From 2394dba47a51ae2d9badedbcc785d49b268b7d9d Mon Sep 17 00:00:00 2001 From: zhangjun Date: Sun, 7 Feb 2021 15:15:45 +0800 Subject: [PATCH] add xpu sample, doc update --- doc/BAIDU_KUNLUN_XPU_SERVING.md | 110 +++++++++++++++++- doc/BAIDU_KUNLUN_XPU_SERVING_CN.md | 10 +- python/examples/xpu/fit_a_line_xpu/README.md | 44 +++++++ .../examples/xpu/fit_a_line_xpu/README_CN.md | 51 ++++++++ .../examples/xpu/fit_a_line_xpu/benchmark.py | 57 +++++++++ .../examples/xpu/fit_a_line_xpu/get_data.sh | 2 + .../xpu/fit_a_line_xpu/local_train.py | 53 +++++++++ .../xpu/fit_a_line_xpu/test_client.py | 35 ++++++ .../test_multi_process_client.py | 42 +++++++ .../xpu/fit_a_line_xpu/test_server.py | 36 ++++++ .../examples/xpu/resnet_v2_50_xpu/README.md | 22 ++++ .../xpu/resnet_v2_50_xpu/README_CN.md | 22 ++++ .../examples/xpu/resnet_v2_50_xpu/daisy.jpg | Bin 0 -> 39696 bytes .../xpu/resnet_v2_50_xpu/localpredict.py | 31 +++++ .../xpu/resnet_v2_50_xpu/resnet50_client.py | 32 +++++ 15 files changed, 542 insertions(+), 5 deletions(-) create mode 100644 python/examples/xpu/fit_a_line_xpu/README.md create mode 100644 python/examples/xpu/fit_a_line_xpu/README_CN.md create mode 100644 python/examples/xpu/fit_a_line_xpu/benchmark.py create mode 100644 python/examples/xpu/fit_a_line_xpu/get_data.sh create mode 100644 python/examples/xpu/fit_a_line_xpu/local_train.py create mode 100644 python/examples/xpu/fit_a_line_xpu/test_client.py create mode 100644 python/examples/xpu/fit_a_line_xpu/test_multi_process_client.py create mode 100644 python/examples/xpu/fit_a_line_xpu/test_server.py create mode 100644 python/examples/xpu/resnet_v2_50_xpu/README.md create mode 100644 python/examples/xpu/resnet_v2_50_xpu/README_CN.md create mode 100644 python/examples/xpu/resnet_v2_50_xpu/daisy.jpg create mode 100644 python/examples/xpu/resnet_v2_50_xpu/localpredict.py create mode 100644 python/examples/xpu/resnet_v2_50_xpu/resnet50_client.py diff --git a/doc/BAIDU_KUNLUN_XPU_SERVING.md b/doc/BAIDU_KUNLUN_XPU_SERVING.md index 29719f99..8659e3a1 100644 --- a/doc/BAIDU_KUNLUN_XPU_SERVING.md +++ b/doc/BAIDU_KUNLUN_XPU_SERVING.md @@ -1 +1,109 @@ -Baidu Kunlun xpu \ No newline at end of file +# Paddle Serving Using Baidu Kunlun Chips +(English|[简体中文](./BAIDU_KUNLUN_XPU_SERVING_CN.md)) + +Paddle serving supports deployment using Baidu Kunlun chips. At present, the pilot support is deployed on the ARM server with Baidu Kunlun chips + (such as Phytium FT-2000+/64). We will improve + the deployment capability on various heterogeneous hardware servers in the future. + +# Compilation and installation +Refer to [compile](COMPILE.md) document to setup the compilation environment。 +## Compilatiton +* Compile the Serving Server +``` +cd Serving +mkdir -p server-build-arm && cd server-build-arm + +cmake -DPYTHON_INCLUDE_DIR=/usr/include/python3.7m/ \ + -DPYTHON_LIBRARIES=/usr/lib64/libpython3.7m.so \ + -DPYTHON_EXECUTABLE=/usr/bin/python \ + -DWITH_PYTHON=ON \ + -DWITH_LITE=ON \ + -DWITH_XPU=ON \ + -DSERVER=ON .. +make -j10 +``` +You can run `make install` to produce the target in `./output` directory. Add `-DCMAKE_INSTALL_PREFIX=./output` to specify the output path to CMake command shown above。 +* Compile the Serving Client +``` +mkdir -p client-build-arm && cd client-build-arm + +cmake -DPYTHON_INCLUDE_DIR=/usr/include/python3.7m/ \ + -DPYTHON_LIBRARIES=/usr/lib64/libpython3.7m.so \ + -DPYTHON_EXECUTABLE=/usr/bin/python \ + -DWITH_PYTHON=ON \ + -DWITH_LITE=ON \ + -DWITH_XPU=ON \ + -DCLIENT=ON .. + +make -j10 +``` +* Compile the App +``` +cd Serving +mkdir -p app-build-arm && cd app-build-arm + +cmake -DPYTHON_INCLUDE_DIR=/usr/include/python3.7m/ \ + -DPYTHON_LIBRARIES=/usr/lib64/libpython3.7m.so \ + -DPYTHON_EXECUTABLE=/usr/bin/python \ + -DWITH_PYTHON=ON \ + -DWITH_LITE=ON \ + -DWITH_XPU=ON \ + -DAPP=ON .. + +make -j10 +``` +## Install the wheel package +After the compilations stages above, the whl package will be generated in ```python/dist/``` under the specific temporary directories. +For example, after the Server Compiation step，the whl package will be produced under the server-build-arm/python/dist directory, and you can run ```pip install -u python/dist/*.whl``` to install the package。 + +# Request parameters description +In order to deploy serving + service on the arm server with Baidu Kunlun xpu chips and use the acceleration capability of Paddle-Lite，please specify the following parameters during deployment。 +|param|param description|about| +|:--|:--|:--| +|use_lite|using Paddle-Lite Engine|use the inference capability of Paddle-Lite| +|use_xpu|using Baidu Kunlun for inference|need to be used with the use_lite option| +|ir_optim|open the graph optimization|refer to[Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite)| +# Deplyment examples +## Download the model +``` +wget --no-check-certificate https://paddle-serving.bj.bcebos.com/uci_housing.tar.gz +tar -xzf uci_housing.tar.gz +``` +## Start RPC service +There are mainly three deployment methods： +* deploy on the ARM server with Baidu xpu using the acceleration capability of Paddle-Lite and xpu； +* deploy on the ARM server standalone with Paddle-Lite； +* deploy on the ARM server standalone without Paddle-Lite。 + +The first two deployment methods are recommended。 + +Start the rpc service, deploying on ARM server with Baidu Kunlun chips，and accelerate with Paddle-Lite and Baidu Kunlun xpu. +``` +python3 -m paddle_serving_server_gpu.serve --model uci_housing_model --thread 6 --port 9292 --use_lite --use_xpu --ir_optim +``` +Start the rpc service, deploying on ARM server，and accelerate with Paddle-Lite. +``` +python3 -m paddle_serving_server_gpu.serve --model uci_housing_model --thread 6 --port 9292 --use_lite --ir_optim +``` +Start the rpc service, deploying on ARM server. +``` +python3 -m paddle_serving_server_gpu.serve --model uci_housing_model --thread 6 --port 9292 +``` +## +``` +from paddle_serving_client import Client +import numpy as np +client = Client() +client.load_client_config("uci_housing_client/serving_client_conf.prototxt") +client.connect(["127.0.0.1:9292"]) +data = [0.0137, -0.1136, 0.2553, -0.0692, 0.0582, -0.0727, + -0.1583, -0.0584, 0.6283, 0.4919, 0.1856, 0.0795, -0.0332] +fetch_map = client.predict(feed={"x": np.array(data).reshape(1,13,1)}, fetch=["price"]) +print(fetch_map) +``` +Some examples are provided below, and other models can be modifed with reference to these examples。 +|sample name|sample links| +|:-----|:--| +|fit_a_line|[fit_a_line_xpu](../python/examples/xpu/fit_a_line_xpu)| +|resnet|[resnet_v2_50_xpu](../python/examples/xpu/resnet_v2_50_xpu)| diff --git a/doc/BAIDU_KUNLUN_XPU_SERVING_CN.md b/doc/BAIDU_KUNLUN_XPU_SERVING_CN.md index 484c27a4..6640533c 100644 --- a/doc/BAIDU_KUNLUN_XPU_SERVING_CN.md +++ b/doc/BAIDU_KUNLUN_XPU_SERVING_CN.md @@ -1,8 +1,10 @@ # Paddle Serving使用百度昆仑芯片部署 +(简体中文|[English](./BAIDU_KUNLUN_XPU_SERVING.md)) + Paddle Serving支持使用百度昆仑芯片进行预测部署。目前试验性支持在百度昆仑芯片和arm服务器（如飞腾 FT-2000+/64）上进行部署，后续完善对其他异构硬件服务器部署能力。 # 编译、安装 -基本环境配置可参考[该文档](COMPILE_CN.md）进行配置。 +基本环境配置可参考[该文档](COMPILE_CN.md)进行配置。 ## 编译 * 编译server部分 ``` @@ -57,7 +59,7 @@ make -j10 |:--|:--|:--| |use_lite|使用Paddle-Lite Engine|使用Paddle-Lite cpu预测能力| |use_xpu|使用Baidu Kunlun进行预测|该选项需要与use_lite配合使用| -|ir_optim|开启Paddle-Lite计算图优化|详细见[Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite)| +|ir_optim|开启Paddle-Lite计算子图优化|详细见[Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite)| # 部署使用示例 ## 下载模型 ``` @@ -99,5 +101,5 @@ print(fetch_map) 以下提供部分样例，其他模型可参照进行修改。 |示例名称|示例链接| |:-----|:--| -|fit_a_line|http://github.com| -|resnet|http://github.com| +|fit_a_line|[fit_a_line_xpu](../python/examples/xpu/fit_a_line_xpu)| +|resnet|[resnet_v2_50_xpu](../python/examples/xpu/resnet_v2_50_xpu)| diff --git a/python/examples/xpu/fit_a_line_xpu/README.md b/python/examples/xpu/fit_a_line_xpu/README.md new file mode 100644 index 00000000..15c5eecb --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/README.md @@ -0,0 +1,44 @@ +# Fit a line prediction example + +([简体中文](./README_CN.md)|English) + +## Get data + +```shell +sh get_data.sh +``` + + + +## RPC service + +### Start server + +```shell +python -m paddle_serving_server_gpu.serve --model uci_housing_model --thread 10 --port 9393 --use_lite --use_xpu --ir_optim +``` + +### Client prediction + +The `paddlepaddle` package is used in `test_client.py`, and you may need to download the corresponding package(`pip install paddlepaddle`). + +``` shell +python test_client.py uci_housing_client/serving_client_conf.prototxt +``` + + + +## HTTP service + +### Start server + +Start a web service with default web service hosting modules: +``` shell +python test_server.py +``` + +### Client prediction + +``` shell +curl -H "Content-Type:application/json" -X POST -d '{"feed":[{"x": [0.0137, -0.1136, 0.2553, -0.0692, 0.0582, -0.0727, -0.1583, -0.0584, 0.6283, 0.4919, 0.1856, 0.0795, -0.0332]}], "fetch":["price"]}' http://127.0.0.1:9292/uci/prediction +``` diff --git a/python/examples/xpu/fit_a_line_xpu/README_CN.md b/python/examples/xpu/fit_a_line_xpu/README_CN.md new file mode 100644 index 00000000..5434f70e --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/README_CN.md @@ -0,0 +1,51 @@ +# 线性回归预测服务示例 + +(简体中文|[English](./README.md)) + +## 获取数据 + +```shell +sh get_data.sh +``` + + + +## RPC服务 + +### 开启服务端 + +``` shell +python test_server.py uci_housing_model/ +``` + +也可以通过下面的一行代码开启默认RPC服务： + +```shell +python -m paddle_serving_server.serve --model uci_housing_model --thread 10 --port 9393 --use_lite --use_xpu --ir_optim +``` + +### 客户端预测 + +`test_client.py`中使用了`paddlepaddle`包，需要进行下载（`pip install paddlepaddle`）。 + +``` shell +python test_client.py uci_housing_client/serving_client_conf.prototxt +``` + + + +## HTTP服务 + +### 开启服务端 + +通过下面的一行代码开启默认web服务： + +``` shell +python test_server.py +``` + +### 客户端预测 + +``` shell +curl -H "Content-Type:application/json" -X POST -d '{"feed":[{"x": [0.0137, -0.1136, 0.2553, -0.0692, 0.0582, -0.0727, -0.1583, -0.0584, 0.6283, 0.4919, 0.1856, 0.0795, -0.0332]}], "fetch":["price"]}' http://127.0.0.1:9292/uci/prediction +``` diff --git a/python/examples/xpu/fit_a_line_xpu/benchmark.py b/python/examples/xpu/fit_a_line_xpu/benchmark.py new file mode 100644 index 00000000..0ddda2a0 --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/benchmark.py @@ -0,0 +1,57 @@ +# Copyright (c) 2020 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. +# pylint: disable=doc-string-missing + +from paddle_serving_client import Client +from paddle_serving_client.utils import MultiThreadRunner +from paddle_serving_client.utils import benchmark_args +import time +import paddle +import sys +import requests + +args = benchmark_args() + + +def single_func(idx, resource): + if args.request == "rpc": + client = Client() + client.load_client_config(args.model) + client.connect([args.endpoint]) + train_reader = paddle.batch( + paddle.reader.shuffle( + paddle.dataset.uci_housing.train(), buf_size=500), + batch_size=1) + start = time.time() + for data in train_reader(): + fetch_map = client.predict(feed={"x": data[0][0]}, fetch=["price"]) + end = time.time() + return [[end - start]] + elif args.request == "http": + train_reader = paddle.batch( + paddle.reader.shuffle( + paddle.dataset.uci_housing.train(), buf_size=500), + batch_size=1) + start = time.time() + for data in train_reader(): + r = requests.post( + 'http://{}/uci/prediction'.format(args.endpoint), + data={"x": data[0]}) + end = time.time() + return [[end - start]] + + +multi_thread_runner = MultiThreadRunner() +result = multi_thread_runner.run(single_func, args.thread, {}) +print(result) diff --git a/python/examples/xpu/fit_a_line_xpu/get_data.sh b/python/examples/xpu/fit_a_line_xpu/get_data.sh new file mode 100644 index 00000000..84a3966a --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/get_data.sh @@ -0,0 +1,2 @@ +wget --no-check-certificate https://paddle-serving.bj.bcebos.com/uci_housing.tar.gz +tar -xzf uci_housing.tar.gz diff --git a/python/examples/xpu/fit_a_line_xpu/local_train.py b/python/examples/xpu/fit_a_line_xpu/local_train.py new file mode 100644 index 00000000..3e0f8880 --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/local_train.py @@ -0,0 +1,53 @@ +# Copyright (c) 2020 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. +# pylint: disable=doc-string-missing + +import sys +import paddle +import paddle.fluid as fluid +paddle.enable_static() +train_reader = paddle.batch( + paddle.reader.shuffle( + paddle.dataset.uci_housing.train(), buf_size=500), + batch_size=16) + +test_reader = paddle.batch( + paddle.reader.shuffle( + paddle.dataset.uci_housing.test(), buf_size=500), + batch_size=16) + +x = fluid.data(name='x', shape=[None, 13], dtype='float32') +y = fluid.data(name='y', shape=[None, 1], dtype='float32') + +y_predict = fluid.layers.fc(input=x, size=1, act=None) +cost = fluid.layers.square_error_cost(input=y_predict, label=y) +avg_loss = fluid.layers.mean(cost) +sgd_optimizer = fluid.optimizer.SGD(learning_rate=0.01) +sgd_optimizer.minimize(avg_loss) + +place = fluid.CPUPlace() +feeder = fluid.DataFeeder(place=place, feed_list=[x, y]) +exe = fluid.Executor(place) +exe.run(fluid.default_startup_program()) + +import paddle_serving_client.io as serving_io + +for pass_id in range(30): + for data_train in train_reader(): + avg_loss_value, = exe.run(fluid.default_main_program(), + feed=feeder.feed(data_train), + fetch_list=[avg_loss]) + +serving_io.save_model("uci_housing_model", "uci_housing_client", {"x": x}, + {"price": y_predict}, fluid.default_main_program()) diff --git a/python/examples/xpu/fit_a_line_xpu/test_client.py b/python/examples/xpu/fit_a_line_xpu/test_client.py new file mode 100644 index 00000000..c480e81c --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/test_client.py @@ -0,0 +1,35 @@ +# Copyright (c) 2020 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. +# pylint: disable=doc-string-missing + +from paddle_serving_client import Client +import sys +import numpy as np + +client = Client() +client.load_client_config(sys.argv[1]) +client.connect(["127.0.0.1:9393"]) + +import paddle +test_reader = paddle.batch( + paddle.reader.shuffle( + paddle.dataset.uci_housing.test(), buf_size=500), + batch_size=1) + +for data in test_reader(): + new_data = np.zeros((1, 1, 13)).astype("float32") + new_data[0] = data[0][0] + fetch_map = client.predict( + feed={"x": new_data}, fetch=["price"], batch=True) + print(fetch_map) diff --git a/python/examples/xpu/fit_a_line_xpu/test_multi_process_client.py b/python/examples/xpu/fit_a_line_xpu/test_multi_process_client.py new file mode 100644 index 00000000..e6120266 --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/test_multi_process_client.py @@ -0,0 +1,42 @@ +# Copyright (c) 2020 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. + +from paddle_serving_client import Client +from paddle_serving_client.utils import MultiThreadRunner +import paddle +import numpy as np + + +def single_func(idx, resource): + client = Client() + client.load_client_config( + "./uci_housing_client/serving_client_conf.prototxt") + client.connect(["127.0.0.1:9293", "127.0.0.1:9292"]) + x = [ + 0.0137, -0.1136, 0.2553, -0.0692, 0.0582, -0.0727, -0.1583, -0.0584, + 0.6283, 0.4919, 0.1856, 0.0795, -0.0332 + ] + x = np.array(x) + for i in range(1000): + fetch_map = client.predict(feed={"x": x}, fetch=["price"]) + if fetch_map is None: + return [[None]] + return [[0]] + + +multi_thread_runner = MultiThreadRunner() +thread_num = 4 +result = multi_thread_runner.run(single_func, thread_num, {}) +if None in result[0]: + exit(1) diff --git a/python/examples/xpu/fit_a_line_xpu/test_server.py b/python/examples/xpu/fit_a_line_xpu/test_server.py new file mode 100644 index 00000000..43a690f2 --- /dev/null +++ b/python/examples/xpu/fit_a_line_xpu/test_server.py @@ -0,0 +1,36 @@ +# Copyright (c) 2020 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. +# pylint: disable=doc-string-missing + +from paddle_serving_server_gpu.web_service import WebService +import numpy as np + + +class UciService(WebService): + def preprocess(self, feed=[], fetch=[]): + feed_batch = [] + is_batch = True + new_data = np.zeros((len(feed), 1, 13)).astype("float32") + for i, ins in enumerate(feed): + nums = np.array(ins["x"]).reshape(1, 1, 13) + new_data[i] = nums + feed = {"x": new_data} + return feed, fetch, is_batch + + +uci_service = UciService(name="uci") +uci_service.load_model_config("uci_housing_model") +uci_service.prepare_server(workdir="workdir", port=9393, use_lite=True, use_xpu=True, ir_optim=True) +uci_service.run_rpc_service() +uci_service.run_web_service() diff --git a/python/examples/xpu/resnet_v2_50_xpu/README.md b/python/examples/xpu/resnet_v2_50_xpu/README.md new file mode 100644 index 00000000..28a2d601 --- /dev/null +++ b/python/examples/xpu/resnet_v2_50_xpu/README.md @@ -0,0 +1,22 @@ +# Image Classification + +## Get Model + +``` +python -m paddle_serving_app.package --get_model resnet_v2_50_imagenet +tar -xzvf resnet_v2_50_imagenet.tar.gz +``` + +## RPC Service + +### Start Service + +``` +python -m paddle_serving_server_gpu.serve --model resnet_v2_50_imagenet_model --port 9393 --use_lite --use_xpu --ir_optim +``` + +### Client Prediction + +``` +python resnet50_v2_client.py +``` diff --git a/python/examples/xpu/resnet_v2_50_xpu/README_CN.md b/python/examples/xpu/resnet_v2_50_xpu/README_CN.md new file mode 100644 index 00000000..d08eb0e4 --- /dev/null +++ b/python/examples/xpu/resnet_v2_50_xpu/README_CN.md @@ -0,0 +1,22 @@ +# 图像分类 + +## 获取模型 + +``` +python -m paddle_serving_app.package --get_model resnet_v2_50_imagenet +tar -xzvf resnet_v2_50_imagenet.tar.gz +``` + +## RPC 服务 + +### 启动服务端 + +``` +python -m paddle_serving_server_gpu.serve --model resnet_v2_50_imagenet_model --port 9393 --use_lite --use_xpu --ir_optim +``` + +### 客户端预测 + +``` +python resnet50_v2_client.py +``` diff --git a/python/examples/xpu/resnet_v2_50_xpu/daisy.jpg b/python/examples/xpu/resnet_v2_50_xpu/daisy.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7edeca63e5f32e68550ef720d81f59df58a8eabc GIT binary patch literal 39696 zcmb4qRa9I}(Cy$u&>=vO!3US%FgU^8T>`;mHi?MmxWj zgyCK9G#`)@BD2NFuj3JRE+88YFSwA?t$P+k`Ai1J{}|@~3BtsBt`qe zQWKgje+xzTlNXedwxj@_CrJ@Xq5DZm0Rds$Nud!|aA1!VEsz$pMQ07{Ba=_sWB^(M z2<@c{Wq?c{f#3VMiGa^-82V5#LHQA=Z~iO&q&zHu-+`8N*8CRqyQxHAGI?koD*v?g zvnX^+Oyb^o7TPeG6jV)FP{i~wKjv_s7%godzcd93U_Uj6LYkN;Oi>w#fx-dVV)0b7 zhuQZ*DlPdvEF<#ZL@OaDQK=wHGASrAsJ&Ot2!e!`tGpuMy{fq%U4xKv@XhHjogis0qv_xssmL8TSlL7XqpbVq( zbH{l|L2-aR(qxK8cz_V#_Z|;)OI+G?jHEv9kBnOapXg{&#b`r3o(W=v5GG+OS^`2C zWPlKGEI4_KP+5%-pb!Cy2>Y*3hzxqUV9o?Xl=0ijq=^ zUc3wY53{PKJP6%i&7L(1S{x6@;_qOJ&}8XTQ~@V}Gc8%mSY4gkulB4!n6rRC6kXLd~SjY>_Z*qiLke% zkVbvBG73N$7@~|q8^RzLkrz^jZtpGs3@M%zlpi`01&WZSCIW}cgDfeWfcDc2QY_r5 zL~k@tLgdgL=vW{iX)+~cDIXaXY!nuz29Pu-EhGB(T}@>S2zn(g_!ICP4lRxdPgYV> zqGUuT^TmRPk^KhJat?|2J==pboPh3`;T{FlVeT+spDRWZ0}3uJD&4T34A2r6LTd?# zd-lCyfDG`C0fly#Hl7T4y~hs)kjp4*lF4Igl5Pnrggiq`@*q@rc1-l}q?&k&XY1mk zqw=(AIy(Kw6AJqq3Rp-T5jWwp&)`c(fbx=dds0CRKno}V(DK>mJqY*XpKS_!HmFz; zl$!yZ1pNMN9CWEdApY=s?PLbEo;pArEuoZ;9C{u+%n)C$5SxyLR+Ju*`r_G>ap>XC z4o&%}V~j7?JW;b>eY#+)5l+j@>LUS$H`FP#Gf*_L;9;;Z$qE4kF}hDnEqD?r+v#>*{xDadqNCU=CG{UQd8y zsZG%kL=>}xe2phFPV3zHx>!8zAi=cl8JOTzEF`AQEc*nRn5^dUk#XmCD{J)wf&8(4 z21+g*$3ek9os6PyEqRHyT-<8PX1+K&_;sNDDL7TDB;3Smi_ zij06LBGNY~&z}`LkOx~HbR>`Z27oO^kGHOpUy##Qz!u<_%Reio0Cpz#TvG>NR~6Ou zrMLkTP?XOn`NG=YzIX!EUj*?g4g5}wAF|)fPqJzE^!#{mC}^U8PqCTkZ6jK_+gddg zztf?(vzFvvITln!m`-X`X7I61<#qVjH)I|C{JQYx$K0b0%95?oKZ6%7#g!t6UsC71 z5_#3)fwiU_BwzDewN+HKC&oSSty&oD2sA4ley6>?kqCaV5>L|L$=GWYVP$PiOHxp7 zX~ohbO(8vO!9Gd7d4E9Mxjp5AD613>3df6&uJ!MkLJ9<)v@ic`S=oX74%lg!0qt}m zG3~w#xQGj`<%G|Dm7mt0SuM*iCu{LT^Y2G{LO+C#L_P@I_@^a;Pm zx|iH2sjy=IdC2PU$y>xQCave6=6cs{U;a+F&7wLsYSHk@PtT9`S)*v z&j7)8?3i~+>WG_F_*P}9pLGCLi-XwO)^Laj{;_pVk6(T~V_H+jRx8?o-&~;U#=OhF zwm7C;%YRYVD=VIOEawd!YA5#}Ta-zsyz*w%Pdof_AaT?K2@3o52>06kN&{ zY-W>z1@tX4`50mBP_kzR0rFwk`KTyl(zJPf+%Xg=!