From 0b359fa0cbf01f9eba908f481201de5e55473778 Mon Sep 17 00:00:00 2001
From: YYJ-aaaa <1058387004@qq.com>
Date: Tue, 24 Sep 2024 11:07:01 +0800
Subject: [PATCH] commit message
---
.gitattributes | 19 +--
README.md | 203 ++++++++++++++++++++++-
config.py | 354 +++++++++++++++++++++++++++++++++++++++++
configuration.json | 332 ++++++++++++++++++++++++++++++++++++++
pytorch_model.pt | 3 +
resources/.gitkeep | 0
resources/overview.jpg | Bin 0 -> 331076 bytes
resources/test.jpg | Bin 0 -> 69831 bytes
8 files changed, 898 insertions(+), 13 deletions(-)
create mode 100644 config.py
create mode 100644 configuration.json
create mode 100644 pytorch_model.pt
create mode 100644 resources/.gitkeep
create mode 100644 resources/overview.jpg
create mode 100644 resources/test.jpg
diff --git a/.gitattributes b/.gitattributes
index a6344aa..97c8e80 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -1,35 +1,32 @@
*.7z filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.parquet filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
*.xz filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*.tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db* filter=lfs diff=lfs merge=lfs -text
+*.ark* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
diff --git a/README.md b/README.md
index 8b1bca6..f3a6632 100644
--- a/README.md
+++ b/README.md
@@ -1,3 +1,202 @@
-# ViT_image_classification
+---
+tasks:
+- image-classification
+model-type:
+- Transformer
+domain:
+- cv
+frameworks:
+- pytorch
+backbone:
+- ViT-Base
+metrics:
+- accuracy
+finetune-support: True
+customized-quickstart: True
+integrating: False
+license: Apache License 2.0
+tags:
+- Alibaba
+- Image classification
+- Transformer
+- Dailylife tags
+datasets:
+ evaluation:
+ - online dataset
+ train:
+ - tany0699/dailytags
+ test:
+ - tany0699/dailytags
-ViT是2020年Google团队提出的将Transformer应用在图像分类的模型，虽然不是第一篇将transformer应用在视觉任务的论文，但是因为其模型‘’简单”且效果好，可扩展性强（scalable，模型越大效果越好），成为了transformer在CV领域应用的里程碑著作，也引爆了后续相关研究。
+widgets:
+ - task: image-classification
+ inputs:
+ - type: image
+ examples:
+ - name: 1
+ inputs:
+ - name: image
+ data: git://resources/test.jpg
+---
+
+# 日常物体识别模型介绍
+自建1300类常见物体标签体系，覆盖常见的日用品，动物，植物，家具，设备，食物等物体，标签从海量中文互联网社区语料进行提取，保留了出现频率较高的常见物体名称。模型结构采用最新的ViT-Base结构。
+创空间快速可视化展示: [ViT图像分类-中文-日常物品](https://modelscope.cn/studios/tany0699/cv_vit-base_image-classification_Dailylife-labels/summary)
+
+本系列还有如下模型，欢迎试用:
+- [ViT图像分类-通用](https://modelscope.cn/models/damo/cv_vit-base_image-classification_ImageNet-labels/summary)
+- [NextViT实时图像分类-中文-日常物品](https://modelscope.cn/models/damo/cv_nextvit-small_image-classification_Dailylife-labels/summary)
+- [ConvNeXt图像分类-中文-垃圾分类](https://modelscope.cn/models/damo/cv_convnext-base_image-classification_garbage/summary)
+- [BEiTv2图像分类-通用-base](https://modelscope.cn/models/damo/cv_beitv2-base_image-classification_patch16_224_pt1k_ft22k_in1k/summary)
+- [BEiTv2图像分类-通用-large](https://modelscope.cn/models/damo/cv_beitv2-large_image-classification_patch16_224_pt1k_ft22k_in1k/summary)
+
+## 模型描述
+
+采用Transformer经典的[ViT-Base](https://github.com/google-research/vision_transformer)结构, 并采用了DeiT的知识蒸馏方式进行训练。
+
+
+## 期望模型使用方式以及适用范围
+
+本模型适用范围较广，覆盖大部分日常生活常见的物品类目，包括日用品，动物，植物，家具，设备，食物等。也可作为下游任务的预训练backbone。
+
+### 如何使用
+
+在ModelScope框架上，提供输入图片，即可通过简单的Pipeline调用来使用。
+
+#### 代码范例
+```python
+from modelscope.pipelines import pipeline
+from modelscope.utils.constant import Tasks
+
+img_path = 'https://modelscope.oss-cn-beijing.aliyuncs.com/test/images/bird.JPEG'
+image_classification = pipeline(Tasks.image_classification,
+ model='damo/cv_vit-base_image-classification_Dailylife-labels')
+result = image_classification(img_path)
+print(result)
+```
+
+### 模型局限性以及可能的偏差
+
+- 支持1300类常见物体识别
+
+
+## 训练数据介绍
+
+- 140万包含常见物体的图像集
+
+
+## 模型训练流程
+
+- 主要训练参数参考[DeiT论文](https://arxiv.org/abs/2012.12877)的设置，除了weight decay在复现时设置为0.1，模型训练未使用pretrained参数进行初始化。
+
+### 预处理
+
+测试时主要的预处理如下：
+- Resize：先将原始图片的短边缩放至256
+- Normalize：图像归一化，减均值除以标准差
+- CenterCrop：裁切为224x224
+
+## 数据评估及结果
+
+模型在自建测试集进行测试，结果如下:
+
+| Model | top-1 acc | top-5 acc | #params | Remark |
+|:--------:|:-------:|:--------:|:-------:|--------------|
+| ViT-base | 74.3 | 95.3 | 86M | modelscope |
+
+
+## 模型训练
+使用托管在modelscope DatasetHub上的小型数据集[mini_imagenet100](https://modelscope.cn/datasets/tany0699/mini_imagenet100/summary)进行finetune训练的示例代码:
+
+```python
+from modelscope.msdatasets import MsDataset
+from modelscope.metainfo import Trainers
+from modelscope.trainers import build_trainer
+import tempfile
+
+model_id = 'damo/cv_vit-base_image-classification_Dailylife-labels'
+
+# 加载数据
+ms_train_dataset = MsDataset.load(
+ 'mini_imagenet100', namespace='tany0699',
+ subset_name='default', split='train') # 加载训练集
+
+ms_val_dataset = MsDataset.load(
+ 'mini_imagenet100', namespace='tany0699',
+ subset_name='default', split='validation') # 加载验证集
+
+tmp_dir = tempfile.TemporaryDirectory().name # 使用临时目录作为工作目录
+
+# 修改配置文件
+def cfg_modify_fn(cfg):
+ cfg.train.dataloader.batch_size_per_gpu = 32 # batch大小
+ cfg.train.dataloader.workers_per_gpu = 2 # 每个gpu的worker数目
+ cfg.train.max_epochs = 1 # 最大训练epoch数
+ cfg.model.mm_model.head.num_classes = 100 # 分类数
+ cfg.model.mm_model.train_cfg.augments[0].num_classes = 100 # 分类数
+ cfg.model.mm_model.train_cfg.augments[1].num_classes = 100 # 分类数
+ cfg.train.optimizer.lr = 1e-4 # 学习率
+ cfg.train.lr_config.warmup_iters = 1 # 预热次数
+ cfg.train.evaluation.metric_options = {'topk': (1, 5)} # 训练时的评估指标
+ cfg.evaluation.metric_options = {'topk': (1, 5)} # 评估时的评估指标
+ return cfg
+
+# 构建训练器
+kwargs = dict(
+ model=model_id, # 模型id
+ work_dir=tmp_dir, # 工作目录
+ train_dataset=ms_train_dataset, # 训练集
+ eval_dataset=ms_val_dataset, # 验证集
+ cfg_modify_fn=cfg_modify_fn, # 用于修改训练配置文件的回调函数
+ model_revision='v1.0.2'
+ )
+trainer = build_trainer(name=Trainers.image_classification, default_args=kwargs)
+
+# 进行训练
+trainer.train()
+
+# 进行评估
+result = trainer.evaluate()
+print('result:', result)
+```
+训练说明见示例代码中的注释，更详细的训练说明和用法见官方的[训练文档](https://modelscope.cn/docs/%E6%A8%A1%E5%9E%8B%E7%9A%84%E8%AE%AD%E7%BB%83Train)。训练过程产生的log和模型权重文件保存在work_dir工作目录中，并以前缀为'best_'的文件保存了在验证集上最优精度的权重。evaluate()默认使用精度最高的模型权重进行评估。
+
+
+## 模型评估
+使用训练好的模型对需要评估的数据集进行精度评估示例代码如下:
+
+```python
+from modelscope.msdatasets import MsDataset
+from modelscope.metainfo import Trainers
+from modelscope.trainers import build_trainer
+import tempfile
+
+model_id = 'damo/cv_vit-base_image-classification_Dailylife-labels'
+
+# 加载用于评估的数据集
+ms_val_dataset = MsDataset.load(
+ 'dailytags', namespace='tany0699',
+ subset_name='default', split='validation')
+
+tmp_dir = tempfile.TemporaryDirectory().name # 使用临时目录作为工作目录
+
+# 构建训练器
+kwargs = dict(
+ model=model_id, # 模型id
+ work_dir=tmp_dir, # 工作目录
+ train_dataset=None,
+ eval_dataset=ms_val_dataset, # 评估的数据集
+ model_revision='v1.0.2'
+ )
+trainer = build_trainer(name=Trainers.image_classification, default_args=kwargs)
+
+# 开始评估
+result = trainer.evaluate()
+print('result:', result)
+```
+评估过程默认使用模型中自带的预训练权重进行评估。
+
+#### Clone with HTTP
+```bash
+ git clone https://www.modelscope.cn/damo/cv_vit-base_image-classification_Dailylife-labels.git
+```
\ No newline at end of file
diff --git a/config.py b/config.py
new file mode 100644
index 0000000..ef2266e
--- /dev/null
+++ b/config.py
@@ -0,0 +1,354 @@
+rand_increasing_policies = [
+ dict(type='AutoContrast'),
+ dict(type='Equalize'),
+ dict(type='Invert'),
+ dict(type='Rotate', magnitude_key='angle', magnitude_range=(0, 30)),
+ dict(type='Posterize', magnitude_key='bits', magnitude_range=(4, 0)),
+ dict(type='Solarize', magnitude_key='thr', magnitude_range=(256, 0)),
+ dict(
+ type='SolarizeAdd',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 110)),
+ dict(
+ type='ColorTransform',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(type='Contrast', magnitude_key='magnitude', magnitude_range=(0, 0.9)),
+ dict(
+ type='Brightness', magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Sharpness', magnitude_key='magnitude', magnitude_range=(0, 0.9)),
+ dict(
+ type='Shear',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.3),
+ direction='horizontal'),
+ dict(
+ type='Shear',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.3),
+ direction='vertical'),
+ dict(
+ type='Translate',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.45),
+ direction='horizontal'),
+ dict(
+ type='Translate',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.45),
+ direction='vertical')
+]
+dataset_type = 'ImageNet'
+img_norm_cfg = dict(
+ mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+train_pipeline = [
+ dict(type='LoadImageFromFile'),
+ dict(
+ type='RandomResizedCrop',
+ size=224,
+ backend='pillow',
+ interpolation='bicubic'),
+ dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
+ dict(
