O^%F;aQ&K|zeo>#*uf9+|m6>{F`qVKu4?|2$|4~ppvae^PiOv9lPHh%
zJlI*n?a^7Wwoz{tr8be_8szZD!X=9On2(gaHFln53L;OEB}`p~C@Ucwk;fLz7Za_q
zG%B(L^XNPFmO5;d7NzoM_T-MhS&W4q>!us1!wnD&$bL5lFPF;R5+Mg<>EP3FY$?R~
zN@v$kYCZKy>YD82@t)@NNsjBv=TGHp8a)ww%rn8?_P79|kJqg>+lK6m7JK?H9kpXX3ra079
zFA2j(Hn9*G5;kr%w+n)Y5dY8JHQ+rKV{%5eg7P!+TFn^-e~;P|FE5`ZhStJV-%JJ`
z1r%+4`rK47i8AA{9TSBJQ|V-Vi!+$^7Gf@3s^Q$=u2jwz-N3_DZ@&}uI5~}z0yC!r
zHS-HtXc;VWOBDL0VVe+*3Y|*n%+`$zc}p9M!5*1)zr7<;l*1ME85nv(qXTC?)<
z9fRCuaz168FxpyLvRo0EnO@FzjyznDw@uLX&}#U+=)lU`rl+BV+Y`LeG_X)Z-}aC#<3
zdH(H6{`mSAmwZYTQURzl5eP#0dzjC`_4#Z|zngckBkk`(
zI#gh{g@S0>cuvTsIi%f4BcJcvS#fsn+gX32U~#^cniVl4omtOz24f1SJnF?2d419g
z?Nc^;`@<^c>j1gu7}o%^+{4p8dTGOkKFIRz0phmhd+jKw({Fv4F57`;L@@iReo*xI
zfTxq@LrmkOjJm_|sKQ_l@2GO8@5y^vX;rsjqV%h?Ik9g!X0y`j*e@DM1lU`qEzYvd
z<~7S+x%m^qPY_6i2yCfkLS9(wC%26s@K1g#pDrONOo11xdLrz`AQ1g8^9~a=U?X1Q
zsBIl$edC%J{@#O$a%B5iQ%+xdsprcX
z`;G$Zc(Ttv$ajCKI1ajJ!Gk(IJCJ`?+sP}CPL(!`HVfhU+(qXeJ|PuDB0mZ0B3Rj?
zUIk#xo`ms_(cm2FNK=1c4;L*qeS07s2&y}|$9$RruwXz;Fb)ZdlpxW(I(<54H`E-HSP`+e1-W~phLLyPR$(z2|D=D2t>a6
zjY|%~f(}XjbU`96C%;=#=w6jvO-v|$@b#V<`;7=!f(ix)@n!;$-8GZ*KW;*0$woAj
zG-DmjL4dq6F)Rf{Dg;J_wOA&C)b_~lY7qeP^s6!+vU~hz80zNY0B<)Kn=iy(qyNhbzaN@CYW4YgmwqpDQ$Cf75h_a2B=Du$
zGFSTuW>1%GH=!4G^*q@k(0jA=0lncps27KA@wpFiE|r&*Yu?FI=JE@%vQTTg_PwP_
z7kjj>i#;b~)(1;HXKj@iR`PfpAlk+2=^c8r>rXP5bDI+B~8i>2!uI%K71P?hhz
z2wyeY5GU%o(tD(vHxh_zcBw}^dIVplXuX{k$W}*2T8)5xZuu$Db%DoeNB13kW#-n=
zhs7FpzRXU0KL3_ve9~^_G5dj?ZpW$rk^Tgyb}({HPo(PCFk$^qG+r4QQ0mf%MLj=**e;gI*FFWBKjkt;?a_H^-c*
zhIyk=bs_ZhRXJ4KU=F02(RWR+UHa+;#B1v(HWFCz91KI44`tLii7E7`xWEh11)A7ak2e@P{{yi;>E_c(Z@`^q$7Q%YgDbz
zg@w~N;?$
z+?J=S*_f&OemwgOGB0@_NrPan(V(z6?`k^>3ph97gS}aL;ZO=wM)1MaxpX-%kg%PrcGRvi
zh(58pWJJ?`+2Q}=NYse8>5E%E&xAa;}eV(C0lJSzF4E@AzHp%x-tda0-Hww1fL