+uhl}sVRbUN?5 z2bsJOONcxQYhjWlJ1a|{^qW5Z5bZFO;hJKT!ix%eTfUSLS<0%{z-d9HfyuS-=7Ojr zmvHMjxk*Q6xvif(+%-po>*-xl5j`Z{@u$`<_=EaUtYoPBl_}0gv@O)uGNsu218W|t zF)P_b3CFZy%09`%EIQgkZ0@PPs>A02B_j!m)^q?$0HL~w)QcBg`>KBZK)fXRSH(Hb zq=YCG*f_L(07i5_MWdGym2?q=q*94M^yl>qk{9(nSBC$$9HLkYMlQS$F`}#97HQ9( zn17H!*|JLtiimpzNP3U-^q3VsesQGSdAm9FSx&U)dMs?}(!)uEDV~NnWloN-`2{Att=J+G){>h;AU+L|Br!C9lVnwp@RfQdXMHx8?lZ2SZj5=?j9oNN|72;i6Ihwkl4bjDtzlLak}qR=Xgm<} zkiX8x2U^}rGvEdAwSDgzL-RcYi)heaWAQr)R#&#yW* zrYuvDIP2_^P$<=X*<9s$!3N5ZX9Si6;nm0ZOiC;3)ip69&n9R=;O}-UuX)`}OR6Th z8VGT%G_1gjgs{%f#vwuR2k&C|IB40waOE|YXt;2cvSRqUjPOO7`B?y8yVBWg^B6{e z{%0lvO@9vNrgqvkm{SE}ZNfQK7j3a$!r?^;>#q%;Tz~)5=#siWMt2eiX zf5=7$}8#tLzQ+*qY%J}Nq3A;);mVkDXUElYZr z^;ZS_8u!s^P-!BURad5-0O>J8J4o?dbkFCwU5ApIyY5c~#5HEX^Ojxr=c=ELi5KVz z;4qG=sl&x@i;!vJ*5#RSlshw$S55iU54JZpr>G>J5vX3`IKi9Sb5IBNzMx##3meR8 z=hy8Aaq-;;Z~m3LXo&`CZoqUh%=1w&IH*nWa+Ba@sDfW(5bFi z#tWZ@=$nqRJnuGgNDq0jrJp7{YTMZux)DR~*t2jTT6y9tHR;OQ53|nwBoRxlKU7BF z`};s^9m8b%02UTLYyCNSuY)!%!q7a`zdY$VF6@y82$F&dpMxm!=P(ZxlJurej1drr zj+RXO93)^nyJ|YKS~CqISDk%vCKTFD?917i27`o+N<-IfuzlK{GnDU)$y8(;d7GG^ zYoW~K>~m}Ey*D`|MHU75$dYm|HS5cv>`?`6^@q!P@0H(3Q;+4U{EL6xoac$W1~8$Q zCOf=y=f7u|@Ll=$dOqCj*^9@Re~7r3N{H1q%tgoQEmN{*$QIHDhGMxI zVP#Cnm5-^YxObn${xTpbNmpKh%ZW*~DH+dr;qT&yx5)|Ea4v+2lq>C&M(H10CksQX zselacNO>BPt~Cf|&bf8q$|E?RA(lO*$t{k9?u*t(675X8!Vql3i9(B^vJs)de@amGX#zHu0_8ha3}n*eadKOBg)$EsIOqpu0PKE61uM5W4*n zkD-G7#2YvC>-U2<`xA-B{k$^oVQqHzVk>F_0(U!FINRB3RMZ&_jpH-x>JC^ikFB~R zb^pE629IUYmjl*G0Cy;}bNk#M_BT%ecANRO-3+Gp)NI5{m7h4!8gWCzhf8dHHG$yd z*={X5>>A3hgMvwZ;01V!GVrAGH!fQnLv@}gICSqupK%$OFEJqg561K5j>*v+$9>8x zpeMtFo7iE^98++k4KI>j7N0&l*F2VFC9PS7b1<~uphHK6lxI9WJGe}1K87}(i+oI%W8uGu*x^kDXu!kIN(9{1YUbL3KrneLA%D?w8Q~U9T#8ScpYAhCX zI(9_uL3}JnQdiPdoZ8@ZIj1YORVzDlAbRw;=vsAyXPve+=PMe?&eLvIqA?d)w^a>l zI8S@szUqMJ+j@AOVd1&QzBjvh^`pHChoA0WMowJ1#<$np8(bduH=OR z-zTUB(>{;>%``rY_V`ga$APvQ))@_PNlS=7#kcBpLiiKD2Vu$u!PryrsL$kNM zKdbSfyIkgf?HoFi?GjacMRlS^fHy$KA9|zjwLrE56T#H$1gqguVhGD)*+kmZ+3vE( z$_44~3$)Cd(1BF3s+b7}$rqVFX&k zN8sCLXiZ8+7Za+j3y6^s2qDx`2E3@C>NAA9-$=?j(V&dfPkkiO6KM#!RISNX|I1ls z7FJcZx2&V%T4}LcYFOv6FvbHNq|io~WgoNgE6YGJj4EH2h=Cu1j8BZ7#+b@42+MkUrR8c9Mh@2KB8Y0*!>7=mN()a}MT8jw1L*C<2&bQYt3vL7K@AAM( z%$Y-6Bdw+$1gqGE(OE?T0|24>;@;gMF4H-~VuH9kBEHLINf0yTX*0xwUM}jJjBS{k z$47-3a+`W#_TttmqTC1-Bm**JWYN+fMsc-K_ znHpNz0KPOU63_F#h~ovb2Z~~%gdI(RqeI6Ff7{^RcrIn6=NBc)FZG|KS*(Rg@sTLU z9kt&YIjgrt8)bZ*(>z+NnhGTAu0N2N8}+XVz*^h4V_)V`e--6_KWuKAJSo}&-?^YM zudo|1$Y9N)*8kQhRkPKWtfA97cx$FS;cn^VYwAcYM$QLE{l5CTt~Z`&*N1JtS@mF) z{ZN1nv@*G_^cG})RTlg(nShh3<8FeO-WC%u)^WFjPGmzRysYIyhHQ5KQO}_G^ARb@ce8pu@xA;f0O36{uS1~6zgus_@o#|< zglWtdammAr^C47y zbB5sE@=+Q~RX%kfBj=WGcLoQNqPS|KHaVk@(&I}Rn{N4Z zfx?_Q$33T=e3N1UM$T{uJ10z!QKz9o+2@W$_hgWyn9rXry0{9D2=A`7g=%Rs>|1K_5+YZ$fix-T(S^S;zjCVvY#j^a(HtmA%V zk`lSXpv;|-Pfawe`vyRmzCc&R3>a2p)FN6aZu}_F&Hs}ua~VSaDl)BTD;UsyVL{L| z{h>cRT!&|x^gY4bg0$WrlhqIQrVb4=TFrL)sP1W>b#TIT(Ur%y7e8n6Gv_zn+D6EnflK050@vv?2WC3Fg6YO)A>65T(Poene)-+9^Nwn2cPn@iKq3RA;r`>(xz_~<@eze50 zXhP|M#rn9_tthzXAnv41)cU*cZtd%@vpH*fl>A1=aw=pN38pTo%^|RW5@rim@DBwq zt#R<0hj?r>WM^ylG`E0Rh+}vaJZncj1?e<5K=qF!p0-elRxPq_ATBK;k|i*;D4^Y0 ztd(Msq3ZT}FlmYNRBdeE6Q!QQM){PHtT8KjYiB+8Il^nzn?aO5q(2bnBg!|VXS42X z&P@l;PLQ(#eeQ;N?lK?uJgoyI!0fV*L(w`r1EP3B)#j4w=8gxk6H9dfd%~e7pOX(g zzI1`r7nDS@efC;M)6g@0Ffk?NFzI(RvV-Ja4{$^oe5sl7yq4qU7ww<*OM^ETHb3e; ze4P>Z$H>^KFDUi()HrE~mw+ria6HZ>TWgzsqSnVua8;lKVB$*?!XU%`LKewiWZkY5 z7+^B(lBQZZ1hEEjiAOZ|FTBE0lF>pikE@05_m;+<09>HZ>nA{YhS^&Ur~I;(`R@L( z96pW&aSF`fLoP#eBB*b(I)lR>F6zjK?rkLCkxbI_#`I&{m5!SMB7vb(&-9pYM{m@W ze7qr;6lD<&#ZA~?X=3shj-%^k+cH{|38B5$*>BpDjq0DsvVJ9H9Y5%*mAGD2FU(U& z_Ju;&))<}Jubffggjo}aZUe8WL+H3?Zr+uA?b}wg}(-zh8NR+d~;2x4r zggi7t-4!+6T`9RwEx*+!sU~EbSoM$gn2(k@e|!3tfZ6_y< z4=`^;soxS|n$p7wXVhkVRRlc>7hHvi=r(Pt%M=f@461$C=n(=}kjuvjP^Zmg2LmWm zElYd)cv7Pw0HP2U%q?LG#M0lqOwlSV45RI?3%%-9VNCJ^YFcKBOcZw-Q%fyCwV(dw z59)1Em*svPRy+J8D+!tU?l&6K(!pjPt3PmvAtm*f0oB0m1X4L1N=h>Bg$4_hRcwo+ z{emGr@PZm~d=M``G|^$0HZH6v8Vlbk_F(OVyEF0_-BnO+Oaok)POy*7#pTM^UH-gB zA-_?FYe?pgNDRuT&p(JLWiKpU5@nN060<1(v+@ANZ8$@#j$`V$Coud&^R;QlERRrs zc8K4yPd*xNT#%LT@%k5C?X)ty49KAwT59u6BrGlX(WRjtE1)YMi>*V$M_sI7kM zU34pPKsJAC3_c|Ju3E1}qs>ilcPV;!9hgWcAipDr(eEI5V`j^uaUOC286RbPSwUtN zyZ2EeK<}ZkrX!xQtzXUDLWl84H$=R)VuE_0YtkoLVAA!*JySV~ZPg|4o|$6Pa53eJ zQdTqi9sSb$GTz3ENYU7w{afk2a6SQ0U3{IcSZ@h}XZ*{ocQt3Br~0=8)p7>oD z$@@vEg~rx1ojR0zw8T#Fb&79A@q;V^G%>4{fAHnXzWVT$^>7k=V|l2gDThOJPIPYY zT>Wv=@b~mt%d(AOe@?y^rDUgE>yq$N#`nn2W@gn&gd=z@VwqRyIFa~JjDPRmlM<|~ z&vf0tRqDAB^qVI-Dy!=4x3UciJ;-}d*nK4rx7E^8(xSbrZ5Qgc{H<%sXqO;TUJ;&E z9uT8CGwkbCM_qJ>3xEQM=a$OTPpo5C)|1_*(1S5Z&bPer8e5b++?O$4df# z@9QtMpnOMP-F0>u`>3^AU3*Wo6g7}gX5@_I7}*OW83YaJgJ@N@MR7O4qrRilts|_@ z2?7gTlp0{2hd_`5(z>^Abu5p?B4|`RYRCgYg6FkQ8rOoG?WL#)OVpiL6Vub3>E)Bz zPL51)AV&vY+Ea6`mSp**TGB1;FU0Pi1n}U^7`wAwde3)^DC<+by!n_kdd=-Zf#mc2}xt~ zniECBsHtGM@Cd&4DL=h~d`;6$g$z}Z{SvG5+?^TTSJg_>EM+dx#cI3aXj6EoSBN>^ zlBE{~B|BzH5xbVu4`Y9hd$G8GD)WEuW)AoG&Q(JGnP4sJF2K@E$pb0xxYjsN7<@`+ zm#vb(4Ho%t;qViki*u*Ka1{k{)_+z!$Klk{X0Chvgiz$I^oHEWPS`n;%hQUJ`U`2K z@$mAp@&sixOL2l86Ew{oKl09OLU_biAIvL5$!*l+b!Dd)pW(G*D?Vp}ofa9L0Jr8+fb4b@DGDJ0H;L=HhidsKqAkNFWHQY5{Vdl>FIA)pZO&0NKwvVr? za=x=o{XtT%#y1}3=KOyntq%3=@)`KOJlll)Eq3#`A5)Bob>t5u-$6y|4*rM?QS@(9 zn_&5xsH9dur(7hXb8ii^3`at-&^`MSWNS=G1>3e*|M&wpHktEJD!2WuuG`^5HrLUj znf3K;jSegoa^Bp*A2QdoGeyy3GaK5ing89quftaHl8`9>8>b~Y$tTM=eFpucSwEYJ6L)t0MedB2M!UmgD~y8 zTBTA~u`d1N@OZo>p#Ntv@)r@zVd=IKX(z>*uB+XHN#Ft*>1NLYVSUT z`}@D#vZ(6TLzHJ8(p>>3%@V4g00phZv7#=N46B;tqE~E)5kRy+eRfLqk zB`n>tgf^(^VS``i51Qeoa3Q?0_tO8}(WA1ftQww^zcWCg=CZ-Otywl!%Xr zQ4TMPJC`sXTGlmMo9%&Te%-W^k%^es+zEM5q9oGeJ?77-i-U%MYL>vm(g$M`bC}_2 zQR(jwwl{9;#;znv z%;a0An^eS_NSm2Q#%nIW%&EIIvOS_$6-2KR<%+hmxXi{^NnieHyvQ%SS)FK6iEn{^ zP}jdGjZ6tt5`e@PGsq_)qfk+i#T8g+!uq)7_Qy4qwjB&Z=;`$(L=UuYWWDjkbqeH@ zYt^da(+#YZj#i|}i~{!>zdBr2c~xWRu93~mUK%*%kQ=*?d=rKzt~>H#j^@KTp8&*0 zREA)~*`#{+^YNDTNH1bXi_N-mr-sXo>w@AsW{S%y4g66|uhN;74SXf67jN|mO890F z^nWvt;pXaxMe*aJ5!u(H0Sx|sNWRXar-D$joV0e@^a=4sXL(`F%9_AIC?kMq7)G{a zS9hIQpYW=AnL|@UjWFF|X|d~_*d!!)EaZF-?n>LCvzqVpF}Y`0VasCQpt#i0vSH?% zJhnwlejarQE(Z;d50A0!Moe+X6h=k%$rGR?>42}|d3UE38`JP67hzcSFhL_1=Ca8E zvAAMIS`OJEI}90@8Xy#gQ&|a|kD7!FV4+p2K2ujb9~BgNhCr-LYhQ&Be+ZGT#}HTV zH)O2>qeN{N-v&I)qzGA95eUE<7bUI>jS>)p4F279;D8P4lNK;*kKcIO3!9PG53vj` zpX8fzq^50V@TuI1x6|B}G$vo9<1(k+gR)EB1*MOMipjQdsXLrhRf|dLxy{`)jJ@)S z9~x&{J0|=2b3}FlJA?6rgd@;IOo791xDO%aMCFr9n_#3bD2X|_J0J;2&c`d-FN!rj=POE0d|=EGJk zMCsByt@JN7nl>kPx7lLr_Z!