+ type='RandAugment',
+ policies=[
+ dict(type='AutoContrast'),
+ dict(type='Equalize'),
+ dict(type='Invert'),
+ dict(
+ type='Rotate', magnitude_key='angle', magnitude_range=(0, 30)),
+ dict(
+ type='Posterize', magnitude_key='bits',
+ magnitude_range=(4, 0)),
+ dict(
+ type='Solarize', magnitude_key='thr',
+ magnitude_range=(256, 0)),
+ dict(
+ type='SolarizeAdd',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 110)),
+ dict(
+ type='ColorTransform',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Contrast',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Brightness',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Sharpness',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Shear',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.3),
+ direction='horizontal'),
+ dict(
+ type='Shear',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.3),
+ direction='vertical'),
+ dict(
+ type='Translate',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.45),
+ direction='horizontal'),
+ dict(
+ type='Translate',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.45),
+ direction='vertical')
+ ],
+ num_policies=2,
+ total_level=10,
+ magnitude_level=9,
+ magnitude_std=0.5,
+ hparams=dict(pad_val=[104, 116, 124], interpolation='bicubic')),
+ dict(
+ type='RandomErasing',
+ erase_prob=0.25,
+ mode='rand',
+ min_area_ratio=0.02,
+ max_area_ratio=0.3333333333333333,
+ fill_color=[103.53, 116.28, 123.675],
+ fill_std=[57.375, 57.12, 58.395]),
+ dict(
+ type='Normalize',
+ mean=[123.675, 116.28, 103.53],
+ std=[58.395, 57.12, 57.375],
+ to_rgb=True),
+ dict(type='ImageToTensor', keys=['img']),
+ dict(type='ToTensor', keys=['gt_label']),
+ dict(type='Collect', keys=['img', 'gt_label'])
+]
+test_pipeline = [
+ dict(type='LoadImageFromFile'),
+ dict(
+ type='Resize',
+ size=(256, -1),
+ backend='pillow',
+ interpolation='bicubic'),
+ dict(type='CenterCrop', crop_size=224),
+ dict(
+ type='Normalize',
+ mean=[123.675, 116.28, 103.53],
+ std=[58.395, 57.12, 57.375],
+ to_rgb=True),
+ dict(type='ImageToTensor', keys=['img']),
+ dict(type='Collect', keys=['img'])
+]
+data = dict(
+ samples_per_gpu=32,
+ workers_per_gpu=16,
+ train=dict(
+ type='ImageNet',
+ data_prefix='/data/oss_bucket_0/',
+ pipeline=[
+ dict(type='LoadImageFromFile'),
+ dict(
+ type='RandomResizedCrop',
+ size=224,
+ backend='pillow',
+ interpolation='bicubic'),
+ dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
+ dict(
+ type='RandAugment',
+ policies=[
+ dict(type='AutoContrast'),
+ dict(type='Equalize'),
+ dict(type='Invert'),
+ dict(
+ type='Rotate',
+ magnitude_key='angle',
+ magnitude_range=(0, 30)),
+ dict(
+ type='Posterize',
+ magnitude_key='bits',
+ magnitude_range=(4, 0)),
+ dict(
+ type='Solarize',
+ magnitude_key='thr',
+ magnitude_range=(256, 0)),
+ dict(
+ type='SolarizeAdd',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 110)),
+ dict(
+ type='ColorTransform',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Contrast',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Brightness',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Sharpness',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.9)),
+ dict(
+ type='Shear',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.3),
+ direction='horizontal'),
+ dict(
+ type='Shear',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.3),
+ direction='vertical'),
+ dict(
+ type='Translate',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.45),
+ direction='horizontal'),
+ dict(
+ type='Translate',
+ magnitude_key='magnitude',
+ magnitude_range=(0, 0.45),
+ direction='vertical')
+ ],
+ num_policies=2,
+ total_level=10,
+ magnitude_level=9,
+ magnitude_std=0.5,
+ hparams=dict(pad_val=[104, 116, 124],
+ interpolation='bicubic')),
+ dict(
+ type='RandomErasing',
+ erase_prob=0.25,
+ mode='rand',
+ min_area_ratio=0.02,
+ max_area_ratio=0.3333333333333333,
+ fill_color=[103.53, 116.28, 123.675],
+ fill_std=[57.375, 57.12, 58.395]),
+ dict(
+ type='Normalize',
+ mean=[123.675, 116.28, 103.53],
+ std=[58.395, 57.12, 57.375],
+ to_rgb=True),
+ dict(type='ImageToTensor', keys=['img']),
+ dict(type='ToTensor', keys=['gt_label']),
+ dict(type='Collect', keys=['img', 'gt_label'])
+ ],
+ ann_file='/data/oss_bucket_0/virgo_data/dailytags/train_mmcls.txt',
+ classes='/data/oss_bucket_0/virgo_data/dailytags/classname.txt'),
+ val=dict(
+ type='ImageNet',
+ data_prefix='/data/oss_bucket_0/',
+ ann_file='/data/oss_bucket_0/virgo_data/dailytags/val_mmcls.txt',
+ pipeline=[
+ dict(type='LoadImageFromFile'),
+ dict(
+ type='Resize',
+ size=(256, -1),
+ backend='pillow',
+ interpolation='bicubic'),
+ dict(type='CenterCrop', crop_size=224),
+ dict(
+ type='Normalize',
+ mean=[123.675, 116.28, 103.53],
+ std=[58.395, 57.12, 57.375],
+ to_rgb=True),
+ dict(type='ImageToTensor', keys=['img']),
+ dict(type='Collect', keys=['img'])
+ ],
+ classes='/data/oss_bucket_0/virgo_data/dailytags/classname.txt'),
+ test=dict(
+ type='ImageNet',
+ data_prefix='/data/oss_bucket_0/',
+ ann_file='/data/oss_bucket_0/virgo_data/dailytags/val_mmcls.txt',
+ pipeline=[
+ dict(type='LoadImageFromFile'),
+ dict(
+ type='Resize',
+ size=(256, -1),
+ backend='pillow',
+ interpolation='bicubic'),
+ dict(type='CenterCrop', crop_size=224),
+ dict(
+ type='Normalize',
+ mean=[123.675, 116.28, 103.53],
+ std=[58.395, 57.12, 57.375],
+ to_rgb=True),
+ dict(type='ImageToTensor', keys=['img']),
+ dict(type='Collect', keys=['img'])
+ ],
+ classes='/data/oss_bucket_0/virgo_data/dailytags/classname.txt'))
+evaluation = dict(interval=2, metric='accuracy', save_best='auto')
+paramwise_cfg = dict(
+ norm_decay_mult=0.0,
+ bias_decay_mult=0.0,
+ custom_keys=dict({
+ '.absolute_pos_embed': dict(decay_mult=0.0),
+ '.relative_position_bias_table': dict(decay_mult=0.0),
+ '.cls_token': dict(decay_mult=0.0),
+ '.pos_embed': dict(decay_mult=0.0)
+ }))
+optimizer = dict(
+ type='AdamW',
+ lr=0.001,
+ weight_decay=0.1,
+ eps=1e-08,
+ betas=(0.9, 0.999),
+ paramwise_cfg=dict(
+ norm_decay_mult=0.0,
+ bias_decay_mult=0.0,
+ custom_keys=dict({
+ '.absolute_pos_embed': dict(decay_mult=0.0),
+ '.relative_position_bias_table': dict(decay_mult=0.0),
+ '.cls_token': dict(decay_mult=0.0),
+ '.pos_embed': dict(decay_mult=0.0)
+ })))
+optimizer_config = dict(grad_clip=dict(max_norm=5.0))
+lr_config = dict(
+ policy='CosineAnnealing',
+ by_epoch=False,
+ min_lr_ratio=0.01,
+ warmup='linear',
+ warmup_ratio=0.001,
+ warmup_iters=20,
+ warmup_by_epoch=True)
+runner = dict(type='EpochBasedRunner', max_epochs=300)
+checkpoint_config = dict(interval=1, max_keep_ckpts=20, create_symlink=False)
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+load_from = None
+resume_from = '/data/oss_bucket_1/emian/VisualContentRecognition/mmlab/work_dirs/dailytags_deit-base_pt-16xb64_in1k_gzn2_pai_wd0.1/epoch_296.pth'
+workflow = [('train', 1)]
+model = dict(
+ type='ImageClassifier',
+ backbone=dict(
+ type='VisionTransformer',
+ arch='deit-base',
+ img_size=224,
+ patch_size=16,
+ drop_path_rate=0.1),
+ neck=None,
+ head=dict(
+ type='VisionTransformerClsHead',
+ num_classes=1296,
+ in_channels=768,
+ loss=dict(
+ type='LabelSmoothLoss', label_smooth_val=0.1, mode='original')),
+ init_cfg=[
+ dict(type='TruncNormal', layer='Linear', std=0.02),
+ dict(type='Constant', layer='LayerNorm', val=1.0, bias=0.0)
+ ],
+ train_cfg=dict(augments=[
+ dict(type='BatchMixup', alpha=0.8, num_classes=1296, prob=0.5),
+ dict(type='BatchCutMix', alpha=1.0, num_classes=1296, prob=0.5)
+ ]))
+custom_hooks = [dict(type='EMAHook', momentum=4e-05, priority='ABOVE_NORMAL')]
+work_dir = '/data/oss_bucket_1/emian/VisualContentRecognition/mmlab/work_dirs/dailytags_deit-base_pt-16xb64_in1k_gzn2_pai_wd0.1/'
+gpu_ids = range(0, 32)
diff --git a/configuration.json b/configuration.json
new file mode 100644
index 0000000..9bf6009
--- /dev/null
+++ b/configuration.json
@@ -0,0 +1,332 @@
+{
+ "framework":"pytorch",
+ "task":"image-classification",
+ "pipeline":{
+ "type":"vit-base_image-classification_Dailylife-labels"
+ },
+ "model":{
+ "type": "ClassificationModel",
+ "mm_model": {
+ "type": "ImageClassifier",
+ "pretrained": null,
+ "backbone": {
+ "type": "VisionTransformer",
+ "arch": "deit-base",
+ "img_size": 224,
+ "patch_size": 16,
+ "drop_path_rate": 0.1
+ },
+ "neck": null,
+ "head": {
+ "type": "VisionTransformerClsHead",
+ "num_classes": 1296,
+ "in_channels": 768,
+ "loss": {
+ "type": "LabelSmoothLoss",
+ "label_smooth_val": 0.1,
+ "mode": "original"
+ }
+ },
+ "init_cfg": [
+ {
+ "type": "TruncNormal",
+ "layer": "Linear",
+ "std": 0.02
+ },
+ {
+ "type": "Constant",
+ "layer": "LayerNorm",
+ "val": 1.0,
+ "bias": 0.0
+ }
+ ],
+ "train_cfg":
+ {
+ "augments": [
+ {
+ "type": "BatchMixup",
+ "alpha": 0.8,
+ "num_classes": 1296,
+ "prob": 0.5