zm7P<;0T;Sh4aV!7z?sVcBSzypCFH=a7BF!{fMP;mXDju`!T<%2$H1Ob
z^9Buvkrs#43_&atr&wTwHYCzR5F;`9g$Hj`!Hnen5jAouXcpnvsik~E?9^kfNH0$j
zG!$8UqJac+mTPiGZW6nr1dI1K3P%*^CEB30Fk?v1izConzb3=Z`He#hbgsF@_sZ;*
z=}e>LG46SW%poHEXBzyilDB=ZI_Uob<(%#Z=I+hG?r580C}nWXNm&J-f#ny?YvD_?@P5D5MOz)?m-dW+=eyx%1
zgMPi*NmU5{gH*p{n#6{(K(`2ti_;*rlF+y*TM4q~`$OTCa8V%}yCYOG9u~haBuq83
zV5|gVd7;$Urv82AZa(4rGkcVKSj*
z$|;iKl#2#0rXp9w>MFccWXufeHd`}ETa(8u&D7}EC1r?6nU~M{-kj;tPWtKt`<1Jn
zuOd%zB&3~1C+2H-71=A}Jq&HeoQPr4aciBzNLQP^P+{hvWYbJh0h1Kjlo?5nl8tTPhOul*{$!k@#qj61k-gLA%VmYVyN4n~&RvzA!hl-hhLb*L=0
z{(^Ur8M~WQ=F?H;w^Wu}&oy1~(%L-xqc`tsbV-(^oLe{k_3LcrhF6(qw%W{GOK#2
zqx$1g^=EzjzbwOS|MEKjmn6#{p|0N_i~pr!>z?Sjt^6JB
zwca-V_vl`~w{QJZ>0bBO`P6^K^7?B@SA4qm*6-K(|Bhvt;`e2k^nMvuX4;ct__Yi(
z{aS|2e_Mv9HA0XS>XyTU(Ek52{BJ1f`X9XvC%p-_AdUEUcV7JMWw@%ca0ut?ILatL
z8@RE>0a4yGp*Cvcgjx-kY+1&bs%BqwgdDXEq}kWRW4undPqKVz89nMe8pw8xrowt$
z3;9-)aQyc2(jG(T{*UhLv}vMo(*2p$psq)bet)fb
z#b?=LF`h9>xRJmj|Jg^bNM_^bh&j7>Ll{cn1eF(x?%*Zx?r$8v!&kUXZw~nrP$2*ZjhNhox5ljzP@1ZwWM-2uUNfD47Y1+C*1knL9d!!2
zWsEX@ce8gK@biqB75M%8Q}G+TI=>aVR}tr)2(A36J7Kz~8B(cr$uXk|u)LGBP5h7GR_|>S1ecYd@dt
zN)XFA1<^H`g4fqN9N0A?F6h>Ou1m5*Xr}_ia|$6j8IsH+>QNUwW<-DiD7B_Wq{>HH
zd@MOR4)QTc-tz~fTj>J~hIcSZFLxvAZ(f7p^b@n&Auq5f`9WNs1u;}wapZ9pe#)%os&CDe6vHzON<(@nVOADJ
zea(inw&G&{H*If)Y9^?KEI`m>g)vj|#*87+ZO%h;5O3c!T6<10?{lH|Hu|ShW->uc
z>MTl6d>XS?A6k+=xz4xZFUV?DVG^!V%n1~;r?dualPIiqcrMKtj6AerP9C1{qYgI&
zBH7U82yz5JAI<-uQcj0E^4c0S`066YhKd-1jQh|bc5Sm%kmp8Kv*irQ3v3IMCZDzj
zo@aSGGvmQV6w)e_v+g+iEnOzJIU}Upq4Lljd`dv5Fb7TE2yZ!asM4%63c>T*sS3q(
za*a}{wOE1BGV&*1cwF@6N}xnr+N~p+HIAc6v=?M3LMd?Ml#AC^s#T!rs1Gd-7uD9)
z6Z#~uKNDSz7W=`-^sDzoR|gOwOt+i-vM#VX{@FdzwO_d4p6GfYT?>n=Q))_9T!5i+
zP6FS%-j{TJa|)Fc%E|`QOZ-ih{qLI+CI=~#idwsV