|RlvEfWsQ}(x@q9e@=!q4su&h_j-T*+fXpN2j91hd?3&ceo&~3_Sg=< zG2?fu@{g>W9Q8@L&4F#S3hQoUPIng@)=BVdh!bVwCtqr_T7h~QhSyGrZS`ayKhlO5 zg|90i>VT~3V6dSOT1BFqfXFe2QN%2IKI7mEm4<4#?JS>%5bXCM$+wd5wE}W9tt;VD z{aCCq*2>cLDIb&Md3(cge*F71+v40(7x?ORTwHcev7Q-Q+-zVRIfP~(GlRLNx}==E4IbR` z<517kWu?F?D$d;~)o3(0Pjgy(&SUDdEaxuz;bmybG9_i9XM1u8nP|$q0-{v@7hcE+ z>u&Gt!7}MN#_*n;s@KU_ozh=pXv-bTIMS=`MQrDq1TdQE13WtlOW+^-Fx={Io6vWr zGl2`^qg=58=-*2|;n(6}G%W^TdVS9+*RfZF=5x+#oZH&9eO=pLeF*|X=AYQN6vCi2Ck`BK{=)=!aqEJD-^p3hcACoYh#`_K6r1cw$``5 zEhP?Is)RXA0%^m$eEMdQT4rNjtPs2euH5YdQUohwK;nNXwsB-6t=XPTI}fss73tJEOCB#FJnXGlaZ25LWC$j zv(T1_*)MRj*B&8NW>T3bhc?t-MmnxBQ%of@RAlsInA28Pa-zB=48mS~Rx_hg;a7?)ns9WE*6}fH`76kG z2_?>d_>v||yYiYyBwRwauiL!PoC#MEF73P}=iFx=;O^{gXqpiib!`2TH=T++F5U zSNY`AlGG;v$MZsP=^yYRvb@cBv#n3D--ZU~3D7+9&hc(+IyJ$D}SOTb=+f+Sa|wqjD74j7l4h;s<^u>YmCCG@H7MIf$(%JhWHsPPj^)W6~UAY^nH<0kzSJ=g03$vW3jW>>=x?`ywpQgNgWQLpw zDh8wD0n)>Knc>^N7EPSip*D5b4k}^NxdRl9q(@@yr5PqJjQjD|&A{-jK*SDpC93h} z{+ZA|>IZ2^c=4&*`(`C!T&~q%BXF9H*Q~Zi@-h2TknYRvU4YWj)V$(_w^&Qdk&}LhFP#%Y&z{enXfmkiiaJEx#i9|{L4N|?0*7GE3fn0jmdQutV3o# z$D5lmG{=Eje#`l_k%p_ODkYs92qE)+tJWQe)!aA7niLo6T>rWpt!XyiDsi&&!%vE9 zp7y=NdvDzG0saK|kRmcHtinfcbYhzwIH8Dg;M4G5CaoPXa`-(_$p^f3Nd>=7=t$4D zOaPhdjP263^elHyf8lUl0=O!vY3PBeb=s$x@NM{4{2kTH!;7;A)czga{v}WJ0hPX& zY0z=r_{REgo={dFS{|GpQA0Z|XFBJy-ncXi{ zJ)99dir)G0)OI!^E18+xgK%A+*_81|CG9eoS>0?I$}h+3(d&R(wQ~0d5d{yOs?6Vv zMl-PXjc~CEw?f{XKcegb!PEuO=(RJ{zdQmy4nge7%8uEx#hAdsRyC#@D+<^m^X9AB zTH6%JcV-NuBchK6{x%${0D*XbmooQ&dj|jnZgOUKYCt#@wu1 zkAeoju3C+(Y%Ud}SoIc*pQB1|mI+{Je zyWjNW3p>4@{o{!IP522{yh9;q5T8^hI?q|v0^r_G*0HSQTG!Xq%EI5Q7(RUqA5y z?LN>wA9{+be%L~lc|NaY!I+FqXL^n#);<5``*CeVg>InM_cwKUp)&p47<+uOdxmV< zGTUx3^^~;n&!gA6s#VRZ%M5;#L$E;V(d~pL`;MWV3_*EsHt6|sg3JU@OJVj;d9U>&V-y*1c5khh_J=hbEw#AO(yxGJMXyh)^8GS^!c zt)pD-6fT6{m!Y9+%&q=rY9tV?s?@ENHTYa%U|pK=yw0+mA$+NHZz^&Lrb5s74jPOF z2qcci;{KI<{L-9(T5yUsOj?l)92lJc^t(4n9Y|Ih($f z&~JTD^F5)$le<}Oc0`L^y(F%6d_VY${=4cU3>J>~T2ebj=wIAU}Jgu^Z)A z%~8f*zOku>8-%hrjP#pL8D5`gga0l~gCisq&{{@T^VOtOk`7vUKNBQjH-zmUt>qT* z&>`s9CE;fK9K$2GH(J)8K_X23kVzj3Df_jYuGofAON8wldFImEf*?v);i|X?G7enW zE5^Fk^6UPFpHPpdi!8ov(wEjwa!Yvu*D^FgM7~Puk~*h6$}ywQ{2)|z{qq4CmgKdL ziGjL&dglquwEz=8e=V<>@h@FnqL<<9g6z@&H1)v8aVj&dpk^_;V^;jsb8t62n|dts zK9@*)gVM;kJj_@mW8V?_Zp?CN?*s;Y?Y@>60chf%_aVHTT2(V#kh>!byu9L_&sDq2BZuZNxjC%r%`>j2Mg=&bG$LB!c(?r@x0hi#TtyyB*s5suKJtH8-K-9KE+FpShf zsVBgIuVmg~DchNcu5imM@r|r+@=jE&V=qH26Nd0YWP%8zV>FO!~$i`C~d`Yoz>kkpG^9o=;|DqpVP)V z=etF=4wPgky`_)0oU$;1cswf^W2itOir7-LMUf_h4{_?k3;P7wV20@Jq~k{@yKpID zy6XwBtkmN+>{e$##xQlcwC3xpyBp=|mOnVS5FZ6PrV_fP#Z<^bm3agEUbnME7d}5w zvY|U1_3%i{dX_NB0AM;(6gA;Uz_Q4KmyFGGyi0h_|Bi zZpbthFQdiM=5@dx`*O8qG7MT&`ORwh)wc#l7D+8zfyT6NQgem zGc7yav}=)@gyC)^>|JZ6;vS#&NcCAqwJZpXu>Wl3S_v9)K17&)|1M-iY4v`tvB<;3 zi4U=fKgy{AnS&_`F5B2SmlpIgye)+{@}J;S$Ly$rNxjOn{W%90VnsxB@Oi5k{06yG z?<}C`OL1XwRv|)g9hgpNC)_e3+eXRRP@`vXus-CoC%gn^0o9uqI7z&)loL$ZO~+x$ zOLG&{*-MffuKNeh08RKy4&-d&yqB;2yo| zwDZzxLW6ah_PTe!!<1{=OGAVskUuVnOYR z@PWBQ&cgJ%F11*J#`QkG514Jr`Ynzz%h#%w7bo%rhKS|=vH}ZarcbRO^}i8|wMy3~(+$S0I@ep3Ie1RU zuR*KU6h(FxaS$6tG(HK4Tgj@d@&QB+J54tVK&i>lD-&a1L}6jGBCV5K6&rdGjK%uI<4WRDTHNr%UU^OUx(yv7RL@5Nt9@#|f8k@PLd$q`o)q{mEAAb`VO z)=z+c8WORE(UlD@<>|4VC)G{EXj=(L{PVZjo1%c+CxGL_^q34mxMS#V;<+aPE?1J1 z{Xucjmm!N*wf9M*T6+(H6q4H8rh)LXpM~vFU8zHqnY>_moB#(p!9R`3BUCE=whn_b z%~TD4Ua(faQd*?~#j-A^@O|=X4KIaZm(0nbF(a#xFNeoFx|mXur0t~{Zln+d}KrbKLuPIoVD;Xbq2c&6^VexHTji^nG2 z=rq}XNTc-jk4Oxb7Cwm7W3Q?++1S=sf=F$xWK+N-3BFD!Bs7T{!nJquCmj-$%&BTA zaet#ti|elRodgt*xIY18@4CI*{VF(e$UbBqNLtK(X_~xe&M*Bq(`HD%t+@X;ypOWA zwPzn;-CC0FS0+22wR}!3y#UP!@LEbEEjhFc?c#BP_FP%FOCK)sdnxH}05m;zcJ2>@ z;Ux(HUoP5Y0>zVd%d?{k?4a-Rpx~3ndr#gxndD;xfFqAzj&kf4r zq+liF6xhnM^adLiIYw&k4+wU|>pLK!o$W}W)@r`b31u_dOCsE$y*~J(`HexQJ~Zcu zIr&)lq@bdpv{l`a%`D?f2pA!XfnPG%GyAyE+-DC?H7?;6AQ8k&fu4F}g*`T{8yb~A zfCYxUPWP&~7&}SC1m3U)l8D|F$hO$M$*;z@*LN<($6;`;Mi+Q1qd)V#tZBB_l76@Y zG(sEIQkqbOR;m`epQApC((8teB76WSFRaEM(nnp7g<>wy?{GWezZf*#{qoD*Rngcn znp`>lOjR@)Ubb$M^{$mA)1T6$bW;?_`KdIkVoO^3PVQ22!7isEzxXxKGQhnjW+qXY zn00EuGl@HOS=(*iEju#FZIr?o0nVY<`8cwhC^ebf zdLqPLYuNeedmkm^cC|9{ibrnO?v~x%WvfCUoqL7dTa4Bl6t?r*DE(GZ;bUdS&x@;| zvG7T+2dH1AYk8nvP}SLzS>n9WY#`ruKp&5@7wzxlp4YM&$>$CEWd|v!NJ&nsdu_SW zzblJt%e`AS3VZ~Q#m4)8tT{H$beF`_jTk&l&E9Sc!7iWegfRfOq2=uG0fJp|Ug0*gxya?O3da2_* zY#AxN#2UhUZ;U`q&Bkh>S(X;Ifb&hsF{%y{Dv7m*)xR_@D9ascZgV8Jo-8vS=MSF( z{1#tmyJxHuc{e&bj99#tvF;fWC-wh&0bX&m+8P+##75~EQmuaVH}+OXxYslHL-<=c z=BR1iF{*^mOn&5I<4fXGJ_@BH%QEqcpFEk%k;U-p(4JdwJd1?F_kD<))d4=8wQ&!1 zQ-4V`D*iMxy|?~N)6M@T;DLUFaC3?SKCw<8(=d zcy7_6jAkOZE&p=jE0ymd{I-GvpWY!d&~c5Q!hXyee~{bQKa!|=uRq(sOvUBT$JN&e z>CNW;RLox?f$_;2(gBDkfX0K!=4eY9g-xKNlvOschvM_dh8)}1$GV;Kf8sSGPJ;&Mqx=}K_}y>5(`{KKen{JXyN=Vz=3@IUM9|DHakdIn3*am z%BgDb=Yuw-?MHjVJAdjU3Y+=bF_eQyQdg!6}GnS=CyZ?i_z z1(cq{l=43}oSah$8^30Q`xC$Z=_dX8AU3ABWV}H(xbE8fL#lt=Yt%g+37VD|am;92U7rt%X2Q>lZKS8DN0hJaoo*21lA?Qq%hEBus_ zmr9KzUewC&nDWGD9s13R{p#xU_J(Fq_}}8gH5b%zH%^^OjOmTcQHcfWYl93A^qmYZ ztsyP9ikD_=4yCJ1?o|B6mG}%EmkMHR$IC7r%2xw5ZRt&@l;GEy8PL{Yx{Z$q88+^C z-Y#(pjNb*f)O33W>lFhtHE5-m9%gNDVHqDb2-b+Shp3+bQ#Y3VB)FHD#(q&+1}wu~ z@V`}7??2SF@U62l*u|@keF5Y47sp}Qtos}^d#9$|*y(@6So!hO3$AEQMEH(E6jji7 zte_Ja>$eGa!qa`*zwMKcn<( z)S`Xb&EELdBEC=muT7zP3S42l+JSMVF2F^ASR1>su{V*z_q3uUc+9OoTTYFWE-b*| z@C(XJkY#%Nd^P&6+w+CM55SeQCV9h~$H~Caxj7w`CxC5aGSBCzUu$RDDa4B}lxxd# zO30(H#fbmf)k{0Qj zun`s_zy^RNprUw}obcCp5v5Zx6j5RjfEK~`7hgh&NN;v%Vl$`TOSROikRz_Kb1j9W zDV=JSTEmw}XDc2YqJnE=-iJeR5e@%^NEE}1DLy$GXN|>pKM6!Dz$Zu{qxHJYhah&n zD5ucmO(SMY09SpHjDf~KEOpnox+s;%RU=FwDRNgojZu2n_4V}|8sJ;M*%=MKy^cN? zrJDEr8!>n3PaW$wxBLDL6JBQP*z8O*IN48(9t*e3E%ZRx!x)qn( z=G^cXTrJ=D$H#Fkul5Y_&a*VH;5zac{0l+GDoy>$*@G3R;z%|sm~ezXb54U-AM?5r znfX$K>;ghHB-M!}t)$`>nmRYl6K4Uyj01E$_PlC20m~3yO>E8%^DAx(UXNZgLeA&< zmMvkPro3R0YTr@lU719DH@NWVVC_Abz1$)~bI2R<7BeZ@lmVFV4fFzqjHM9(vLVFO zqGxHAqT9nQItxP~e#7|)qB8&3@goMP|Nf=tY`^x*aweF-J^T-L;1l4aaIEvCv6^K} zFmC*><;Sir=KD@LHAn)MPrFHQ{<=g3j{(DB`MU=4mC>dlq!xNjV_~M)t(TeF1cX|J zaZpS^PE7vXDc7mA&??7qiMq<%BW-k31X(tS)iHVh4*-Ece!oZ=F2(^LH6o{n>iAZwDQtIX z>T7CU;mj@$E@QyS$Zs)^M8^?V+G;G4P}*qZe}(*~$rMLSBx6x=>PF{n)N922J4^gN z8%^R7b9|A<6-STaIjwfid$XJjXJV~a891m#JT~){pZ@@U5iD@AyRXSA#ZSfl`En1q zTc&H=B|72KTIURT26q@8Ra@zJ$^1Wbwzi&@qLL{e7reJ2<)?RHgRbNSWyCh>i)HH5 z4F?d^$6Iia)6DEQc)%xS;{$QA^-R{>>-aYFYO&B(d&_SxmClm0CGkoC$sDx<*!I~z zw&c{JqT5+Z82fvF$dlAx?fid#^eEb`wy144J8acA2EgWlM=YK$w!34cr||^r^V40f zU&O3(w%=O;bW+I3NYqazx;|u%c}&5>&h~A5imH%4U(_D+(OS8?j9X9&($6eOmoN`RZWs-`@Pc1WQn&b zK=;87M!zGNM<;h6l0IQcqqkY6q-CJ{%@yV+1`q*cu)K3Q zT&2ujGyV1KTvt%vT^zRBsu*jd4GuLC9aq1HHKKICBby$GgBmYJf2uQ_X?EX7|O@b zv;P316>OjX0AI)c(PK|pT~p<$iLEYkL2x>T+_C7j)@~h!x+gwL$%LnKalKapom!W5MG(=>x8 zCqr7@Upxgi8x1@^1LgX-sZ6hd!y&}|H%_*A-^s)7y#_F^*yiOq`-}^7NBr47=|X}o zrN`z|`rG-N_d{Dp;vaP)j7Dx_ZKQ>~m=Dc3033!sNhYEu^O32S*TUXf+@h}2Xdj*T`6GCpTu`&ZmFh@NseTe zVt?