+ },
+ {
+ "type": "BatchCutMix",
+ "alpha": 1.0,
+ "num_classes": 1296,
+ "prob": 0.5
+ }
+ ]
+ }
+ }
+ },
+
+ "dataset": {
+ "classes": null
+ },
+
+ "preprocessor": {
+ "type": "image-classification-bypass-preprocessor",
+ "train": [
+ {
+ "type": "LoadImageFromFile"
+ },
+ {
+ "type": "RandomResizedCrop",
+ "size": 224,
+ "backend": "pillow",
+ "interpolation": "bicubic"
+ },
+ {
+ "type": "RandomFlip",
+ "flip_prob": 0.5,
+ "direction": "horizontal"
+ },
+ {
+ "type": "RandAugment",
+ "policies": [
+ {
+ "type": "AutoContrast"
+ },
+ {
+ "type": "Equalize"
+ },
+ {
+ "type": "Invert"
+ },
+ {
+ "type": "Rotate",
+ "magnitude_key": "angle",
+ "magnitude_range": [0, 30]
+
+ },
+ {
+ "type": "Posterize",
+ "magnitude_key": "bits",
+ "magnitude_range": [4, 0]
+
+ },
+ {
+ "type": "Solarize",
+ "magnitude_key": "thr",
+ "magnitude_range": [256, 0]
+
+ },
+ {
+ "type": "SolarizeAdd",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 110]
+
+ },
+ {
+ "type": "ColorTransform",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.9]
+
+ },
+ {
+ "type": "Contrast",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.9]
+
+ },
+ {
+ "type": "Brightness",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.9]
+
+ },
+ {
+ "type": "Sharpness",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.9]
+
+ },
+ {
+ "type": "Shear",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.3],
+ "direction": "horizontal"
+
+ },
+ {
+ "type": "Shear",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.3],
+ "direction": "vertical"
+
+ },
+ {
+ "type": "Translate",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.45],
+ "direction": "horizontal"
+
+ },
+ {
+ "type": "Translate",
+ "magnitude_key": "magnitude",
+ "magnitude_range": [0, 0.45],
+ "direction": "vertical"
+
+ }
+ ],
+ "num_policies": 2,
+ "total_level": 10,
+ "magnitude_level": 9,
+ "magnitude_std": 0.5,
+ "hparams": {
+ "pad_val": [104, 116, 124],
+ "interpolation": "bicubic"
+ }
+ },
+ {
+ "type": "RandomErasing",
+ "erase_prob": 0.25,
+ "mode": "rand",
+ "min_area_ratio": 0.02,
+ "max_area_ratio": 0.3333333333333333,
+ "fill_color": [103.53, 116.28, 123.675],
+ "fill_std": [57.375, 57.12, 58.395]
+ },
+ {
+ "type": "Normalize",
+ "mean": [123.675, 116.28, 103.53],
+ "std": [58.395, 57.12, 57.375],
+ "to_rgb": true
+ },
+ {
+ "type": "ImageToTensor",
+ "keys": ["img"]
+ },
+ {
+ "type": "ToTensor",
+ "keys": ["gt_label"]
+ },
+ {
+ "type": "Collect",
+ "keys": ["img", "gt_label"]
+ }
+ ],
+
+ "val": [
+ {
+ "type": "LoadImageFromFile"
+ },
+ {
+ "type": "Resize",
+ "size": [256, -1],
+ "backend": "pillow",
+ "interpolation": "bicubic"
+ },
+ {
+ "type": "CenterCrop",
+ "crop_size": 224
+ },
+ {
+ "type": "Normalize",
+ "mean": [123.675, 116.28, 103.53],
+ "std": [58.395, 57.12, 57.375],
+ "to_rgb": true
+ },
+ {
+ "type": "ImageToTensor",
+ "keys": ["img"]
+ },
+ {
+ "type": "Collect",
+ "keys": ["img"]
+ }
+ ]
+ },
+
+ "train": {
+ "dataloader": {
+ "batch_size_per_gpu": 32,
+ "workers_per_gpu": 4
+ },
+ "max_epochs": 1,
+ "runner": {
+ "type": "EpochBasedRunner",
+ "max_epochs": 300
+ },
+ "evaluation": {
+ "interval": 1,
+ "metric": "accuracy",
+ "save_best": "auto"
+ },
+ "checkpoint_config": {
+ "interval": 1,
+ "max_keep_ckpts": 20,
+ "create_symlink": true
+ },
+ "log_config": {
+ "interval": 100,
+ "hooks": [
+ {
+ "type": "TextLoggerHook"
+ }
+ ]
+ },
+ "custom_hooks": [
+ {
+ "type": "EMAHook",
+ "momentum": 4e-05,
+ "priority": "ABOVE_NORMAL"
+ }
+ ],
+ "workflow": [
+ ["train", 1]
+ ],
+ "work_dir": "./work_dir",
+ "optimizer": {
+ "type": "AdamW",
+ "lr": 0.001,
+ "weight_decay": 0.1,
+ "eps": 1e-08,
+ "betas": [0.9, 0.999],
+ "paramwise_cfg": {
+ "norm_decay_mult": 0.0,
+ "bias_decay_mult": 0.0,
+ "custom_keys": {
+ ".absolute_pos_embed": {
+ "decay_mult": 0.0
+ },
+ ".relative_position_bias_table": {
+ "decay_mult": 0.0
+ },
+ ".cls_token": {
+ "decay_mult": 0.0
+ },
+ ".pos_embed": {
+ "decay_mult": 0.0
+ }
+ }
+ }
+ },
+ "optimizer_config": {
+ "grad_clip": {
+ "max_norm": 5.0
+ }
+ },
+ "lr_config": {
+ "policy": "CosineAnnealing",
+ "by_epoch": false,
+ "min_lr_ratio": 0.01,
+ "warmup": "linear",
+ "warmup_ratio": 0.001,
+ "warmup_iters": 20,
+ "warmup_by_epoch": true
+ }
+ },
+
+ "evaluation": {
+ "dataloader": {
+ "batch_size_per_gpu": 32,
+ "workers_per_gpu": 4
+ },
+ "metrics": ["accuracy"],
+ "metric_options": {
+ "topk": [1, 5]
+ }
+ }
+}
diff --git a/pytorch_model.pt b/pytorch_model.pt
new file mode 100644
index 0000000..fc9e9e0
--- /dev/null
+++ b/pytorch_model.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1732b9c9c2daad2b55b013d8a4e75a5af9b0bffe1c0843561849194f50a2bf7f
+size 347270855
diff --git a/resources/.gitkeep b/resources/.gitkeep
new file mode 100644
index 0000000..e69de29
diff --git a/resources/overview.jpg b/resources/overview.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9ce6f2a5d707f8f74d6b5940f8be826be0d85715
GIT binary patch
literal 331076
zcmb@tc|6qb_cwmcV#Yp}v1KepmL&T=wq(x|6-kzmE&DFJP?9|rLfI>6v6fx7N`xp&
zhOs6YYa@f-OYi0L`F(%CKkoZ}+*f8^&pEGaJJ)iqbIx_z|FTa2Oa?moIsgU(AQk!t
z_NRcqcChOe0E~>lF#rHHK*A0IDhNSYfEtYZU#tTY0f;|zH~RC6Ousjq`ndIphhVD!AJi=*=dBKVw
zJrCf38>;0PK-oL^d25=Qo<6AapXZk6;XSIcovYX#qe82D+X<*d2@zKH=wXe$XGN
zZwP#-i`D`CPafX@vxEK%LUc1HPu-IchIl~)Uh&mA;Qb(fqK6-Z{_u_{aB%YJMY26;G(*f%5Rp@i@{{Nr;5A{DeH6eaa|6MlVpBDKs|CC2*qfVjp
zP&z14aO!KnN*Bmd3QzZ`)7
z%*NmOXe4RiG#oSVED{zGRY54DK
z{M`e@4a6%%7oru>gJ?k10)9j@q7Lx^(RhIWZeQom>^lG3n%STJxI^>e`fr|p_&u2M
zgEkM=ho9sV$wf)ZU$yu<1^Ywm8EAQVhxocWyZG~KLZ^chzrLr#iR1jzlF~8&9Go==
zc>q{v|8s`G*x&t!7JC=~)sX%D{f7V04Eq6i)d+3>(f`ndXCX214uI~v4gtP_f7ORP
zcu+x70t4+dF2D-}fC$8oG>``>Km%w417HHqf(yU~TCXn9I`;)ZAPhuNELZ|0umyHtFc=!f0Aq!5!T4ZDVPY_8m?BIK
zrVTp{GlN}#*}B|
zAd`^U$TDON@-1=@IfW!4_ox`Ca8zPc%2fJPmQ>DEK~(WnSyUxd)l?nOUYe!aLZMLX
zC}ETWv==X++)&q0NvH>?N>mH#6N-r1M5EDMXfd=Z+5~Nn_D9E~@1viho6!U4Su~lN
zj+&2Jnp%h2irSMphB}M-F?AF50QEfe4h<8{Q5t0$6BDL(;7=#!!7_1ol8Il>w8QK}XF_0Nq8BZ|kGdeOxGUhPWFb**O#L!@l
zVl*)}m{3d><`t$7^Mi?oNr*{{$(|{KDTk?^X@qH$nUz_J*@W4HIf?l(a}VSNS8p~S5+QB-DMPY@p`dByYEo=q$BbLO*!Y0dho-LRy
zhpmx~z>Z)SV%KNC%AUghl6{!{HwO=gCWjNpO^#*-N{^sK6
z(&lpKO5>{KnmB|!bnMWXLqUh~4s{&*!Hwls;dbCo|c+}La|-JVhYCLt9v5K{Q5W$Mc_cCzy^kzj%t1zO7RWwyhQtVcuQaY&=snnpftt_wXul!PZMMYf2UFEUL
z0$v#JgfGGqRgb9Jt3FgEsPU`WsTHUZ)DNrMt3Oho))3He)F{!I*A&%s(|oG=^Q7cS
z-;>oRe`zUeg=@8(LY&e)m3Zo-HjDN-?QHFDI!APzb)M*ubmerf>9*)m=^5&!>5b^~
z=sW0_>yr!=48jfG8qyn@8)h3$pFVcl_w;KcxRHTThS9k3QDaZzS`%QRZ<1j$VJd9u
zWBTR{>Wt}`oHKJ~Qf6UhUFNLjw&st`x6YnCn{sySoXEL=b8Y7_=dI6|pZ{fX$|Bu@
zU@2i4Zu$Pgp$o1T>a5UK=dDVtHZN*lymN8ZTFyG&diav?rQl0FHivAk+BDi?Z0&7d
z*-_b9*gdwR*qhitvfsR{e>wLu$wABEuEUa}hGVAVqLZpqy3_m>)hp>&7MxX`Gn^M)
z)LpV%ezAB^*{c<;PFLmF)YH_vFgWAK+qu!In)7|r}7q3^a*C%f=?|5&5
zkFw7lACj-3Z;2nw?~-4gKNeEzKLm&b#05+Rss-i*l7r3%y$WUy_6+_IaxCO#$b6_y
zXz?}VHHT|$VTZ$_!=}PD!yiV#A}&X?MG8b-kDQItjVg<#jdqXjjggE=kJ*T|h<$yX
z=X%t2Vw`T=lX%8>-}sRPm4y5o$Q!OVKHikMdH?2KqGRH_Tavf3Zc&mPlHT2xx_$5V
ze)5&%k16sgd8w#W&(z^G^|Z2d=Je2XVun#hO(uV4V&=vj+dDm3vRV0eY3~NyCEPoG
zukOC!{nY!EY?tiMIVW=}b8)#hbALT>d@zuwk@qa0C;wLd?*i9?FAsGe);tn=lvRi-
z3@n^0I$zXPtXN!La;W51$!@7}|P1dHG}9VN$S|EXy}eV}#l
z)L`e3{?Pl+#-Be8pBw%%Vm(6m;`n9Z>(#HTqk*G4V^QN&yranxYPmd8Dh)XkmGnCoua}0A?^L+E=3(^Y>i#m(_-!FcjUGiSq`4PX&yqx<}
zxkhC24lmcXfX)X`O4mbVGKdZPRq~+m^@H&aXr=7rAU(al7lc#qar@APSmt
zZ&!5p^`7D0_`b*fzPG)v{U2T6!0v`$fvohkQUDk(0Kf|A5VRNnvIYO-fc<5^LKyb<
z=Yae_@L%@g!3#)-0MDSuff}#X0>E?VBNPSI4LuK3__G=SE+g~r2cjG#zbi7*e*lc$
z7g9~g`}^xW0H9w3u^E#QNArh*h2DE-q2430oj
zq0rPcv~*C0Y9;`OArNpRf{NtaNnRLu!GOFIw5m
zK9K}vF8jvN(6V!IaveGy);RuAaVu`Pp;lEg)6S!O`i8vx}>n
zpMOALP;f|S?De?#gc~;#(=#&fWZk`YKl@Q(QE^FWS^49aud1qRYU}D>x3$0R=W?0@jZ4Dkg=A`wXR
z0bek95cEMXBdHEcp;$D}pzVEFk4Q&TV^5|%e9=lHAY(>iyX^akmR(SG>gehLYkxTV
zUt=uh|A@1H8T$`kW02Yn|C11KID!g+Kv1EmpoF4^v@{fonwI)cqWwG3|4EDo3G;8-
zhbn66(#P?-z%-8O3nK#y`O}rn&#vEDF~%|07a(B<
z{VmD5M^aODYeUbex93-4fd{uwawhyD$h{n+98W3NAm;5Zz$AUTIE#f5d6{9YH-)o}
z1Ltws0P9c-&rC&O_|5(Lge&mAsGlORiWnV62H+d+D(UUhDH=$b6bER0bO${^2hg<$
z;LhaV&6CogNrruoJa$gi@{oBg|6XRD@a#_&Q`m2Wf!`3R?NRH>P*5qXf!8lDCJM_G
zEc@{0cHCjrz<%mZiwYp?8$#r9Z2`*OqDqM2Y?Ka~zaK$Dt5xMRd1WUhYs?IKtF2VJ
zfk5wO1#N|;b-~#4SeZw)K8I`IouzGG^69el!T}{Ki
z?)iZXexIe$D0pDn7>}!84-c%UrKfc+c5s#`3mR2#luSkSS?O?oSF4&zmCEgh$M^N-
zxwP{32vGU+1{A!$T#z!nOIjhpLz=Uyuk7M@)mjl$a5nNcbAogVB_`c;Hy
zRO$t?h!`|yu|qNi^2_*)=KNItg#{#^C}Te(JZ{gw(4}Qp}LeJ5QtIlGrm&gZU)dYil-4=g}YiW{Z$e8=-N3j((5JLsj_{R^mYd39Y@
zu*-&XvB?F#BKG|bOByhKM@GG3K}Y8bAx*_ay;8oQV!%H$|8T_w!15+KWb=(D%E#(?
ztH1y}ckV&-uh-LQdU#&UT5)%uEMSHDibmx54JSQl_>`8Gr(XahgoiVDXq05X4wj`VX1smBzENV~Q
zU5kdXRfJvFV7`=4mlkLyVfrmIB+)uzrEF}*!nL6XAxjzz-DByYW;1Kbw6
zOo|<1xFCWlKqL=M`i#^SpktHZe%Y>s`SoIM9rWVKq^+5LcwR2fOjPi8Rq0WE^pZvo
zFomB8Wi_Dg)V1Qk&!-v2MA#6kfx@i#Vxm^o%o+otviGBieKL135SiKRRPLu^v6aOI
zAkwrLWBIHA4T9fSTU0xI>pxUg+HUf&wZeFN4VC2zxz~OI2{G
zJJ348~5=T4RmPlIXPPZdpKq^;HP1cp^M~wCKGjL}0qY&!mQ3n2QladopynfH|qs3b@RycPbA?