zuvi_l)D@C@mJf;5$fv^tF(Oo6Osaka(4Gt|>9&}I%;^bp_4vZ^n;Pg3E4|v4K7Y+h
zPnHGT1#Y_!JiEjIfQdJ-n+(5w++Aw5z?&?1P<8P-*0Cw?$M<5+r;~!fh*f*EZ5Rsw
z({?1v_fANpaKGcd#uV+EF4BQTz-sg}I%fQ;2RQXo^JH^zdb&2Q1c6c7qAi_s2LMhF
z=lQ7MhC(6TflTwPnfos`Ym43N)j_#j_(KZ~oHh#)3cXDk#A+uQbU=RJB_xz?dd!x6
zM97zNp^vyD>lJ%Et1453JmpL;-fW+0+VafMwm_B`*?@Yk3*NXNg$rTEa~Um);4K}2
z&O>&m$1Vdx_
zo$*qL@-dcO>09SDn~ygATt{>9(K$0XhV2h98;2<0HwUN+U=z1-}_{M#fK8~U1i+87;!Lmm+W?Zd0>cMWq5jqUnx$TFLZRCX~@{-oG)6#Uu
zP4?qYW0<)OAxNF}%biQ@bS>cpR9c!^UL|a+)E6CtjzI=apN1*ex^A}FxtHr$Kt{snaAD4vBBhnm3*=p({
zB6~2mk57I?KYizbuV8JlW@c1jcje>UL;VfJ>Yo-GZ#NsQ@#zbzVK!UB8?TR_VaAnC
zM;key5^wt>=7`|j6cb_M}++XKOeax6wdPk9Ra2xJ$_5{%U
zbfIkV?}_BkOP1`v7=&G((uq*ry$x?%F*W}j)p13@Nak`SzQdnV1|5$VR~4tB5{te=
z-h6OJG<)~+a^f8-2j}Q1!Cg$a3>w8rK-erSq6v%%F~w+uV=cgO7T|aZ&!d4{ccV|G
z_n6!RoqO%dk;Sw3(v=0^N%PVRU4v=^;u@UhdYY6d8(p;i2*gg5Wwn
zMHGA08EYE2omWwKgsI{~=3tJ97rv0-GjrdPFbCXHTVuKivkhD8M$#j~@Xex-t)iE+
z>3&Y|kXH+Viuu-=$u^Q&03;Z{!#2NnFKM9{igARm12n98KD2Ge@RODC9qK9&znrw^
z8?tc76L;ruo%HZA251hp{YK0bY3#UeEIlV^!7CO3CTnR!kMk1eDHrF<08hKQ
zGK7nQJe`r!U7La4p0Fq!izo$W5OKQb1)AuPTl^Ok7ZrsEZ3{t03*Z$S@wF?7kDd`<
zpP-%WM!H%@Y_=xUY$k+H#b%xWk!-zR?tn=Y#gT2r4@fEU(a=o4dEo7VA(<~x_o%Bk
z^x3ZU*$_#O_L4Y-lDV&w_*oJnuOj%}B@+vL^;eT?;V}{EFHxanZbW(t-cMUk;-kD2
z-1|w-Xu*m?sY;Px!e$JqYhM`_vf?>l!B(otuzopTT9$TN(VX6$Yy=wrMUWc6O$a@%
z=1Eup*Ymjx`SN1J;nehz&=*w#=qW8edXX=Q=@Wf`FYTIS?MPIynlswnGaAt_78?@r
zCKJ)<)AdL>Da&4DWy&vsFqzOLL-a^F7`%SSnT4ZcCMjkLq$GvctBZZh`u-&R^$AX%
zH*6V=EHX1Ykr$JsH$_-6>`0fiqvb`GEoO7O)N60jlJ*=69n1={^aMPGI-yrX^q6NQ
z5-vjtc+T0eH2%e8UM33|iPvdhTfeEy%wD~mX6GE)l$@~coDMz3am8FxTgf9K(%H&xI)kU&Yy?+MJWXjC1BAf
z{IbZO1v926S07y5TJK$&Ur3cgE2A&*QIS?j2_|YwI}nnqW2eOITA+`isMnGCG)twM
zg@UjGowTt8JVMcnhB33A<@wDGBu7haUf`;)%(Px;Z1;r8wQxm|R{Ol*xxOr&riN9e
z$BQo|9>Pk@g=J%Uv>(>X>R6&zILp3P