}-eUnBPv|v9Yu>mpJ@I6+&#a!=oQv%4(g3ir5pP5vQk=l+`l(4+>U_PLxeauUc z)Aa0^n|+cbmqQMe2J;zJHQMFaa#zi^U~k?5VBPh7E~5ED>X|cYa@buM+YigPRUh~$+DMBUgLlfmqTL7uXdK~oAC)x z`@4%bOqjI02v>p99*PqwRur~x6y%eHnke~fv=nh*1luOi_(wgFTV`V!vqpQSC85UH zT3F)CcT7}MH+_Pw>aj=erEpwm8GQilfyMF2l$k6lhNR`{;Pns-l<^v zsuS^*KC+AO@)jk^$>?Lpd^kV;UXXw52!|i=&NKXCm;V60dG9099zu`&Dj)eWPw@x; z0Ey^2emddmsr(vyoji{{1F`$7nCl~$0~^}m=vSEV4iEmqv!T4)t=KfR^)Y$}{(kNOZBs@krME z&KI1d;=U)i$RQVM`DBpctclnz`2e*RFJ|@kBHFGbuBj|L`>>hBpZw0w-yoy2NUK@0 zNe{05?+`~{ZO(zuEp22DaOT#$5vni?B@nw&vJOAS0=a$`Wx%=0NT zS2&$487n$gVenjyc`qt&kV4-mE-O7>PyY3jPgzddib1)7!z_2tx_*w+J#>N75*oC$ zv~kl$BV#}Vvae+><%#r6ryBWPoC^N{!#4{r5z)hRr;ON{tiuDfNeds0vn?F~>B%{9 zs=MD^NlU}kKf|s1Tu3}NG$J#n@dd8(m?t|!Mc1DfH3vh!aPUhBAU+^H?J z61WMT%6f*&gBwea;UnrX^;}<<@iL;wO1wPvo|otH&CLd}T$-P;WmQ>K!;}zk^|Ie3 z&b&b9I%#BRV?ie}JmV$Ao`ZD?lFMeK;b&Xh8y$9SKB?n-Jnxaj`Hyy!$oePk*Je@N zpHUR?R(tG`Q~BI{M(hYKINNe|!Yi#kb-wkl{vK&=K*-?i{{X`_A&~OkqkLnpHHj4@ z_H_R2{I&D=M&w|VO|rX&p|V;yYO0p%=G9+dW22~Ji@bKqcV?cr<$^X@&BpjUtC5>lxvcw3mxR~8KN%D;7`VUN{)-+Vk{eCFi;8Kl zQB>4Yx-$flOX4y{JcgDWa_5}%!i~n&RnoqK`FEph?X-llc;f~~L>@=;H|Pk)0+DI9 z*`%@m0K4^ctu?VzcvTR1WI4c$2Lsio?i4;0sS!S28opgmg--nx&YA7gQy1mW;&XXm zsB6Sil38$X_i=w1{`ttC4|DJ59Wu%xTLiDWaD*+CcsZ$xM*jdRq;nd15bSN9T#?e< zto&DNmPsV2j`G(H^tw6eMS*}`Ule3xJbMl@_ zW35)}%yiYZ3!Fl!gtfJF?h-d2mF?tr=F&6I7*WYb?r)E2FX=Du{{Yc7nPvC=`;Jmx zu9k_vhu@hQbxTZfTniqmnKhXl70>thE&jNtR-ozp4b zD-wiKZ@x1Z`Q&ArVHV^&K`qOPX=45(ZVYq}8HXUZ#A(^_yJB0o!P#`a8M{wL+*Cs( zgBpCtf4^0l;RtRMT6{~L18Js)S3>a84xllKG5KAckD&{DVzxn36Wt>+KMR3f-;MG9 zhCMz%9>*`N)T!l2wua$BINuCCp4=${2xJ+J!ynRe6=c?#;ocW`t?WkXHm5}FIhq_h zY`1qc_aiTEe%4(YLOD1em_%t`(@9l~tC75>kbKU6bn|W%Bb2Oo{-*x`gykrMY>_C_ z%0B@6rpywP{+p6pc$vQx3~>IJVJ)|)zve?LKTwt;{l)cs2{iL`iOD9Ga(~iN1>%tI zOUKQn7`ssa0R4s!oV zi0EO8AsJ|R9b7k2s;sp0w1#;d)2LPj?Z)nJ-8TA|m@-~Io1=sz*@VAZC8o%82qbNd z!TBs5Z3&iRskk8XU2^?dO;t_d2A4b>hhrL6BI82W<=|n#8t1j(Q;U)*D5x6X(Tw3; zK|scI37SY8(g59CK~DS(g)HPUF(Js!Ba*JD$lNQcn2Rxxs%tqr2-FFo%XLuCnQ18J z1P$`A2C7J3HWksRI$Mx+y*psEc~Jwgg!)3P5_#F39w zZDyhZnCer@3Ce0hB-~|oF2#eJ^qxe+MUV~?kD|kQD~ei7v}_cYTge&H6*+2oE-f;_ zlA0e3l+w`FS|)$O!lJUvBS!2wQ^xqU%PUV$B={=uj1_h)l4;ywM5?wv5J&|TbX&k3 zGD{>J`XdD-f;ogh9P?8VWO=GoapWk`4b_BgRqmDxG{+)5J(E=fOK_#&0hJDOt11}F zES6Eh2#k`YTP>AWdf4cw;xbJk*balKK8cYTdxKii=M#VjGFNo=alx$N`)H}CV@-76 zI!At9%^ra*P3oxU+!RvhFTmFa`#gfd17wgj^-dmnNXhvQ!D=0Ef-p<6eyVfa0Qo5T z49?#=S3P9h-sy79m@TUdqP^Cbs?U@bH@beF%}Y^Apl6zz1y+2ikw*zu{4^@fcKV{bI*p^eP#UQ%`3>SZ8&TD)W zd#Po2j*YD^m6}U?IK>shp47;15s9noswvp=(;yPn2aFOk%THc_j470LmDO?6@hnhB z;|e!1un8PIxg`0{{{Rr?v=Ms~lHwz54>X<>o*UlmkL90Q?BxXbpS} zacPPDH|AP;9=w#Q`)#7@LHs+wiSN?cLrZsu2jV@w2vyYeOl@2%Q!ZQW=3FLK!N@%aA$_-H$zqlWTY z$$y^E?|0${YoKcw=2pcSBsB8zfb4eEKB_F1+FBc%uXYN?QaY&Xm%;PJAjib;-H!bF z@0AkWZMfa(8BZ>lq?xXUSx!%g*9;uA?UHf0!k=g1TK*oXjnO#7V>+~r+J0k^%Cik8 z4ac{`$4~wWGSx*sxc-bPy87G1bnJB-JSMn!@O)9u-mMuK0Q2Uy_P!0Frl*Rk*a)VO z2QcGyj>pt;`YTI4rr&*N5kkspyk*em0Q$GT?zHwxT~`lC=DH13@_cMNkC;6_y5I5M zGtaHQ#7Jv68o zw%fcRxI-1Rw9I$JGlfxGNC7%puB3SGm~?GwLUO7?t`swkZ15>?x2TiGT`d=5&xj!80N zE2!MlgUM$ujC5^rZb4lRoFW0jA%(K*GywM|sEzDTTDMm*O_-Mb*9PivjhxUxO6 zg0fCac(;D9qK{bnNyjA#p>DzED8!t8q3D~*PK~=Luneh5tN4J?$p6snxv+fc8Ia({Kqs<%+@`5&z!}|qY*sEKaKF24u5`CU<1w~C% z7kh+yAZMB=oHu#=#B^POtqo`ckl7+8HZ}$|xC3C6do0Utm9pI;?RBUuAr5Z0C6T0T zhAfT;LG_lvv454oCzrFff&n&`IPG1AwB{&y(OV`Hw$Do5BKFV@-_KRGDA! zw`PSq;T@7YyjikMrOL~0O72_*iBGT8(+~usX zv6yoP4?VlBO%;q?sC;I%j(ZN0-k_}mI9`>meq7?y$aYSD3>LB~DKnArf_@;ShK?}j z91WJO%cseamxt!1c+5{}8-fSxwI*irCHuDjz51(JE)j>3MF$c47d5+$S zs^?{Evn&nuE&l)tWlqJIGADNaCW#bv)K$kcvuKx@joTl$_*2^V0f!H&$9PVPK;{=Q z?btcH5%ox^zVQXLc(l<&b%=aa(w`1eKcVAdeF*xgAw0Z3#!A0~Zr2A4f{kLV{?s^n zD@ZZCt+iTIi}-D+G*lWMSmyoP+WqSb1M^Azbu6KqCir8cIX) zQ_Scc=rYsuY5v$xHb$ODbi9Nz4pCJt#dOynNa5eX^go-fj0#ea@~K zh}p~s4r%pi+WxBmc8{^*M(sYT6TE9m&og@KZW;QO^3 zche&^y`$B<&&(g9X5Z~vQ-ND-W4FO{5@Uapzsio;#vgOwOOv7&kCI7axY>e%8O+kTTh7&ka_9bD;;!!mg|I+?36vN!36W) zp;lL3Jfjf48+df`5rH9YmN(aZ%EwgP-#o6CX-twk;&J9vXC5*>50c@PBdJPtNbaab zU$5$r>xf|7{;H&>g2ZB9I=D9nQUdhBvz!)IqUVvbzg0zX;sXP>cj~EKso04UOS7Q5 z?#DcN!jlckGt7m_X?T`OnaFFig0(aia>xd9a+MdkjczZadj(r0cFNM7*#NACx}ded zyFXQ;g0N@`*Q%Fd^xH3E(r}JeHO*BlxQ3o#K5k4c#-N@0hzs=c6Il~^X z9m-Z%;wVKnH9g6(o!k(oy)Zi(g2M3y+FD|CxT+d1B$?BIU)Q>z^>T8;MV*mdoz9Br z*MM@ZBCBhLa=FEp>SSYl%Rv>?Gy|*kTC_q^NwS@?NWQMC4Q?{BcWZ$m%^b>?RdQq4 zNpj9yD%$5~R(ma-HC*y$YAskRE=MkDJo2J61myDguiv6b8!d&zOsC9;VmMD30B-6q zb`l>h_*Cq9WxC5NZ49?tgcNanQo9+)!a9}TtX8P47U*dqXbodPT%*}V54~QcsdnNP z29KFtox25@s-M-j;JZY4+#eu=dF3|jzbCSc%D)ik1P29C%aX#0DoH@tB5O;zrrV*L zx!{zS)wh#U$gnVa&-&Y(>)c23NuByEUd1BaiO>)5H2Fp6jgCJ>ZNH63EWSeY8 zLS=U9lAW%oezzPjZs!%HqA{9w7>-$uk8~rJ!zHzFJma#0_ZezH9Tdq90*OjBO81_n zI$e#AM6QX!wh}YG=~+%gSRb|%5_58p?unT>pHPH@5y<%jZrQ;i2W6?p45GRN>;Vi6 zW1;|~$|FjehRJQ8#fv64a~=H^j1$LG7<&)Qc1oHi&lzKe#b<6UxKIbWgtXPi=8cc| zCjNu-<(`KMG_d~w1ODatEvv0;s0H!il759y&rc*SERm#Ry0+oIC86S4Gd(FRmNTMy zc{Y*<^=pUs>a1!hqN;A08V_UHD=O#m7+Zs#@dpY}aSJmuOf^jyWlUgd9*6S}L-be) zKi(b&tP!%NdQju!k_VFvqtG3*=tfs)th7x}m&N?X4#RM&YiMb%RL`ZWg_A)40Ev$u zsO)`Inte>Edt|bKcAZ+cLE)L3H?``xw6g6*9YcD}PIN!#r0jl}T?)s=Hp}#St@VtR zz>F6oxc>mo2ksY1rnoXl*&`m= zqIk<*(~w7SdaT=Oolk7hdb!5Mn3MMEvs^{M_UrDc<7lL;Jf@6n^ZiZh^X9ZHk*^tP z1Qco+MM+_gM;W+UD=f7>?OhIsb7RKrkaub8`-K_p1TM*Pw^H3{UrSdIk~a^lPjGr4 zRm<=8DeN_#1c$_h`7uq8jCMVUOjV??fcP_&KyGxmKOn0eT3F%$10904otu?MR}Ps9 zy`qYIGNsaMlv&bMNh@=xsVsBccO_TboEUI@Wn*ex!OkRbT(-x z89$;sb6Q+7wYKVZmbBn3S=6#>Mtm5`t{9c$$Xuu{Tn$%I@VMuu%e2Qx$B%?18)RgB zAwfMXWHG$jAstsANV8shpZ+p3(s!n8>EKnJ4G zMN&%!>ZOqQZOU6}p23Ffv@6rBaEjBt`;+RbNp3Y^Rm}4DbILc!`93Ek=~p2QsHr)* z4TkQI3Bb%EdvFt^!4xqgv7bdBUQT#7i@xNNWk{E5JxqWtBY1-(bK%cf$#=0vp2IP9R4I(!QW+k z>0I|=o1&VQ8Cm?`3aKTM#9q@T+pM_rJd@cJmuH{Hg+8*w6tWXfkfS$-i%z-c4QSk` z@X2=~7rj1bV=i%09NK$y2Cm~XV1>Jq!Wmx>DV0~TDbDQn@;lwo_ zEo`}H_#$t>&!rYTFnh4Ly>vHy})Wg*QsgG3UB+Q{~5W zrVKDD53=5P3$9Xu-5~Nt*KUx7xi=u>Q;s~7QhX^fAu$~;!jsKI55vb|r7-FVL{bBG z7)FTf?4v~u3>3v(D{5JTvxmZaCxSkyHIXNSk@1Z+cM0(fJV9-7h0`4q05a0?`G-*9 z`h~mUE;6FxnqMX$tc(v9oyT$N-tF~MxPr+A{@Fbg+jxz@em)b>{R-i7GP=QZa1WB2 zwvs~KOgJqgx1s$(S89s7y|3;?Yb5KI?`C=#K(4h#QVZGqxqFUB=Cj-v$FT7OTSr}@ z3yVh~vnKU#ucFtgmRExC0BPUHO8fd2qUo}>Ct=(7M7(86`vbKg_5*O@q%pRbQg1K(p5PqQ>D+p(O;(mvJg#DEQK*f}o*B?~J05DX{b-K$ZI+SX zYsRRYe+$PW{l2R|Z}x*mD#aaSgH8sThWt6~f6CMQ{4z$_qm_?g{6WJ6D#~e+;NsP6 zbaECwrH^SXb5Afv#44JmhdXtl;(C~8zE?{FgQIk$f;k@KeO7vM3)QVV#EN#~pi+<= zK~ltlpNT?@V4*1>Ee5wD2+L`IYpG~{a<&$`i!jlsE^9?pBRDGhhOY8SS zR;G@dO;*AI$Wxg!Tseh+rMjK4{!0Fi5(=F!qw-GO601MDQ~gT2mlCZOQ7S+2|P#fEqHD zc54`Q7hnU?OSH=y%Um}Kjd72umQLjGXlZG(z}#$^tETXdKvwixGy};*x@p1Ed96Gw z6|(YEI;TOvK~~huqEa$~k8q5*nc#CM8KigODH>6RmEGYSp5{8o5(cEblTGZEs|u`_ zB=LL({{V!j`c#n8{eqKEX_j-hB`hSOxfv+AJ__)};Af(}i-(<$Hd6lp-N%8bWe2E- z$W=1XoQnp>Nt!aOWROO77Zt&G(={cpgo$+011Hm89 zXYRcVN^?