z0e^y-FSbDtTSyA1}22K2yEeXve
zF!NTz(|Rf{_43dB=G_#w3GGMt%FpCw*5i|7?(U}b6$@z)N2&pGS)fLFU_My2^5Zz}y)TWRBRcxm#Pz>y@(Sp;t^dz*KnY1rT
z%JB1Z3}Q4bB9|7}7=t1^Mrz`3vo&KF!D42Tf;E3TSzX&AL7#uS8MP<;O*o!+OC?;(
zAm{X;&?Qx8Zu=~*yrJ&P(jZCjcW?qHf8krmg+5OVpB~a?X{^_qnGt}%)^}P
z$R5PW+d-oYRiW0uNOa<2Iy>4!&-GdErOm$dH)9^vT9azaQ_kMq8In~s_|^7uL4I^|
z_Rh}eHr?a5S)I#*W3|gL!MKVK_qU(hi{=x$u&_^IPphKf5f0szJ(G;15hG2ayU!Xf
zkqq_8n`J$ONW|in!t1<;ySE?ZEX)rv;n_8Asf&GZnh;vMN3AyJV_ivMySNp*`%sPQ
z3mI)hHh(NC)`>UbjqJZ9DHU;!-Q_W2$wPorw8y`-@SC@J*Pd#qcRo;ICy3
znyNL7!I@e?Gs!&`wQ9d
za9bcYo^ec<)RqN%O(oo-XcZaF*t%M&Jq+?X(f`f*`Nf7CU+m6KW#X80%L*8mxKyQ*
zGv(39aQD^xEzQAZ!b96Ci>q%!QU{1M
zTkhjGZT{9b(0n4ykJPudoiQbM_pNR0B=7w2W{7{c9HP*6!RnNew_}`>3!yG~jC;FH
ztP_EQFWG$E2gq&mi&YMy#E-$E;Fal;veD)nb4^3yKTq&ddwhAy`trywrc?l&vh)iD&L9;BJa|KWkM&MzinR)F=1IyblF{li$_iMSP@Qg0u
zVgcZ-%Qw)RzcMDUHB0_FEisxX@wJptdt-*=dDH31g@quF;T!qxb&Yjt0{1^Wr8{!+
z=b3Y~>1VTq`6bRw^NrdP6@JX`Tw92I)s=wPoaVy?pWGtq46b;D=`Du);j_~prhoq!
zkf_Awa6uEHLRT2i?Nf#LWhXDG3)xBDB7L}3y->!G-=>J0{7LlY>Sm%Q??Nc9A!A8<
zncyB0ps^%$#730lRm?s(r%)4`4I-u|B}+&bxgfbC7MUj^=$AZE2CM6O$EYg0%@_1c
zgg-sVlr}e31G
zqQ=>a&m^t3JWN(0iNJtxrJHrR^~m+Tep}u!1rlBB@b9@|51|*&mMqS=zGlCDRpa$%
zIk6aa?%-Fc1uwYAW0(SupSvjBu8xV2`))%$?1>LqWtrDsExmbuqJ3d_tHSRd`>5
zb^NG$Z%AjD!Ow!+5r67em5TRb7MIdmXSqLCg;>7o{N<4IxA`);)yd_|NtUpqq@0d?
zjwU{eSu;tniS^T5y`jjDsa@--0$DZp3Wg4X2O!}YVS;r7r
z;(f!B>^`mi%3%vR*;K(KjR6W@!$N(oQl!5`1;o+nNZ3t#zAvt}rm-aY?5
zB~lpwSd0zq#*cOLlU$VWBU2K~ldBe6O}oh(D=_)A)#07DwKk&4iQ^5w*;71lSGq5<
zNj(}<;%*@3zVT=4k9Zp)yZb!SeKI~zbk%6G>&Ya&Td2^Z8cAfo?6oI>47BfGYs;@T
z`F@###Q#oU7b5RL`StXW72CAPR|;?W#$+^htK=%b61VQIMy)d5M<;kztBCTydQ}s%
zj3+z9e;4JijX#sr9j8Cl4rhc_mhFQnPjdd2Kl#?wO57y(rp_>hkI2>3HPy`av7G0v
zMRR+p+Zepl(Pyt@_S~s^hx;X+3ouo_-ZCvb<$68coE;M9Tkse+?cd%aunya{2RsGzS;tqDS;(jen^-I|0DVJTC&>xvKhUp3lNYz836iy6)Hh$
z8@bCpJT9>I(z2f-OHoQieZ753C4}9d_Y2Ux`J$b!Skz2Vfrm9_;K4{D@RSa9;1RjX
zNd0?&NwIbAE{@VL7loQn6|r4$Uer^jj8A7iNKl0wE5@PGMq?>=agoQ)WpI=@;x?22IAV&@0i=J)FJj>AoQ)a3zResHL7FZf~nQIx=1Gc*T{L0wl<5EcY$;
z)3i=XF5G}**Qy2$q(uhtCyOdhq$^^JSmTaFad2&+?*FR)&Y&VJ$VkN}=_vwxQ-}A{
z;eh$*|4uXZY=MrE6W=t^v6?+|88KgQv?uEbduPEGGlZFF^LR
z^nt))0bzazY~w{hQZ46LY$>3cA<*goLb7_Xgdl;JihFze4P>(TC2zz=$AoScnd&e0
z6na`H%|&7EWTgU=417OER@GoTwJ`!715pi;_f0AV%cto;CjH#3Q
zY+soq36OrKu-|enx6${pckuCYVr4m+r!QQSWoDUqP9x+pXc=!oY%aOZ@Macz>QzGf
zE^?X!k}Z5>x;7jBTJpNzbhxffW7YscJ;C2y&s3M=em(Z$yNTf0CjwjS(Pi9bLt#Y&=R<_yR
z$gdS{Um449SAX?unVEQlVls6*@V#kcb>5iK^P1YGc)AOD3K)hev(DA-*bgaznq9|}
zo=(0HRz#Y@AiKa72+j{bc#|J$95|C)1*Qgvg0KNCxB=(6`P@P8h+n7s5h*5_MA<}y
zuS&oCGxVY{E&TP9!0Fva0LtGhy%yv%n7~XW8U|3R9z;YAcoIo$Td{s0#3V&bY(*Tu
z+h#^F=5vkTe0rfXt>_2K@w_wYtMj>6n!}693_ItmqmnVBL1zpJQL?wEE-_3`ChMj2
zobSBm)z=8Z^OM2~hy(d)ze1;*+2>blm-hilcdM3cOKK-#-Uf$VB-nC)zfA2VUiCnE
zRHd#zVvFmf)zjw&_2&;eDk0hY;t^Y^Y>*=nitO{>C_3=>o<1Qzu=<9J5CHAnokkgr
zmYJk0`8`VbK^P9Q(AI0YW(FAQ#0?6$h|yXHK0G)uu^Fm2t1oz_qkHDJX%`Q<06G6%
zTP0N&o9o$wqorg+jHZX_>=vnCX{L5)@CZK@`r-#4U)
zP=ki2K2ub`15}gIsw(C*g6rSzK<<%DSo9AO#QvvUA~^7Byy%Lc0s&nHUSS8`BE5P<
zUbt&Qa-_0xz6fz37m~ng8=$ll%~gb)@hBGX9>*Ms&WP5au3s1md)hpo4I6_T>c*}3
zAjn}PNrPN9vKzbM>8G)@?w3eohQ^t~lj}Z8>6P+eY%)uS7&9gJa1m92MG29vPHWbV
z-z+9Bh!$WK`Su<|Ui}_Up-g}ZP4o;`2`uiJ$X#;*4ug4ARKC*6iP%`^M5^cFg8bDP
zDLF}JK8`zQ==MhE9XOt5vZ0vGaM!v86QLMmm@sGYwy#PyIH`$97Ja{9Bx?uniZv)o
z#>fKbX2wMJJ~D6UN$k5k7aYB_U_Trb`z-3O+Z{TcSZb&)uHq{XS>YeN(3*4`3v?CpUC%jFeUfc%^&(uyBg}n^6nKbZl
znW&t&b{IX1yfvtllD1^qXG>w{BdAjjPqO=aC5MXFOel0Q-MY9|IBR6w1CRG?|5!T~
zJ>EI9S)u;5mRM4_b8@d&4YeE3GJSp|z>7t<#J0OQw)DEV`m}?O@I7ToYLCysRKR*Y
z5$w9{#)pYLy+}FohTKdP39DoavmJVdw+Lg@Pvj=<#d;R4&vv;994r0#^GK85Zk@Wv
zqYihDy
zhZo7qOC6WrT6ocM2$bE3I`%la1<=E6qxnb|3sB{xNJKQ_n|<)`f+_lY&X
zJJ8K-+P~>?I!hr(?c9CZJ0CT+uf%B9*?Br~7GM@G67DXGt>#QCboTL4Sjf=?Lk!Ao
zk9y#6eTv7A+UJwj)_SMRyA%_TTXnzDgUfsS)%~PGx8xs+B$PvT5G}=WxQUxQIpv&C
zx{8i-mS6Vq3dg^gwj9nh*r#(d+O%cZVW?{{(nk2%~df=ve@
zq)^8NIonakO0R9r>^_Zhs6O|`Q*FIZa;60Vd)SCa^ug+t>F9Jlio@)rH2=l5UOYP_e
z%eS{@*nYSDGB%#?+HHj
zt9y-wJS$%C_!G7d9;}(iIZ?1=8l-cXs#w7ci(jt--l4pJ(V^uV)29gDU%b*nvp`oM{g>y(bBP*>I27_>E>khF*yuHE~Vq83OVQOKKQh3J9@%WUVpCn
zR`ZxFhU|{HmA@t97Pb$@+D*}0bfh9a^0{xAF3wL?`(RhtfHJjPv?m=*+@tw4_2tyu?
zYj7WIJP=zO+@sUo2SV$FHsyWLX3nm=TQ|{rcf08ZFqX{oRRX;+G{tS
z9SOGN;1Y1tYek>x8ATausL7KRc3i_OcPqn;rkB;X+y=uGrk1(e;+eNn*i=9FE`R@M
zt$Xak*D9N{k%z~v&Uadr3t~hEBMGHyvSh>6xgCPSvJ+ufMK`bQr^CCrFLR}
z5pqkWL-w87xS~t+{a+bOTKD=r`{qn2SOx3V+8g`8t(oH&KXB3=7#?gLgeOqlTJ2oZ|nr>bTqt*Q4RUe(;;
zC+98(x$E8aldDfuI3K`unX`=FTfap;rx~ado~ntG-%N*Fw2h{;OH{i#7^Qpg8GpKx
zHN9dsbgKW+Malw;$XhSLm+!`p==8?+?|$9UQBr?inQT}y=Y<&0fTYWU9Qg_4D0RAR
zyH&I*G)tAy#?j&TC1(egw%95Q+ptYmCzMTA{c7pE#T=})6-CxAaXEkb#d!5~R>Q58
z7at6szPNO=q&A1!(ub2Ov;3H5*HV7sK5*fS-(w}SN-T4f+ce~1_TI;yOFVUEu9t1#
zVeM_Ew4JXur;LAG@9Kd5WOG8_`>S
zq+ju}T`o^4x|Qubi{|Ye9>`ZkM&Icl@_mnqQjv(YtqJN}8C~z>6{*_I4wHLGj%>4G
zTgx)K*m3gvoO`*j#O(3FVY8D0$C>ZN3M_L9&c+Ovwn~si4t>wZ1xS3owRNN4qk)IG
z#63T_{PXFR5dSAT(R&q4?~rL+nehoecH4A!d>@B7?@_;UX9=TO?fo4+_kLQ7m3P+R
zRrKApZ|>`Sd$gg0F3q*-x8;na&sV?xT4-IAmY$qm^580>oQIPQEzk@(4yJal?tPkM
zk96vzx=1l4<;VM{umsz19j8mpq_n`Z$u4#@0H7QJ7Buf~gXyMZ$WG8N&cooF(WBrzP%*u}uvjS$y
zo+W)dVy??zH=4xJ(iaPl?`C!1#Wh{N6(@}5
zVxyD?mkB@}*V*(>MKP>e5tu@mcV7>|e!tn#iEP!3bYF>MAGT$xYK{rq<01!@kKei`
z*35j%l8D_D1**S|Y
zg;7hy@FJgxsKQ-^0c5g@#W&=YX~T@%ibPpb8j{yvD;%*~T}w%EL5))Wj0gQHI!