%kc3iiJsHyDMYU{Ju0i}q_5A|dL`jye_Rcx%a)6BhgE1eNd?)RZ_FXYoaSf@LyFFC{v6)zO6Zl>?Y5U%z?EB%N0r3N!UBHy?U-`EG~uaW5Ew;Ak)p`1fe-S7~s57;0R3_G^#qS3=Y|tm`Xqc#(Fw zDJ-KacD5RAQM?HtkQRbQayu2t@3xm(E)Ilb!tXpp{$Afj-)f*`R4$3*EkB~>d~<4Z zXy2%L>7A?|?c`LNJ6$&8*^Sj!t*<`lz=nb_lkht`6z+^9$kn@Q%vv%lcw?UNq>LOX z%&e8MX=KthgkZIZ| zC+4sv(lmT!J4++M*~jj!M^2g-tZD0NbhjX?@m5G9Ruy$j@@dAht0k$Gj@jF+{Y`jf zm|MXn?D-_mpUM)fw+{|husdZ*Syxdnkn%#?DTZG$>(yf1oF3#{)60CY&ZSRt9uUD% zYil0yol2g9xvpTzPE^!@)4HBFOtCb=WBDez!ow$wf6L`00M z3OfZ{(g5i27euz%m|6*pZ|JR}v_}}(gY2x?t!Z#W4JvOe$k1B-)q$nB!kwf6uve8f z`~!fhr>Asefx7e0;w@F{?MG3ha8OAROL}7|)>FKU=~5`@-s6`7tc5lQCC=!)J=&$} zSlKn2g%^ZUIj&*kRJu3}j>!sRsWNkf*CkjZOnGjLS%6DkmL4h z=N9|@Ws>O)Wfabd>_0ZWqnImU>g7(UKV`!+B=ek}!fU~JM=Y9FH-LDWZYbgSZe}A9 zbBi=PxaYrReJx}BO%G{m=^16rw)4&aP~TIA`5JN5m0eqO5wNv{ z6W^jOhiHx?zvwa2)4OunA(pp6j98pRSfCcWSLo`|Ol{ z3GJ(76Aj}MIR?KPeF_i%08z=8KN<8Y&N(Zh)-#QdE{RVW371Kx5dO*5yIVmUf}~81 zuW&Wl8=`V7>352w`k?d zW>v<;wnR~~P%b!pUtfsk(}cykLYYuUj@I_!N>Aa2JANIiOktP7}YrnHVwS2`z{T23Iewp378`cBt+|K}dfPPO^v$Y^_-fTmdP$BGbuS zS1BY-8n9PM5Mjqu%VCky3=Nd(fCl|i>2aOaYpg`Dr!0(V9U$D4nJ7qSlx1JV0CPbA zCq2wKatV>Ov`Fbl$(l!F;pV1NTV1!8D*2YZZpm&!v8A=(j5P?$>dP0VWu&*iJGCY{ zCq`D_+CgsZ3cj+|V;>5)3x-OgCs_-dp31TD7J^;+txZL}#BLQ`RSSzq#ua;#lN_>) z?AX*u#>y1)crt=Is|wm@!d=-?)x5hT@(Z8VO9`UZo>-ZGL2<(Y&m`?t*UNx%O3qTu zbdB(ye1*==c3CmC?KSM5(~;>UbdD}3bQDu5$_furnfXViSzx5Vl6czp4z3G3#+IR3 zNZ4rL214L62PLMdsdSxi5-22!dV&^cf$i99%C&cG};67h2>W-u|?J#eo|$=b?rj4MO8`JjR#;9HfQq`NWa zQ*n@0EyC+lbCW>Sb!-|pC0}sQ231r@*{X8J9rX8BBM8DDQzLV7m(qI8tN#Yct=gtG>X!jVe|4I)h}O=trO zaE-WC^xC?q?RLp_1~oK6!9jf+T>ZkIETO#AN7S=C!mVzaHg<_&V~PZH&1Xcj7$`DU zyQ)^Eld(z^Mk|CVSlH2tMyPIdaPHrKlgXQsF!FX0?7YNb0pur==k^E=Fcb)@&;21&tLUwJKdn9dsCD?sIe zuuC;kxNhRQ_*Ln_MSBq`CRyAr_yiM`pQyN(lY&y`)c*j9S5JnX$(FoCp0)J_B|&gG zT^oj~J?7T~v-n0=DX6Fo>bAbmF3p~e^saZB7ZHy`xLnqUQMbrAKldYRSw?hPQrsO-KYqK6w36u7phw#`tWKP0ARm$2I zPxhSliv%qvcT_h6?zxEOwDfEIulY0eO?rZ{l&m63YjUh7hC{{z5tkJvDpGgoou_~S zlBY!w4djgN!W1gW2-E3mVet@gg=k|14|fNmgVrVViumGiX~<1KX+iZ?PLD_-R@M;1 zmCt!+^zJOv3_KGAm}bfqMPpcSND3VH4-K0Ym%H8N*W__jBdx`GMwBW{&{bVn=N2`< zkfl*h=?Du04RnTL{GCfjYP!2v2cJ@^y;Piu<61{SqaVuwSjP~@E!8!uq&x!S&1mTK z$&HqsQpzEk6-!PD0}A4vk)RENjT1viI97DTU@7+uNOC$hka>v4vKHa3z)tc- zkfv?E$kN9^cIg@c-y*&dNe9Aq?4-+nsiwhcI8D1tj)PvKJ)LLxqXxml?z3ETd=y+YbEwsV*c{P= z)!Tn{kcW&hv)*#3J*06FpM>UPWXFpmTb%ReR$isJRVVOUgw&r0AE~L-*5rU3=Q#fW zWv2FDOe(mZnd9N6jnbc3+v3fNjFNAC?xMu21Y1Hinfu zV81`pk+#jG&a+FC%&E;73ci)7pks$n!Z?)Pn5ptmOejnSoTCGK&$G8J(_a zXeB|j@g&?ka}_PiqCQy~I*#i4s*}VAWtzF#kC?PBdt0G<)wS^?ZRs^>#n(QWqP}}$ z3}cyXr!q+4jdE<(pYFNy_`dNy+MeSzQ{|Eg1dL(XpJdHNx#8!UOjiE@HtRm?^)|n) zn!0DDqBw+mI^`#|DoXoQl`^!)RMDldleu^JS-WGTshQKtb0m%~arG*K-$d5q%C#2g zbOGqKYL%Dm^?i(seL9qtUgr0NZ(i)tMNsb{{{T#$o{L^39P+tjQwC$vT~YBQ(J(RR z!~<-eVQtqgFk?@%cInxHf{@ujgocsLWq;l{3=Waf2Dp9`u~pTablQ4HGMS{zc{uVO zs%Sz|cu0;@j80?zVFeQgRnGc5KxKy%Sj+$kJo z7#l5PX^u$7#WN+eJ>cPBX)A6!S)&=GjAQ9cF4%MCRI7Z3DTWSJmCB34aRhVpS^3-U z^9Qx%yQ_1fX%oMbq}i&qF}sDbw7)G2l36arg`ko_#!BfHTZ3I2aC$2md3yy*cP6Ym z02L}4fC8?LVD^FtN)O*TQ`wwkd^;|qsZX+YyQ-iyI(nemvJb|x=c|++8ZQ2yddf6< z!@8$ig+Ue3py0VpisIJ}Ye!Fdn#oq$oH%8Stp%qlmg!E{L`FSTlWLmas;oM~Ad*uXxI_vdkTw!<8q!M%r05IC5OA&3dv$hcBeb z$w*;wV8Ox(`D!&t;yGmm#aWu)CvmpYS5^WWPWCsUcFM+=gX-p&T}4J4%r(6}s_K$h z=^_mTmkyX#H1{UIVgnDVev9~DNmtcoWX^`~DQnKo30AscLjz;fP6sknx79|`yTDbl zO6doc;X|IDN3`L~tb z;H<^+vPMSjC0tKcj;k|!t80Pkw5fSzJ2*x{=JMyc&UQ+wWM%HzO09dPo!RcB&{3CQ zk=1O*dtznEJERpQ0dUTcg>ALk4$DPLL<4yvs-+$o1S+=bOEu7Q$>|)~nlaU7eW+BL zdP{W(z%Fr+{JAcdJqtv|*yL9n_Mchf@jKxMEv1Mz9=%o`y8Y1G^2;^TvAG#GPT8I} zadu0w*>;a+w8n>p;Bgo;z0&3)8<{qi&-bUqgGO0dF(6OS+$JNyqGNUp1Pqo-1bxz1* zqSPcg?%}~gs-SDZ*<3`>=JH1+I3tp!WHZ@r6TyPHTLfny1y*U=$1Roem^Ft1W2#?= zk;!ITQcr@W8F!Gxx!3s z1OgP8n~sWSAw>!if-;g>gULyWnt~3ELn#v|j5X=jKRlC8ZV|`LM)@ANBo8urHtF3m z)b@4?wBv%rWH@lKQ%8_k({V?a&A3VPOeysCfkv?se_%A@|WrtkkgO#1XMtP6@6`vwWG#8H`f3xzmHX=sZ7i z@-#St2W5$_xA;yk0~y^*vsKFm2CRUo*#`-5O-%kyzeifq&5Dw0$A7ZM+U_C1dACbygcPc26?r{*H3dTfAZiM(R!b|4c`SJ4 zlw9+uVZU{St*2( zyNKYeXv32iN@%Xsb5hFCCx~sf>m6>Ps1rOhAh78;?u$`qOLAvPGB#HC39%a*?6eT8 z>@v#GBB>uDRrHqP_Uny7w38#0m24Pmw)ra-%-1(!3eL$4&>h-RTUR7|YQlC}8?`e2 zX95XS)zSh=Pw@$HjO9eDoV;u>Rb54U(KS-q#{J|p;1uAm!*QumO;F}J+2)|jb#y0k z6-sE>$t}4wXlpJ3tvigWI-=@;a7O9@xw@tJJe7qzONQ%KsbJ5S)K|9F9Q{NqYDzI4 zHVN3{1EPFkp;(pg0*xsSvNnO)N36h%rm8DMK<=t5uMG?%Q2VMkY7$(Q>!rSk)6Bs+ z1nj)&#P#OB_i(D!w6hRC>$6?1{sdJs>e@J7@6&J9a2ko)emwb$oJz@!zO0@)Le$4% z!D+&jKgDl$`ZhQI^HkQ6o|yr9qF7qz9Uy7x@?LrNJxba5!inT%l(D*F>40(iki7)j zmg56;si2nD*;QA%IT)ImqkJYjEGe!!(@q#$u(u^y)q#QKL&d@75lV5%GD*slX5uaG zCEX0#hUH6p#+64)IcGAYzOw_Gt!*WW?IT`Z9UGTEu6`;=TKmOl@ZGxSS1RJlq2}hP z%T4adC4U9YB$`fD-6W;pZ55zwov$U7$01YJ7i8gGnDw1whC-`Dh)8%PJ|sn zI#W!1Vo!o0gsCIhIT6D^LBUp(i)j_|M0zAOm^f3M(b+N?mt`?bP9EX4n+=gThJlp( zK*DNv3?@o>;0uA5Ow*mz8mUS8AyD%f=Ag}5!VXUBCtDw-lE%#NONygO8>uE;YbyJh zj#HjW(wgMg$DDK&;Xxw+?vo3+rNblVX=9?KAGevE=7PpvJ^ws*9JysTypfRBO@>XzM zh#1N-vKx6ULdIO(D+Ue@eyb66u6T2URn1jP-Vban{aXx#>KW*fIf>eOA(K39m>-1>>e!c$=EDwU)8}aZ`it)09&i37H<7a>11eTa0lIP zJ>8}pV|7go}rt(RKv)AdYL zR=~h>g(JXl(Szb@*HB2@H08$&Hf|D5?aAD&9~VA4r?fN!zIIi1s|7`LfX31r-wB&j zDca?`2!7iiR|TP_Ex-zGvxD?783`&)RiU)a^59`CGI$~ZkT7LKtGKo%?on09TalHY zx=idlbp_LEEtcwvlfxt4xPV62VapR;#O^W?tt$TjJHU1cSh=Nnoa!f<<}RG?n^Jru zq0-S5i`%u_$z-@cUp6*8pXxv0E@@Mc@Bj|jH}18Uhv>02m>tfYkd$nB_~ zuvhdI?1RC9)IB z&i6H&b591O%*@79_yRMxSDy{r^y<0Q_0DX;{f<{lxy9r;>K8oXDsn@)s4rhYQW&AH z*Q8}_tg`BBS2VYWi3dOm=vFdxv$+br8iw8&=L26%H2{@U4#(9uMC^d3Iiby0ayJDe zl!rOOjw39S(OteCgeWsXf=E;~4-B3#${nNu%~PQRQm&;Xyw1vWM1hoQlW0mjli2=2 zJF6i(xZa!3iCfrN%`a*j_n=$bTy+2J9Zg%{wYyyPUiB^IuVyg6ZIeOq+` zsH-6%#{p?&k5*MhTwCQE>QjSEfxJAKy2j}TAvHx*T1zz@)ccDAn0u-{UKFc<)xp(s z!J!A3myeKmV2-K!Q{+@{xM>Y&-CEF1;xH{bE|AYnGF0x%FsZ884n{XB>Qh3=VCu}7rJ7oqnYkONGvOMq8OjCnOyX~da!F_m`OrZbUQ>SmZ@ zXlG<~6^?cMb6 z7(1??X0pXi6F^`C3af-{jisiH7L4O0`mKXNiHl~q@s>I(9gWDwLmxl_%U-J*;$x3@ z9?H)1aq!w&-h6;nRFo!1c#`f)bh#ypDBWwGe4Jw{B}WRreCv4&Y*`dTpinC7po==3 z1{5=`HHJ$WHPOxQKF=8}FL0UQ<87BNy>Qg_%A&`OK%JZyez;!fUmKg_HzT6wTtVSi zP6U2!`Gu3J(8D>0W}cT@DDcij#QBI(Q*>xABh=*o0ARC2%`WGgPwG|o7)W1ysSS6Q zAX+i$vZS{{`Kw}K^A|mg-3lYJvnN$QQr9Q$&e_6{d$pd95w@7de!HziZomOi+VnN7 zjxm7H0LjVZt}Nf~4oTo-mY?7}V6J zwQ5%^qO`jf0ynoUAuX-h8E(pO6n7Z%FtOLF4--2r-E?`*%7(s>JA5H7Rs@VCJ+QJ~ zT~@Ny35~7vLMw3bbER0ao5It8s7f0pl0@=a(VPIRnTIO6Q{@;|4J0Y!JgkRk?0|Mv z^WbMHc{vIVM?4i;=aOh5EN9H7vA{GaTQuY+87@XLlqG}>ypq~-REBg=nJ&rZOb$0p zo#_Rl!Zw1DD8|7E9tu><&gq_}<-ryR`P~p6WTG^t%e=};;?f+)Oj&MKU1d4kt7b?G zcUASWw|>d;o)E^eW$q2q(~-$#tE3`biq_nuYr}vm9a;&Sy5W4DJ!0)q`a}n1z^m&_ z-~bk6z07tAS9YfXn;B&0>H`}UcT0y9Hd4bi9$ybR=Aa6OMCapl;1U>4dvaBUkJfX` zD#fW~5pNo}O{^TM`Zoh9VNz>iaXWHRBV(g}q0LvF85%xl086do}h$O zxuZM!tEnlP$TVSP&Zil3(P}GlId%St$=|Et>NxCXxV4$EZT7R z#^;y;S&Z>s*U<#3d$%I$hWb1(3FJGXt-KjyaBH!$qBJwa%|Wi;Zb|A}b#(jqP6N5> zs`@eQGQ?*timLPDNa9ZUPf@wi#s)~rdbZP3PW?