zTJy2_h>#)K7GtHB#4!3jS(xF}3We?3bV5ReTuK~YHAhs{3Pao7rBWNi(rr>-iNO!7
zzO<3-lB@&&3mx0kl}qEIyIJH1t1_*VXgt+Zir(rjd2mCVkyY+(-FR>J*<`srQ9)m3
zY`|rzNRddA*;UM#^Ry41ugY~F{PwJ_{47B}-kezNj){isfiO{IQcdXa(_b8lf;0>d
zYJYxi)VU%aJE6I(!8<~d*?PKL*6e;s?CqW^nQbbcj_ll)%{FFvq4MUTt?%om^?Ji*
zZz_022rwQx7C~A|a1l9D64B9I%NV_LzP*C%VDSl(F>I@o@u;O;h2Hf?={$QE^3aDb
zy7k1RaXQgSd}xiXjB6&`R}ux_W&ScZjhl;7s|&FSdrVk5m?Bf{EI}Op+HU
zk9^zX`pzVRk=;^A4K0e~-Um*Ti9zc0_zR?8n$6JdQ@E?2dztl1BUcsW&bP7#=UR2n
zTiY>t>$T#4b(xTlO{PF6`w9&?V@kMvVdZJ{MMLw*XKvmTO}BT;1?l8I<~s{sVHO*c
zatxH7k$(O~V?06yn&i%W^1EIMeC(7vX7|{e%CX6>lT#+`l{^t0k+cmoM0Lim?p9Qg
zoN_O}`t!2I&C;x|!XX;!0Uwfk>+ELoToC2Esm(45yQNK33zKY~RcViGt>62&2VG6F
zVOD*Wc_l6f*-4~_&FUZ1AFJC{`1~S|%Fl+06d4yrHT6bSe!iRuQsANWt0REF+3M-p-dFO5gGYBiYnm1S(z3#_ufe^YsH1T=-O
z8AYDPLvbQ%8!)mtB#G{m8AnFdl7*4mRvledOM;c%#CdOh~iE*pU#hq
z;BTo~6;G;p*gY}xPG{i?;*GY5KF5T1innx?3>H}^bufFCq
zn?$?p5l+q1lT>FGJW>6I7VKq0)@Q1kI7Odqz9TvsvDWu?G=(A7duYS+qU8G1%ZK$Qk_e6|
zx0@a73(38!TGOS=j=fvA6>J-9>4R;|7$<*-vnBS(o?3tTbD`lnucmQK)7F9W14fSC
zQIBjj3v-5qXB26&tCU6~(-$`fe^&C>r_F0bH=25mS^(C>-9S=bTQeQma2kEP
zNyxSH%5YflsLv4n(-OH$(vv^>k87U0VBvaPOUCJ5T69OGNb`$lkyEcJC=O_!(;UzL
zXtme!%Uh;X1VIYgqYLHuVTnfJk>V8DDn>EbZ_
zJz>5963cvXaiJ0vX7f3+IFSdIu==|nrAuKY5{rLzt$a;r&l6mwYJEMj
zp_8FKH0w}bRQHnZrH<-1OI~5W-8s3fPZJ|`OVFAkY3qAoTKjHDDIMu||HQeAZ+FXc
zCPv+-h;2o&<|cx3{G?#n0-)?)G(*2B8wFQ|LIO1oyli*V^#x=Tq6cNi8S@&;U^7ur
zAd%!j94|V$OBSFtp$m-Gk7&0m-_ZC0#-@Zu{GV7<_8}P*A=Gf_sNd0CYMBE?e)-W#bY0JBZrfoU(M(}fu8>s#R+SL0;B)L
z;VdO*e%t1`mI%^((fHqk7|3pXpmwIKSV`xaZ}?w-7P@t_ZS}+1z93Qe|*wt^YA>s|)VgXomx}=c^px?xfLm`!h8E=%JaM`X>JT2{v
zV*pkZ4aIo*O3QqUVio37;4FMl*C&*cgTR}m{}V<&=+mEPgv^B-OBt@s1j&NyF=(9m
z-TrKs7OE7~oj+mZP*htU6ve4pZh^jo?^opfsDBVeP8`UXYmvRrIiKn}DQRon4~wql
ziQ|Fp{LQ>ua_5pH0dyTEJc-JTs#j}V1R^f;o<2Rl3J`Nw^Nlx(4(=TBc~y^FNWooD
zcwUR6L##0_Dbx&?mbz8)8VVxEDRf|%Y{`aGeK#A-i>~ej
zW|UsKq&auHXRd=)`X?6`7j8|A&yHHqZL~>&&x@a@Ots)jTG{yc)%Gw~+gn#$T1aCk3
z9*56ZHl>>E3`v=U`c=NS6)J6f%ndY7Q>tvU_vN6If0;|z{cP>+iNvm%6seav=QQ85
zz-Lm$e&JwH&|f63`4wYEse69Sujr}?W45}QNDiH4?%VG=#bS@l@KTQ#qz0>sAU66s
zZAq54yul{ZJ^h=ual$&JiiS@&Jp$C#Cav8sNS^nos1f;^nqc`(QVH=vV6L*E**$;k
zO&i~<6Qz**65;yVa+TwDm|2>khpTG$i&s^(c_v3sNSxrZ?2K!-KRHj^kx0+F;g!yH$mXp!LZsK@T2wzmRXYpWdY>43U+?jj&%DxxW}f%qI(_K3$(G6P6AF6A
zicA~HLDr%1uh25Q$2V_&=FD84mR3&@EcSB2K!i*k5p%+{(TazAZ6|=75UgBDh@o8S$M3;#nX(>-m
zNr!c+1?UQE%K@b0H$&HTjw;8sUvz!#`=0Cm={d;(6>7~pcYLn%
z$9yeM|q1dgUtDhm%1p}48nY8N(^5$+sd-i6km|x~ZD#-e7aJ^N%jVn@DDg203VIlJ&ymxYm`|
zO$lWaUG5G(DpV%bU&l3^M8@MBbEuvdl{=wZnhDSI4wK88X_gxl!(^^DV?LIazLR?q
z*6ixOa)P+a`IQ*>A?(`~pR^>4+Z0H7h$$proc?()pDOZdGsCF0OZHU!sK=E-|MRnV
zJ{Mkl@BLMS=UmeB=d>C}Bw+&IA=90^DF3C?^%qR74Thymn%(?k6Lzs0d|1eWI)*n|
zo!2GR-rC5I=fiC515R>~>F1{z7WIuo#Q%q@_l~FXkN^KK^B{YaO;L8T$>C7%%#4)G
zP!xykb?k!>*|TI7LXl+eJ*mUJy{_x^dOjb|
z`~9nfp<40x#dE0{t}PhcNXH4^kY7h4M{I@W!&p9
zj#FMt(Q4Y;i#_35F(hjgB+W^My|5@VO^iXOr1PXG~N6KKM5nORC@8Y}bBN
zD)H&mHAXxbebVq$gV6iIC~WG{5B}*41=hZw5#xU$2w@ioosEG0(Gz?O_A92Z*pzbX
zdb77u=2D?GS=;ALZq?cPP@%pgbDpiA*0k?Vc>EtmS+QPwCE9RkXTpCU&%%Ol1S`cF
z_qqH7ApmEDUP_2f`Mk#X=o>vku%@i$<(rXfE=#+j4B`Jhkac*uJx-ioD3it+
zVhJb74s~D6^JDWx`Ni6hT?jFo*h`Akt&paKLt4h~}
z2QlJpbdd|ha0cu~_dLtj@d_G|yQ8Kf<2Jl1!)C%FYHNl=Tl|d0l9Vvsbm&N6iEe8P
zcS+^+wLjO1G3xf>S#}Xh?@{Dqkaw(9c-;84>Ev$JXNHS&iDp5PqM~n81~6JgJ$ua|
zLD&a+7YWE=_-*UO!57u+3$ZzgbJCeiT6M@5xVj$Oy80Saidf0w6&X?Y+B`$9p8agR
zP<40%0<$ow_W)ZRRMv&R!KVf*Hlltdzj-Gkb5@){(fpA&n9
zb%R+Ez2$CPJ$lf|>K4sOfWIyFLs%i|5_zzOoI&jl`0rpFf4xrc>@B8rfiBMV4tOaochcWs&vV6-qH`Xr
zT)cpWsYXG+5Rc<#sx?karB7}D`{D5vyA$oN#OXe3c1I|6{@1kH)M|?{`J>V>kLO#n
zc~VYCzH~DphwE6_W=w(cdbJnbk|o=QL`9YFU4xgm+9)k6%(W{CHQ$Io5|=9yLoLrN
zyR-3pd^~u2aVA0ehuDB^^-=xNja9mtg0ADN`V&{ws>zRuEur@tH>PfrMjMhfW${H?
zldjq~wRJ3iy11jlQ^Ru!VV^x+dT58`iDEdtlLz$mIUxu$gxj8k7|EOc}!)epelOwH6^8w@4u)J7>
zDBb81?h)BYBI*`*%L{=&0!dFRgY0=;T*{^_RlbA|YG3#Cbo4x|z==ozzIW6mvQw8V!9vl
z*TaUMUAV*(6L@(&ixVG;`J_jU^0nJusT;e}l={oKv$3N5y*gcDlTBiss0RAy8#J3d
zjPc}*>U8;(XZda*3r?=BH3$FL`AK!w$o}k|&+VigWGO$Yx_GQ9;iclRFh=*LW{GAi
z8WCOavUu(sR&}%Zf-BYUaF3g$I;a3Y4=5gX%5+A&U
zTk?9$V{!`FX8Y)N%@JVBhSavhLZYxf%bqgREI>XTn;EyW8J#5T#I6iqCN{rR&yc<8Najf*`V$*$)GJ2rf=*ggEMF)X%Z
zbeWmk#yLf!UGk`)Z3izR7qfXLRoa$4eZS>l7E4TN*IQ4ew>r$zHk&6T=64h|f2py?
zY!=s8X5Mz8*}-42xfj${gMaH1MOymFxX4ODxN2KPV8~|qx5Mtl8I?3~;B+=q8}96l
z9g3?zrZ$Ie!?r6+3R~LVOz}HQ9``x-V$!>{c%}Zi?;V@5(uZvPN$8I`C0MBgB*G
zXUhCz%A>JO&MD5(Cr5q=YJKw3&Bx$%dTF#7DbDf8p9i-S!PGV{T^;MgD_hRCSsK`X
zR!c8W$NVzqi-Ec1C$1KBLF4J?zmRh?Y2f$c&b0PS678g}Yz7$#o}|21ffTNq8pX$Q
zl3{KomlZTpv~R3zK76QT*~T*|fPu9-N2h0Zi^Eh4m74=6^ho85M6TvXHV6!yNIS60
z_m%VRabfdeW&Vj
ziT8&6i65rB%Ng?b!}}C~f8-o+ILBu|rs4Y4!Do3%Lm*DT=A|x0FX`jtuPN5M1q2rI
zK9RV|h$ZH)km>M#r1c=S?1z|s_Q}te1czPcGs+lTTqh6h8RyxsY@JQ)&{`8?El%t`
zQ*zXC8`}fPSaff#9QTsUehys&0x6Cc>fCvuELPfLIaS~0g5T>boBxn0)7f5IDYtoM
zQW+u`RZ(2BTmOc^Fqr2CXVDbu=>=~26_kDKF_
ztnF>}$v6cQ_?O~k7_tlMgaUW`%}y%
zS>N_o)l=J0Tzz|sf)IoY;lHti4e?6^gshY?vcxri3Mt39u<*O?l@9Yg6SMY}BGJXEZ`4~4
ztV5#;PUJwe!QUNcc^EO$L=P$xEf3SLWiN4bk78s?5i`%7)V5x^TwT(8MgVnmkyOu4fA#bZGa$ALDDGex`u|Z{8DJF!ml(QIj%J#O+@>Vf%u=`V*}cT=%r-B
zc>o;}P-t17PD1hsfe1RJuv#f}Kl1~c92V_U{LQb1lQM4m^soXrX~p;
z{Q}a1ek1XkiZM`VYYrKh=e^U*+AWO^M$1dPn1RANG}^nbROCyYC8f7|UZ?YlX=jF_
ziMr$weQy)~MEJVJIgwF=a6QDWch@>!fJ{>$$_q9R2S*s?5VW%4X+;yMz!t-a0{+wD
zX~Pvqd%>UzkI?8ZW6xVrDFN`|Iz~Uy@a7xMnCKvqEIRz%Glav#PJU8U(B11bd>poo
zE_Ay7$<&DT_s3DEtC%X+mf}aSc4cnt=j_wC#enM*8fB9G3Lyi`A>uc^CHqeJAg6@M
za!#8v{#`o~$MP|%GfZ$KET|pXdirL-WS;q%z*wTB9SM%(&!!
zi3J&x>eUJz=64)!5x4Tqs3j%k;p3LCl-d0iaHE|9F$AG}RPf?SOyOj!ml0<)TS6IL
z9|n>POdN~M?6zAYs>-11-~j4g|>>GjzDH2!n_{V
z$n3NkY)0r<@uh9_+H&oWOVnhz93Q;8Uogqg!H?Y?`7Brq#xMrH735BZMT
zXj#X)w;`xDJs>kT}7q69;%2cs-cbgV4SC?D
zGa376WZNbkW9O%@MhSNL3iDFIhVF|?>(J?EnEz%EX#AdM?eQhI3Z|M$Is{uPATU%f
z8KFJEAzl${QC4b#JnpY5l`Xqk`F;ID4BVt8&W$AD^|FNA>-HCzDk3&Q>}d|jRp~JV
zPAQAwL*Lmm%&#$_%C?v4Ux{<3;$}mh-csI;&nx0mqbS{c#zVIz8{O|vV0OD!OG1M|
z^WpCTl6b$^1MsD?ybxSrcW8~+-{HaQlQyyA#3SIycP99^ZKnC3EA#9rf0;d+`kplN
zJttmElTCP%H#D@B)RU#U$I`(5f|;40{KM%H<`Djq!YCG_3@kbHiB-|})Q%()x6lz6
zu|Vzd00-*`VpyMN9p%;&a*`1#+_Lu3N@IAJlIEB4@UoGpiae(g?$wDFs#_Hwc5=&K
z4c8nlaPlyr#NjyRN5+iSQPu9iF9q2zq%iXvq{LD^vY3e#T3@dn>ksHdK^y35{R@FRhqn7G*A^SH?8ohs-slB|r07W<-)cmCUKqV4c?F1fdF%4rM-E>6Tmx
z{!a%t^O!#mU8Lyd^@=pLayV`;!{5W!YvV|>
zL`VI(a!nk*DkSVIj&3d2+d36LdqAtX?d|@s!4SIvj*qcdD{672;nL0Y+E$v?j<@jzjqOFVUSruFV_U_GRQK*
zXV%r1M$a{@l~c;7Hk(WPV3#6?E-dhhzs6}uko`hv;zl;p!6~%9Zuhkn$nPr^+oS{Q}c@Zf}U0V6UPF>TpVI!)4bNd(*=;@O#YYulhA$!S%Wc&!l~
z7mHhkW!<}ARI!R&ifGzuR5xA@=iPSf`kh`Oc8?1?1r$$hLPk$0E!@u4_?&Hp0PH?=OcZ~@yI4#y-(d<=&y}&
zheb7vg{1!;9kFOz
zh!&MY*KUdD-|2|CfZ59KXbwhP+Fruyfnb85Cv*9@jTsxK^cA+Yqn|_)M;aa852+=I
z_i6R>;MSkz(P=?J_!=D43Jpcmcm?(HqCN-9xml&DEzc<;AH>2h2Lh$N(}
zZq*k0-wQvrY`;&q{no^+g#-D;(R?No^;gX>~yOjw3m&$G=+2CbN%sFqQf7}=Qsf&v|VEzDdsdFMJKu#DewR4-{A;9dylq58gu-BYe(sh!vpoxg
z5&B~;1l=SWK_gOv@WK-v;$;4}Pm9PFLS#NUx=pALi|%hk7O$ARGb{X1lc>=Vners$
z4o!&5)cu|-&3=f-*-_23fxKt^*|ZgUFjV+AgC8KG-7nm2@HO1S;giBv$vBLuJ#+0s
z*NnCxwSvK&0E78U*zeKW$cs+j-IJWT-F!i3$E<8*zeHxP;zUsFZVtJGTG@8@-7WA|
zvLk1mOQ(cmT(xv>W|2uB_WnX6NYA(lAS~+x)?b762TL`_iny12567f?;~47fAMVb6
z6?aZtl6yh-6jW4a_Q($U{aa6T0tdXZxhftpnGT6hw1eXc11h~B{2=pfXXWhO(w}#8
z)2d&;k7V{GZ|xf2=Xuf0aInB(Y+=_Vhdme!^S6E{D)MnkBlFU}ED#?U2yHWQh+;LJ&<
z#D0xu{xP3>e~_jX=khz2;9`_Rd(-74-E7N|El&jxzr4n;4zCTY9X+CYlG@C<5{1j5
zjc_QQ?iyM_JT#l8A}BQb0$c=JuhKmQmnmp{#!2n-Rjs)(Ic`V$vaPe1%rNR(7brXU
zEPh*H1>J?pArB>)+vXM)8eiI3pRlMtYxyQY^-%#}Q?l6CW6Dt%)xAN67zEKZWxaz95x5DmHxRA*hJUZ}q
zOiP(Om832tbDGdvxIQ6E`Eka)Qzp4gbloME{aR$2OT4_IR?3R3i@Hp@ztCcc?$#Bs
zpCsz5<`fkOl_~4-cwT7du{BX#d5+
z^jf|<*L!m{?8}$2np(N?Hb33#6M`-4L))4zniJ(tfq(n&VJ;{;cyO!U71}Wx3iyZM@%j
zR75a57yT8+njI}UL6#HuA;~?@pBZboaF~*r`i@D}dnrIu)wm#WT7{c~BDtznxHt*+
zJV%Wj%@3{{NnmZlE1U?Se7WpN9Ewp@NY9hP*N<~AMMXLlz(Wu1miW01Be@)QGU6H(
zN8X2V7`l9FAx?>FCPtI=u19<1VT$;1x-BQ&83q#FLx3&knxuQqp$nPnCQc4r0GIIXYO2gt=mB8tv3E#qDbnC^%8F}7%-1qm1@75Q>ZH*rimpVAspqe4DF3Ss_2
z!W?z>qP8DR(4!br1kYll{s0*+HV1T6(e?~NIh@r4cuz>m+Bq;SkNAe~2uSAqLPHO%o+R7a%_=7sOU2J23*^bs
zG|l4QIynNJ(;BY>EEuU9EecF#78LGi0Fqji)L_3(>tY-m0YFF;NP>@sa=6v+-vT8S
z#Zr=NEeN@?-*wY~kt$X`XF1jiR`P`5>7nrF|4Q78o
z2jrT^_)}n38%oR46oC1kgeqxZp0kneYJ@nrq4yw1PzL|BH)(9o?@Z
z5HNpoNU0_2r5Sx%mXb3(;Z(@2q<%;d+(71RklO+orr7YzfmxE?b0!>qS)qWa
zTWMWIq5GDw@)Z1hyp!5-DOS#kJ8R6pwRw=4k=6^42SOpt{-#i%dDea&qwcZ_;CGP3
z;lmaq8yqnRdX(C{XH}Z%mzr`DvWMH#os9PJ8djzk$Yqji1^I_DLseBKN(h$bt?
zAONeWK%SYct5mo`54Gly@{t6mI|0Um8ZFUpq}dRFECiAUwOT9R03)nvVFV&fHaR|9
zZ{om#=GON1Z;lJEx0?oelFLdno;;)Y-oN{aGoG?bBpEC|1%GqABStYha-JPG5$OSf
z-j)smnpr*emn5c3+Gb9Ro*GX-G<>EfL214c8czY^sb5{BYg2zFhRLK)>J`wb5)wY)(Ss@V7jagMM>>
zFH_eVNyYF%BVUK<)27NC%bDM%o;_kToyA3PNl4==|1HCCRZi|oLTOOz!-LY7>=9eu
zrN}@rd22HIu8y)xbSKa1*Z3sL&hGX!mHYSL12`PCfu%w)MhO(p+w;WBfK
zE$g5!L{L9{2M$N!JnRT7p}>m-rXf!qwN;G~4~Y`~i;SI57xAJ3Rs4F2AG=PnZX~>X
zm6%uC@~}HAsFedu97?qz=;7{|5;xr9IN?dUx$mP2uf2iMx3@h;6NmXk9oXwqkyM2+
z1Bc7R=%rSwPRk8gXuOLm%LxM3IT&tzM)|sKg1~?5Q)wAt{ok{b
zyFToN!-VWF%L*szZ+4F?{ou=-6Eo_Iur?pAydW~Wu*sFA0y9RvBIQSRbm7gkf1$F0
z5V&}L&@;Bq-pUhdZuP87R%zd+ssk6&yrg2KqM)-!2uwj5D%N>*`
zcM5(qaichHWBXSyp9Tw%#vNquRNELDG@@(e#C%@XjKWbxOZjo^5~_xnwE?DTGK
z53sFFHeT|7*-Zkc2IWE!15aBLNAJ_o`m11dn@gvV>ErJTC-sm>PghRH*<3Z;+0RG8
z@!=4CEuY(J;oty>PZ?NFm3M;uN|2pj+%!jfolRO34Nf+f_TvPDkPxzH$0&=fhezNo
zucTTxQ8f#ju*0!THC_k%bzn)P9(0(8%vCn{-P8-rKIInmef@GbgrbKp;Zf#kSZaZ_
zYPm{ZOE5;JjspUfZ|Q$yXh|6X68)r-fij)W_Nx
znb%3ZwmTc}S>yX_6>Sn`#i`ble6I{nDl8A4zjU)~DK6~zTK$QNzvhry|Ku}`-+yJrWfg9Ps9Bb$kKZg}!w
zxYM<9#kjno>m98nhic1!25fl{e8`Ko(&Ms;8bvaFrWY3A
z9WQj;`auC<#lZgxd1I*RBtg?gtcRYo&ty-hBfUJDL?5bxqc3f8#+=r(7GU>BmBpF4Bd
zpD~SX9XP_(KeSvIyK1tCyj&K>Z|HH=I5(suw4L~7sAZicr?QB@G^oIbcsShrksZ0e
z?&z5fyPo~*n>(A-eB*odLmQ5-&mPL#o$Y39laM$40Q@~l_rK5;wTR%qkpB5E(A2`M@0-v?=Ibigzj3hK>uv3{1Q+xOsCJW1(#Do
zZ0LPRJ!GcrO)cIQPY&kWAI2|CJ-V!!lO>ZXM1NqUPG-z=OORyO5+hDv3B|DNv^~1G
zur&AcEM?hIc7(&M-niJd{v@D7i>bdlk;Ufpki;?m-NUPHD%62HSXh@5GVs!G1ZxxA
z)Px{6T#&B)_JZ~i{Ou!S>b&!KG#~_qjj+y29ijfklELo*xk7&uD7FUb>5<63Ym(Sno;Ak_Qi>7L+hvsf2b(jRM3u2;U
zyE*k?xWN+|d~_qwbTr|!MxHpvYc$>zR1oNlOd=HaJ7k3;!6NIbV_Sh29
z9WH)1Qt?XT`SX&lNlb`Ll;Ff?7G{V%fDE#*LUqe@p;ng=AkPq9r+2v8p~rwco34{Kc1EFNSs;%Vd>u)rzstn{~Ssc37n_
zia;75Td}xjII+Kx9hsJOHTaux@26f=NeEry1i6aqgrc9(p@`3keqf((b$EWwSd$=L
zZB@Ow`e~|2jA?JG@l$hYHfFQK+
zA~Y2ICwUHOz^i@Pg9zV$Vt2Nn%!HqDE(q+iEWfkyn*lX)JgPBP`EnWc_=b63LO|lM
zxk&SC4HF(aoi~QFoGAy{c#(yd4GcMf-_N-5PaTR(Zx;tWU*DZxe`0u|QC?kLt0&h7
zLq+Od^5Jk(0j0jNPvT)g)E*m!GF*t(X{D06^5QSwSi;M8)Z`8sH<1+k#KyRs`6IVe6toTN)x`83u<8i!({BqP;-F|a;5Uq#eaG_n(#gU4W_~IsJlSyrT_O5Dy;DlDDIc-+HpnhU;_AT8uN|}02b;he?X}@uxQ%>_-a;IbG`=+mP>EEA;3D~|
z?@+e7RU^{hRI515gvgD~LB%mF7tY(}~+I@4~9n>OymZSPMOUn#HAj>w{L;%h)ASc~4Z?lqhn
zbr<}7$E_KwGdS67j1_MICdq>_U9pPZE+7t)-+cuLyT%
z$o}NX&AND+sSKfbOf4Dla*NFo9y*2`ZjYwcAAZqe`wM+=N(k4(Q?hIlwp!rX_RyK;
z86}tVA=PP(hv1pCnf7DrcaLVL=FN+ptl#Q2hHHB?MnGv1t0sZ^W>iKTt@jQG2`q|-
zS$Jz)^&)Kqrujvh_);c9eWJwqfXjYF?xtgLKhl${u^^+58*gb11LNFq{>GEFBO^C~B7QL;qe58r+pdLhOjW5DHQfvzp!K?cfz6xUg?OQu0
zCT79|sKj?@?;URT`G^`b(%w*!Ihr~#B(UTDY}%d}W?w`;biSB$v8ss|`vZ5&F<)bw
z?7K@UD+lJe`V|)!5^Dz`#P~r9
zY>aIw(jC)_407fgqgB@#?kg@RbyeZX47>y)uxyZSt-jUH+8$CSzOm$250F`IR1?SsT?S?rA|S`Ej6na++!Mw7+_#4lg%ZAjwOk1SU3BkQAr+T-A^%yl7XcM8n
zJ~s+@2n8n80;LQhY34+j7;2sq^79YQzTi?6vj#7*{4$Z@11pFef;@=P%)0?}_7`#DHi8WZ?i7|3gw2QR~0?