||!VyL^VOnxqa5ib@B6FR|C?)6v zq6ws2oyg)Q-qu7YhY1Q5Dsv%3@O-YaEp$7TcDx~_wR4cg+y=^{3q zxR2R(E*-Kmf|-mA4GkL(%f0YcJw<+^=;C!cH>a_w7IYNinh=p)S)rr@x|J!=(gK0* zhIi_JCDFfxM7~PcsUKH2i1D{ft073%Gm??vwif)9I*RtQV19`u@;f9OjVtA85u7Z= z#@|xVay2c4a`?RVQR|~&Va^b1Ax!1B7(mnr!Sz)wjgjU;Fc=HIfM@$HM-sr@dYO)U ztZn|FQQ52J&NRU6e9qt9RkdwSL-!m$>r?MFwNqzzKl`aXDFLayQs;j;dHO1E6qz!4pcAl@qXgNxW zBn*`@o?N|CQ^}|e)ujw%xD~5RC7w#zFR4yI^-`H0QxvhB1wE1ZeOA3} zb}}PQ0&UL&Wb}g#%7rmg?i$d=Q-LZB(1*7n9c1P_l*Dz7AmpL|qo*4ADKk@z8*@>N z=UB%@Ij37jReOmdhN(QHw1qqklxX3@vZu!pJk?lCmuTBT>Z4S!oT~#C2Fg6~?3DQ> ziYO{-JR(|C5LS8cN_OiI0K2NrKCnYBob_3nokOTiq-sR?83!qFPUkoeRZUA(;tp!o z-qKtNS-+{3lf*`*Xel#*N}52^$wQ_&z*BLYE&V)r6GS;$qdQ~SSWOrKSJe3xrtpTf zk+CK+GODhU7K6T2(Z+ZR6?2*-keni{pG=M&P@_8C^ITT7 zw>Ki)a>BsWwAzott?X2wxDpUFWV5kS?cmaaNJB@fdz4059K4qAYg){0l@0Wu?gD0O zh%VX6!invuiuU!%vx$LcM|gPl@vJZ>SzA*ZtRT~!d!e7i1x##XrMDdh6*0aP{m zNT~0RV1H6LE8O7Yl7fDsBH5fa7(I|#iSTDW&Xa$+c+zQU+tFvGsKMq`N}G{+E-uP~ z>fWT_EIKrsBZJnRsyrgs-aC|^c>wNJE35JGI0BRJrQ76XT3Tj8^$5OgVCRsWZBg5G zo)fsZ4MA8#CaiC`QO=AFSMFwxil@vLw$D2YM(Y7jDbC9B-(Kj5!YU7va!Bl$#&nu_ot;$Yn8#p`HCEE$z<>x zP&i{xAexS)q;pX2HR3qV90KYxjC>8;t__G0l?Iv3nCf=m ze0lXu&$#r_sgbQ~HVZ#7xu4qKj%~00NfR zWGKw?v}HKQqj0zFjt7!CqrHna`-r=orx zu~p$`ntWy77l@ zu{>9EY*ez6?qN{ZSs2wO$0Ts-$dA0qpUi)B6pEJvGmd$jpt*?I!BMVAMfh`?PC?x@ zG_4KBaG=*#7VQ|+QQPhJ3!8&k6N6pcDWgvtncmA$jtH{KT`oyD9FvjMyI(zsSG8Ol zRTM37X)g2%JUlC0@s`GS^isJ;7-fC`0FjvG$jIh5!kxCT zEwvA`x_IFrvwU_;{Uw&Kq9b0m6~0-|I%ZQ4O-&wMCXiL!Fq)&nmq7{#<73)m`H*1lgUK8tS*a#J)7dBRO2V~K!loSIPL z0Ltk!S}OF=+XJ7}YA1BJNPGB;qG3G9R2OTz$tA>w=aSIDL*uiegK*))Gf7vTalu)X zly_QrhCD{+)lO+^@uY&YRd&Zv)H$vj^eROmdqzXtK1B%S@W}FODe2hu;lLcGDr}5- z)Xarf1>#y#1BRj9TETvXP&2Vkz~wBJwN}|-ZbtZAC)+hksyK!!Qh(5bHsi55as9&d zeARHz9yJ4z3LPMP|3K z0vy;{NItu+ZDO@L?Qo~2W=pCjjvu>#UWVgt4>ri0`ay=o19Xm{ZT*!yAnZ-AVU?zq zPIXcXX~G@EfO) zL0V1e8EU#6j>?*{lSWP!-mU;zc3JD4IKew1D*~%1X@glH7(u87hgiz(2%xaI|l8Ws5zPOwtWz zFgD6EYZ+x}*SUgvh846FxoAa6 zPBtZ2#?2vV*CvL_BsWsz@=R1ww;YwgV8&AuMhMD$rJe?DGZWK>${8OMmxo@HTc_)M`^0SsxvqpM+pa^Z|~R}6M?2SrCFFsR~%=5QJoh8t@nglaiBQE6Lz3~7~SCR9|6bt_RpNXXjUDyXeGB@Sui zRirbBx@v3FB5q89 zMF)ab(=IBmA~14QB~K7FfaSvj)n>SdjA^Oicry)M23Iq^@wCs1NU1et!{aBCuXeRU zScu6urODCmym>(Zbbc^0b_)rA;#XWjYY&vHMI5tBBcCaAoOdLZVqdp4kE^!JUL^P{ zs~bcLnu6J!GEhmVZmCH@B$f(|@@j0{3LP*rjPKQSTVD)O(nQkMG!TzSWP?Lhr8s$R z`NG}O>7gHNo4SompSd1uToMch0+SV*o-zKI2jr}w8}63&fK{hdebExOUd&>e2x;B3 z0<&y`NXI2rHA5mkVghn1_=`4p`=J(>0FrN!>Wa9beqV(?gtD*?426upR2fI|-MXE@ zx04*ADT&B9B}E(8sj-b~;NDSNWDwFr$@f<=n`A4gB#2vW z@Gu%FIp74pdWuFJ!i96MXN2sKvqI!dx?2ALu~ACZ5=B#E+)Wg6kmmEspaa7jE6G^u zoM&|wrhtZnq=h9mDLzJ9TS%&N^UXDvH+GySCt+joQUh`^x%mZdaaiE0o5)SLz87W0 z;l5+?8?QF;4P&b=b(K#!GRjFk^5ReKy$O75w3b@>vcd##I|TP51NUBWxCb~BoSjDi z_5oq){{YoUnP*|O_9bi=UOckK=H;xY@!hvZ@BNphIMtE+l+JcKla9?Z8Lz346KE=Z6p$T zm7;>A95sZJdIfJoWK>Ghi^*lHph|SuZ!xl+YT@eI{5kD0xAj^E@LWPKFfMjFtqEL! zt8B+pPpSOYf{EIi3ah)=&9Yr=1PtyJ%Vh1)GF3TR+ruZW7&Me_RxNNLQqs-|*sB&y z0Nkdl3HFMq_zuf8c{&*5=&z@o79G}N`BV&U0H#jf2_q95bE#+!$yw7ebBSr{tt-ci zBZoy!$Ov$q+kKXWIOoU$#}tYZ;7vryYUjPDS*bDtmh9@>Mw}9Czx6&K@JLC zGg%(6*95nKBipj8xV*+1^R_ouLvS12K)*D5xBzah&#aJ5WYqAr;u+y+)-?f)Wosj^ zV@7q4s^L^ud2OgND3!an)3{8Id~`Luv}lpF;f>ZCkbXNbdordAh7q4INA@b8398--a}M(7FIb$Y8ybe}56-|gCRKvXp}%DH4t1`b$W zU)2SnZM#!JBXksM_wZNrj;~1vigjUHFa|dn2{h1naOWKrRmS7)jyjG4>?*P`T$ZMs znQqrG9ONX(;f#W_Lx>s3xlD%c$yd99X6w%2RHeAkhw;ah5ViGY3{eyY@Y67T;&~;945$jdQzhc{_fAYo(xM!*H~8^eJi7U%LLIB>88Y zD0o)xZU&6NcK-mI_;9+f3h*pcF}3oK5>Bpi*+pZ4kJX{M`70=(Y)#lEgn#vm8@eGWr>bA1N#te|N z*NfB?t#ccUgYAT>6-0W1YDXVK;fOab{6~Ho1eN~!ha)UFZ_Rj}Mv^-BMKyle^sHsMP zNLbT5O?u2zG9MtQ}067akEc#!uR3$xTWE1980jXjjYKY zWGkv#U?Hv=ioK`QHn3H)Ugd4pzDZ{+h@O1VYdlCBvDMWbEh&-OoPO$GhUOkyE7F=s zsjbtf)gTB(r!}EUk6M8N-3W%G=UiVUevaBRE{zo%&?RCplMi^|$~W zr9IWlbVgHS=-WwkhCC2*qtw+ls9>!4s_`3vxl*AnXv4@;$5AN_vd57%^ek{Eb)BlF zF~7><2hC_JCeXu&ES=Wo?Pa57JSIo#{FPHe^HyEEv&iLZxPH$iLtD00EwY>5oFpSU zdajeg6qo>!u}^UI8i?qm-bZcxJB>LC^_H&)4=&2~l8vncE!zrf=~*K*hVO-|PLeQ0 z@!+MTqz!PuR=(&Q--IV1r0~u(sbQOLrrj75Mh;yRRkZkw?6Ft6Q0BW(9*d(^#NQ0; zu=mK^=MdGIt+EnWCR%x8CVCrlOk;d~5LYyo*9P z)DT-X^cwpx(K*;~$-pW_4AKd39&ivrK)Xcq=wpfrH1GK?Jz4`U4%4=GC?J(z>d4I1 z_KhWt?(%MVttU4e@5uxfR;S$9@u@kY(v_o6B{n^?gb-LZ9sM6_Ev>}r3r|J6P(fnq zh&80Av7~O$q0J0?`2q+kwRTahk=q*@;ib6wD{EWa*@hfO=peK-&l-qhbzpRL_v=k;f$j7G}Tf zELJl{Nz=Nr$yEZ?FCELr~mts`QOu+&m~1JyrJNMXVV zElB;MW@_|&g<~sRL3R%$5Lp_ZyicMmPIft@xy~{GNkIiBpR`w*aY6x)l3c)J2q3iQ z&}uBL;CN%?kzc${$_Ogf{e+9|GZn1`uGvvb8+=Y(5J6{0v6hWQ(c*Ga<0n=SK`D?| z*>7zblx11Vv$jD69bFMmvJG&{3dr3WGn{Xf5LWB$5KOe;z2kU2mrdbjj{%!?K?Q1$ zxV2YE+|v9brT`%>)*Hqq9Peh&BHJ0a)tb9X<%k2rLa(-i=!R W##>Ru&v$Q>Ac6~0A8IH*&;Qw&gu-k9 literal 0 HcmV?d00001 diff --git a/python/examples/xpu/resnet_v2_50_xpu/localpredict.py b/python/examples/xpu/resnet_v2_50_xpu/localpredict.py new file mode 100644 index 00000000..2e76098e --- /dev/null +++ b/python/examples/xpu/resnet_v2_50_xpu/localpredict.py @@ -0,0 +1,31 @@ +# Copyright (c) 2020 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. + +from paddle_serving_app.reader import Sequential, File2Image, Resize, CenterCrop +from paddle_serving_app.reader import RGB2BGR, Transpose, Div, Normalize +from paddle_serving_app.local_predict import LocalPredictor +import sys + +predictor = LocalPredictor() +predictor.load_model_config(sys.argv[1], use_lite=True, use_xpu=True, ir_optim=True) + +seq = Sequential([ + File2Image(), Resize(256), CenterCrop(224), RGB2BGR(), Transpose((2, 0, 1)), + Div(255), Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], True) +]) + +image_file = "daisy.jpg" +img = seq(image_file) +fetch_map = predictor.predict(feed={"image": img}, fetch=["score"]) +print(fetch_map["score"].reshape(-1)) diff --git a/python/examples/xpu/resnet_v2_50_xpu/resnet50_client.py b/python/examples/xpu/resnet_v2_50_xpu/resnet50_client.py new file mode 100644 index 00000000..b249d2a6 --- /dev/null +++ b/python/examples/xpu/resnet_v2_50_xpu/resnet50_client.py @@ -0,0 +1,32 @@ +# Copyright (c) 2020 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. + +from paddle_serving_client import Client +from paddle_serving_app.reader import Sequential, File2Image, Resize, CenterCrop +from paddle_serving_app.reader import RGB2BGR, Transpose, Div, Normalize + +client = Client() +client.load_client_config( + "resnet_v2_50_imagenet_client/serving_client_conf.prototxt") +client.connect(["127.0.0.1:9393"]) + +seq = Sequential([ + File2Image(), Resize(256), CenterCrop(224), RGB2BGR(), Transpose((2, 0, 1)), + Div(255), Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], True) +]) + +image_file = "daisy.jpg" +img = seq(image_file) +fetch_map = client.predict(feed={"image": img}, fetch=["score"]) +print(fetch_map["score"].reshape(-1)) -- GitLab