zw|vD~-~s9G-`z_GpveYyh&70k$J9%4pZsmx(>U@pfp~H!$DhE)8NdJGox}Orv<;
zw=cCFkU^E_{_X#3Wf&nQmmR6P89WlsMGney5v80<
zp@}~5bZ;J2)me5lJ*}h~<-2oOW&F7b#G8I<@EV#NK=4Cq*hWQknlVfyj#i(wW??=1i(d%wD<|-Q52eXf%k-_cU6#)nP~a
zd8;De;s)=HjU=kBmRwgK3iN2vbrU3Naouy(j7$5u%nYdb94-_9VRt2Az;~n+Ao+Az
zcjjOF!1sW*rP2*RAKm-qvGS`Uz;{4P263UtgU@9GA3_uk^3SQ+IV5~en(JCB@Th+Z
zkP%di{53RvyvlxoH(w42mo-_q--cX_LRCo4Nv^uxf#^}x-7?I`$}R**qSTsmSrws4
zgWMJ5-QVe$|Hmi*CTKC%%!1o$Ql~=7Po4KV|KG1J2|1wdYx!D&e@{guQIJTM(BK+3
z@{dVMO%j{@PXj_3aN^B8ph=v(5*^wvO;5ycl@7VCw{~7h)R@s(F3{6L0SgF;O7Bs<
zcmBV*)tu!J_jJmZ)k>hyP=)JH36KULbI&~>G;y#0uYg8HrEB_OxBgB2_MGlrJ@Yn}
zd|Lv5@S-eVqxG1bj%strF7$39C5DocG;1(SU7+{~YybeNHt|jjRgb
z7CSdFXvdxseL%Bvdgrr|u=RthmCcgSIQJffP%8TViaY1(5vgsa?&lT(P=AzP=ziWH
zsi!owxhrQte@S4X5n}nr)dN^P&cjZa{mNNIC`nvj2CW($C_?tkN-8Jk3Pf2n%UPN_
zExi4iLk1(ob;^XxJ7uo1wUPvD|C1iU$ZD6F$=8hb-kEW711RB1W4Aqo0HetwS;%v!
z9w>7&XZ-Z(huD7KOgGQ~^ueBEx`Ef{^Q7{$%1%NOyM%@(u__0-PDDUwKx&{cXll3&
zcM)pmIvJ%C$O3D|>$A>fl31StZWt#ROeG0!$ss9)hW?pThSZfSa_vD88nSWhRT9#G
zCH&JJrR%Z6NdIT0Zm)QdC-KOA8?dvYrNY`lj|uz%$jO}{c4x{|IJgfu~7dYqU{(rwaM}?jp?LBh-EITr{I}m_>#!CD{
zR5BBxT}S>YkBG#ZotED8@)~3$#QS^*f%v1DRA`mt+5>w#v?m6gq+iU
zvxxZ7uVhXaU3XP^&O@LJ>JfP)Ke(rLnHhS+5w1RxY=hID
z__eX2^AF^hgUOQr@THq(w2?%f*G0&Ecdttj!c@bbpy#O0JVdNR2auBp@0u+Y*cuHl
znR`DUMu#F{?w{G@*ys`&eLb^YC^LSkcrI*h&91!Fp&kT8D*#AkQ(kWKQlh2{iXzG|+
zfqWroS14$T^Z33d0DJz!UyfwX#-c~8e;xVsgCf5L4MdH25m%zfbxfkODA^sVozw$*
z-SiRyUcMu=dT`$%4xSruga0u_0e4$)w0>?a>)?LgQA+b5%OL(!$Fxx>T&x(?{3xw=
zv_eGj)kYDbmaM{2G-kc6NSa8!T^Ww;es$o?XgAHWH1`OvOF*WF&>h*jjStN^a_4@_
zR7K2FH$IqnRI$p}7^IYq)moTVH^L#dRB=+ba8J9g;foGqyrjiyXl_j#y~$PaB_aLl
zamJeL#d1nTTzdu0*G~;G?P#qZ%bZYX7fS
zY5u0ienE6R86-`e{kGcIG9`_$r$s9-ppcV6>!(jtVZ+o(1HP8y>E}#su-+oHpnJU=
z(Vlsv)Qzs
zAR4f<0H9L3D{qtVv15ejJr&9tjMFl6wk$>d(MwXC+Rj^`OKQmK`e*Zc8%qR#lJ*nm
zm`H)mOUOuxGH)4-|7XZ^*KVjlWSab#*s&V$X0Rg)?)JTRZbBofutqL3M^mi03GyaY
zf3`acIR>5ezhkFA&ynAG(^SixRCnnmeF|)fIG>>Di#SVt?}N5e5x>V-@<`1&yn2av
zZ(}~4sI?n1usM5lM;?g1?1-9H<4j%VLJ+3+lA^JK{O2Q6L67$DZq~`akguZgMYBkwsfOHB|I)<;0#my&ZoHnZ
z>w%krKRe>3B5CH|;-qw|R=U19c-GM2X)>2rCzC89Fr|2rj4ZYvPq3w1x3T39un%*G
zw?&70I5FiWa}@axxIghn3~~R3-iw7OJ5_#1mA6_j;C>|@)5Ul=D7ovTDdsc=U*{Vv
zq@|_UROc%zRj2DrxrBR?{UL%TkmJK&DB@b9r;@`WPj_=@GuOU#iT_pHqqLWgV{U6R
zi&y=E=HMGX#QofU&Z=2IHaZ}8P%uqeN+U3Zf72d^bvRF1cl-&2aCai5R2>iBgf)T5
z87Jl=tZmcXDBs6-JcCFPm^=ji86*Dd0;XW=$2;x!zbUKa{na_;jm)=9u0MIn`ZGEs
zuEOjms#UL{{r4Aze9l{^GZWSKS0&zfn$j(pdangeS*2i3?fB_(icdw1Mh>=aj+zcT
zUmlY_Sxa^Bdj(2!&my}Fhlf(0TDeikU-=xX>&tzx#>w{@!157&VZhw7uByNvOGb)T
zSQ8}5VVbBU1*{JPLMp-MgZ3#h*+4;QZxMTw_KugI1VxjrD!fA3GFG6NU1Hn1!NLno23OWNTRGvyS7AH1eHazxamzocZNQRD(ft*
z5S@AOegPvakW6}?a}Tt^8C6+X^UmKi;a5>kA+i4)Or$&Ajq*gXe#6xWabG7503w0F
zW}s7WCQ&FSW5wg_L0+uuOy&Ow5ptb$6{H)s&vz%fs{dvwR8v{`{z4@h6y`*oq_HZ$
z(Cj|)YkmgNv>IJVLQ+z{w5%GPDRjGcIZo?tT;n8Ai22c0M4$UO$R9ET3ssaS~}J9eFY@`K$m}-U%=)
zwtG>WBGte_Dd!&rhdfK6;ISr*C+z&T3T1!|)I6G}8X>f?uJS)F37h?u5zH2NaEP+~
zX4&tDpg5!CT(@);-RRa;Xj*t7`2y>AYeIP;umX4O#7h&>P}q5F0*!*Z>+)LD0De($
z%@8p#Gcf05=3Yq&M>zoxpF7%)2<%QPUVwEV)cuR|OuB~an0p_3eHDc;r6gGs90VsK
zt8>ky3hn;}I?k06`J^krc)$n&>m4~7;7}MTI9DPsgbk>x`Ix2;K{u`!d?tKio!j9T3
zp>e(TGC0XXyw=Xlz;$p=#~wX0NL1(sR>lk~LCi=@t^CT!UQMD;&N%}9QmGP>+}JAa
ztdjS#3yc{+$kRcdWcl5e)(r(dwt`*oKnzW>f{!AdK&+CGr1BqY;OyMIIr6V)!1^z<
zL*lKgTHwejXAUwufr1c;e7|CW1x4PGd$#Tvy)N?*-mm8OXvEKW-B;@YHkW$
zI++3kWzBN(`2vssrFI7Awx(lpY0gdHS@Klv|Z^;%C}s0`5Y0`fB$M4bpTl(o^I_VI1FZ=~9vo)ANcq;Va(4
zE-HL{q5Xlm{yvuYuT6>ym^F|NJ-)oLD{v^)a+(VlC2AE4`U~aV2XnG0w)OF7A>L(d
z{>JV>Zs~WQPDq*|re!Jqpod4xmDd|{Ur%)qjP|+mGXzKJoga#y7=^D1{_~|kxTlUZCKZ27{Q|D*$<%I{vYeznu%!75{a}S1
zvS>KSV|yHeI37cXwQ%Zdtk8m8TvBXArQrXQ8{D?B0-H^EmORi=p3KF-A=sA<4d60h
zhO7lXNQ}hO%z>m%h0C1l`CsNh)XUP_nW!sCKL-L7^;PBX1LLN?ko@yR_jK>kMqY)k
zX$S}>1B7HlYPyM&NU5`sQ^M-tpf0Q~GM~J%2Rz7Jh_%<-=He^1PtZr!ZRHKPAR+?R(#)&6mx
zF&Km)Dl&|s&?Y-sh89~fqEaZNvXs4S$qXv8Z>1zNJqVSkh>)G3vNVJc!&tJ(;cdD8IxYX
zSecjJc`#3S{~OneYq;wqblUK8r|pz=TvlcV`!XPQzD93NjAdiofT!k5FRgDuNDG@o
z0N9(hcTN2b$?_RQsl!&gXzqf9fAQI(m6-VmkQF$}
zdvNd~URx?2lI=gJV$R>Ms&v-%?u=PN6rjzCp_M*H!DP=LVGBlj
zl0ih=A_x=`1{;tiXaeNUAb&nJ^fC#tvNV$m-`K`OZrsarMzY}~@fr~Ls$OZNfJilj
zDHP}YPnklH0w%D`Lo+-Hw@0NbGsE$r7SJ~G)q$wbyR}|Rw?ST*&Du>-L%~ln`nuNC
z#Pk@eQz?gV=I;MS3MsBotg{e>JM?~n`s5|RIn=eeC%rB1OW8z=-gb~G4Ucex)w<@{
z9n6VLtWPic)RR!7GAR!G{gpM*!bEzw_%;aZlIJ8ou4D(o!O2==x&6@fT*Ony_B+>n
zg4>iR{$T#Z8M+h26;`b|3B#a_!2>EfN+2=^wG14>f)w;YSeUged
zvC7gte%y5E+_U=;CMdT}YLzOm(}waXJ_wDgNow7_qmxYWY-(?u-Yx3+Z7Wg?1(nqE
zbhY|^=Ihro4yuW&x+a%f&oXP4Vj~lmoe?{pcwF8kKG2unRCf726(?Pd2|WeSfxAr0
zN*1BU^~JsGj&}~7h-JNgDbiAwUzn(Q&FSIU$d#MdB9)}ZYS9%eI0Lz^$5dx!>FukQ
zq6ueic0V09kZ*Cf>O(Rv=A23=YVKJ50ZU%9-NUum@F5Wpeue_prbwoeh__;Jdiv$V
zpMnEwgHu}*9h_|yJn-4iu6{EE6~o?%)n@;FN8ewZLw2+gg71%?d@AD}>X?-0F-F$Z
zp3Ly2_^6rk>iX=h*Y_Q2$h0|?TsMY0WR&2iL!^LMP3Qo{R-UG0irRkA3__8`!tp*9
z5c(!0V1bF;hJkayUKxonX>M4@6kC>|O(bq|FIn-&GpO~za&
zo@H{GV(hF)x^KZ$9RvN72J&Hw7a{)SCrBcP?LbLaS^N1Y>d5I#WB_VuT);-*;!3hv
z3n&GH%$xM^)=|9l9eJ6DEG2)5Twi;)_}&S+D$vgegX|j
zpgbOF?hHKctW!Ha`o{!}l)HlH^-I(Vs`#J`%OvE27Ha!x5X?-38G0E3-MLzZQEKA&
z@uY)bQHPsx3v1^oKUM!sO;G{q-9A_lFMMgU*+iUw7QVAwuV5m)wi*0}wsR_{94xom
zsl`hL(XHlYBsd6>=be4frUmz{TkGU+H_DkLg0>1X4)fsKlsYq%#C4ZbpCg1oHYsXZ
z?!urFCV{efv&l&YFpw0a-b){{iINllKCIOS4maaO#HBDqsY>}sl=M_U+~u2pEDZTV
zCV@v_WM|Rj;dk146KOOZP}{57Fd1YPSPn+6fdK9Z5U(ZF+J((~l;M1(LfkI@0FV0C~3je7IcS6z#F8ns>C=U-1
zQk*zxn&VGWM}a-BI#e0%kP3*guIhfK&_q(!L7)#ne!%nS4iMD`{Yg4GoRt}jtxnNp
z5K?o~vVAbc-B7y@>`=jMtnY5jfk_s5pDlKR(l161qoGdO
zY_vwctMfer$&5Sx5#cqow{+Xx-hRAsF^hgW^hn1
zbbENY7Kj+Kw$zrsX$5Irdgop)m4NKQTlxtp=p*E<;6K$Ct^PaShbt2}W150`zijR3
zlO#^CTRu$td&?9VxSTzkU#<4gWDf6#bmc(cTRcfnd*OHq743
z>Sq}9!yk@>otmo+(=PzyWfBBF3&Zp>_(-1;3wQIcB`i^DLu)$`smve}Qwb{r?{)(_
z@u+#3nw>Q+g1ClbjI%|T085UnG_!JiG%-q<=}en!WN#sdmVWrUH1on#@a1Xl)T1qx
z(cIywt#w6_IMj-LL|hDV-$R)1nhPSUF9gRVu7FFHZ=21
zHhX&qTcC7+@+!vkHN}PT?9j^y4%BKbime8*Du_nIY5NOudT=
zENUPyuA&)x+Er_Ku3x}R)l(Ch1t-C!z}}1J90|j7{bb^rPx)_X{5}*$m3dFHOd~Gs
zUF{vIXZO_5PCu<-_6+a_axvLrLe+zh8JU6
zOr`VNfMFu1K}>5(feF%8pS?8OV>5BsN_>%zbKFZ2
z#aouovR>9uUg)z0e$+5v8_(&?dI{R$h5X`wJ;xX**Y`D0vCTD`fr$i0Xr#D%*^%e-
z{^he7KkTFBRn0_T&raE{j{2s_`)1vC{k50P|8!k5>3NNoZuPcULVif~R4CT>Rr8KqL00VitaAnAwposp_
zp>x(BeSLjKTuxQo&R>){Jx;iT`h~fX*#C(w@a9mc_!l3}3+Hqb*m`n8L#YPvVK!MT
zYVcj(XeRJFVwL`yWLI(ClhMW#Ly&>M4*{z}%T(_OiK)RELrI7_Pw~Vx
z+D4Z_Y>b+hOK6@FhyQ$u`uB8P)Zopysto$xJh*5c=I;)3Wc?h_c9>Bq&tg=%Se#uG
z=k}5sh;lQ)A^Q48hs7}Ve7syUV*cZwm-ugwZH0sHAy`42UV_K?Mxy2cA(#VBedVa=
z|MUrmRGR%i_gA
z2{zd-{S#m}jc(iLMu3&Z1#B{VBQ|s>^
zEJUyVRdK)sSIl`sFTRHj5ojlsqeQjtg()#Ny)LIM7j^jc1lSe>7D<)e_4gH~>F7p!
zk|O+s($J#y{lG$JQ_WR0zeQ1yfdiomw=j=_Q7Wj`kNv6Edq9UlB39Q2#a+b+_4%sh
z#c7{H*QyAeeF2<%hx@`Q*>#=7tVTntRIxT#B
z7xum{dOR9xlV@afD~}=b3pvs~qf;IA4${p^qZmm_yRG_3E#=4(l46*&bsw)sTp5sD
zkxUvBzzwBZ3lvx#IHdvdhqWhh0|o|P
z2F8TlHOpnbShhuYI~In$3G+}dXh)h0gDSI*QTpmUfTg
zrs;^Yvzqaech~!7jswczMtREy;s)p3>T6s<;)>uvqD1j+LuI}L@Xnl!zWcbJmM|uF
zH!roLCuyUMqJ1FbM$X?DOrT?$zY)YF0X9a!&!8sRlg-o(S^NPCMg~9>5zyc0n!Hbl
zz(H+=R73tcP>In+LLLhPl|JHPpBM$OWzT9ciW-7`O9^iRF6`06X
zJ(R{=^PRI;uZ0w0rnQtSscr*!rJ4fqpQduC?T}QU#F&*|0|#9rCnZrqJpeL;ry4@S
zI?pa;Ju&g{)dAIcXm}-22J9psMgkhecRbdsyYjB_%}#iT|MBM8(X=F9T>*DE9Knm1
z=3r}Pz&Odx%4|u!oa2)O%eC|gvGT#flaD9a>QGMBT^q$izVRi(SceeuX1prQj}^%LMUri@Ut>1
zZMq8iGbgka!dvo|R@9--T0j=%GFy%jsC!s$kh2}*r<$}lzP}q1Ff#6Cdwo=vE
zer6LHQgWbPvqMtPcWNkYb00h-@IuHM8I=O5Ls1(>gD`I&oVit>w!WfEB9Q|<>`oKK
zoAAaR;BZj5nHx%Y3uw#AQr=`&qKEeo1Tz#cM!?k{8xO)PSf+U?MiP`xXrP?5GxC8X
z5g_Jk?{9#qF&j}nO2V^THvu<|lhq;Bm2pAK
zRXhi;(r05TApy!YbMIHNc9vi|JRE_2np-!}3#OPo{*mNJ4xH-{@xFsAVUw#k1dsD1
zg0-?IDaoV^*da(InnM=ltDp<$U7yG_ltUEC)=>y#JiL5t;*SWP=t+Gl{)Iy_BEr2t
z{iK_laz!o*oAgBJAoR9F2LovC!n_u60bdJL=F;+lA+kONIc3U`z#9#IV_(m~AL@h$
zFZ4$kodkt^MLBhs3F-$(Nr3?XJPB?ezzpT+-<3E>nnQ#yIrR!mX!RuE3@2Mck-Z2q
zetRuF!Tp=QiJfzup71dMvLS!WP?E&yC&FW
z*9AOvbZ08pGA@rm!pN0lQS8)QDDbHX)znZgK}gv#y(m`#^z2c*;M`c|#UxlW)d3w>
zy(jb_BUs}@O(N4do8fUhS6RR#?cU0zesvxer*U
z&`z4_A?pGJaEH`7j5(wVmK4bHptz5F*o9kwuhEza-kqz&j3IKSbs|Gc6ES6KU_%<=
zY@ghq)o4H*{J@LdK%2F5gpwn>t+-^c0K~OA?j-CU@-sM_I|(+5m+D@c0+`(34a8C&
zZ(ZHg5fl(~$x3T&_mPv7zSf&-ai{uo*f&W(&;q2Oo5Yz=o?_$M0WD5mEMy;
z$e&s(P22w9zWI46GtL;;7^dhXM|T?e<+s
zA_Pl8>`sFa(glwcXmjS4SYTj*>DLSCdM`^2Hw|UALfUf9n6kq}5*OCe-6wz+Rs!8dQI-Iw6R|U0w&(&6p!Nf*1Ja^I*itR#
zKRXv_Z(*Uyke_X&#SUYTDK5x`@G$D_>cL|qLvh5j?A98ZHRi62GQ(Shyrdi>R_B_3_Q+48
zN~`j%fqO)J$LE7yaV2m^;vudYA`kTQPzFmV0i+9cDfx`_7EGxsRJFj?Wy!t>tqV+Au~gsARk>Pi1dOZ?NA#
zEZ^TYklgpi{tkjRcy;N*kRl~Q_iVby(Wj?Epo4oCW7(Td2`~*9&LDvE<}gd-{@T$)
zekvJ^cIDqtx+A7@ty&WNgP`2aCx1!i>KJAKzEbd|<_a2qEf>0xo%cxN^RLS
z!)0Zi))kriBV$1ly|1$MALN0({N!q?IP{VT!7
zsejeo)W7=#CGtN0ncIGu9jQCt5T{B5HCYYbmCSPtA^P6Bux;<@%Cm)|EL+2#rI!BF
zo%8c!U)4V?pL`IbUG#aB|Lm>$+e&&OY9DU-uh_pX>%v_Dr;)8a(_EOlOYftlhi+cF
z-@#gH{y?FiBA4@1Yb^W2D>JukT%t-BkLaL~HZ&N<;d@&TnV#+^to)!*i-A9apE!R$v~dFC>Kex>G?e
zDO)b0%ko{2OLx({@@571=zXzj#nvX=akf8dnin<&-K@vG4Qz@<3Vk!ltGhuAN?0)+
z<=*RX-mYUG*+qfpdo0Q~wJGc@u1>BwOqKym261oZa7)2OXELd_{LvZbum7yno2oxW
zSC<{O2<`Z8c{48A+*DT3*D?9bZ%C%<7Cn7s?5hDR=&xUPomJA$Eb`CV*|`O(x_7Df
zxH+6ZKUjN)(zGXgi@0Ie>x@(ONAX^H`SC>*=@TUdie!|Q1|Q1yNuW=yScH9
zIlle=D;{?V)Ov;#dT%6Ox@?kqHr@K_PchYcYrnax*KKvV4Q%8ep3lyoyz4M1
zXKne>@)H~B5=>3WEND94IlXB+
z{|f_8E?S(|*u?uj&hm@H3G#6%qVAx{&H}?}+ft1OrF@otZypTO&SvTOo2%H453;xh
zB7}o-1T?9V(o=?06$wW+V>Kq8OiuJNMT4sng_%y|uIR!WSGlr_KU(H389Bb20ooAT
zQ7lvRG1`gog|N2C9q?O(zJ~Q@Vf7|zE+>%2ul=uD8x^P(+I|cS;$yOM1UqezeYYDg
zG>9~^l$ANAye>yDzrsHTn09_Z*ka<}x81E(PZ(f0Kkw>-c^qbtgh{h!b#(X`aTBUg^;cg@GVMBG0eNc5h7T`0#0^qirTAh4F&l0w!6u?
zcvgK~s%1c`D9~ZorhE_=0Qm})75wbM^tPWwDB|00{Bb9-IBPEd
zx7KtbjHHcqBv?Bj$S0wh?=DcagWiPWG7=R
zfE*(V0_38AZL-unkE^~w-DLzpu!%Iml`+mPv{gKh+XWSv&CQmcGToFCqz-`_J5sJ`
zfj(fAW9@$*097JE>N!clGbSG`Lu^cSK!6E2?M-*Z0*
zXhnQT9Dr*QK+jKJq?#I1v)oDHByiS+C_ajiq%7PtCoZxjhueR~1cLIoJ7JaKKa?8b
zEnt1R@naAkQrA92DKGa$+J0ERKf4^<${(
za~574VV5if>5;>;2Em3-e|2XXCC(*b$%)YLK?&xE1%nOTS0Ha|_IL|Gul)do&xPbo
zOF`G4gZ}eEO49+0F^-ifmMW$PE&0c+xLzY7d+>P$RE-IsOIUt_Z2%ZomE)@)WkOeP
zIppHT9#H`<^1`x$5L+Jr79cd+AacZg3RFYE8
z{A9j5Y>_aB%bi-wo2{WAmNx!=z$MYJ<_UL<#J=LgWMhg<+$&;aU(Bi)n_ubK6c9SR
zeO30HbS*YGcvLs%ffLbmYD?+6rv1Gh^A{0@W^x5Z;0`_5xGRn^p-Z%$7l6Il;V6fz
zQ_x{Ox^XbLOS#+QB5X3Y?9