From e55ca8a1d1eb94bf3bc40cd6c598ce5e17f60e3f Mon Sep 17 00:00:00 2001
From: lxr-tech <1838593642@qq.com>
Date: Tue, 31 May 2022 23:25:01 +0800
Subject: [PATCH] update example-12 lxr 220531
---
tutorials/fastnlp_tutorial_2.ipynb | 2 +-
tutorials/fastnlp_tutorial_3.ipynb | 16 +-
tutorials/fastnlp_tutorial_4.ipynb | 10 +-
tutorials/fastnlp_tutorial_5.ipynb | 59 ++
tutorials/fastnlp_tutorial_6.ipynb | 59 ++
tutorials/fastnlp_tutorial_7.ipynb | 59 ++
tutorials/fastnlp_tutorial_8.ipynb | 59 ++
tutorials/fastnlp_tutorial_e1.ipynb | 14 +-
tutorials/fastnlp_tutorial_e2.ipynb | 734 ++++++++++++++++++++++---
tutorials/figures/E2-fig-pet-model.png | Bin 0 -> 57162 bytes
10 files changed, 916 insertions(+), 96 deletions(-)
create mode 100644 tutorials/fastnlp_tutorial_5.ipynb
create mode 100644 tutorials/fastnlp_tutorial_6.ipynb
create mode 100644 tutorials/fastnlp_tutorial_7.ipynb
create mode 100644 tutorials/fastnlp_tutorial_8.ipynb
create mode 100644 tutorials/figures/E2-fig-pet-model.png
diff --git a/tutorials/fastnlp_tutorial_2.ipynb b/tutorials/fastnlp_tutorial_2.ipynb
index 3aa27c86..4ee9579f 100644
--- a/tutorials/fastnlp_tutorial_2.ipynb
+++ b/tutorials/fastnlp_tutorial_2.ipynb
@@ -867,7 +867,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
- "version": "3.7.4"
+ "version": "3.7.13"
},
"pycharm": {
"stem_cell": {
diff --git a/tutorials/fastnlp_tutorial_3.ipynb b/tutorials/fastnlp_tutorial_3.ipynb
index 353e4645..172e1232 100644
--- a/tutorials/fastnlp_tutorial_3.ipynb
+++ b/tutorials/fastnlp_tutorial_3.ipynb
@@ -29,13 +29,7 @@
"\n",
"### 1.1 dataloader 的职责描述\n",
"\n",
- "在`fastNLP 0.8`中，在数据加载模块`DataLoader`之前，还存在其他的一些模块，负责例如对文本数据\n",
- "\n",
- " 进行补零对齐，即 **核对器`collator`模块**，进行分词标注，即 **分词器`tokenizer`模块**\n",
- "\n",
- " 本节将对`fastNLP`中的核对器`collator`等展开介绍，分词器`tokenizer`将在下一节中详细介绍\n",
- "\n",
- "在`fastNLP 0.8`中，**核对器`collator`模块负责文本序列的补零对齐**，通过"
+ "在`fastNLP 0.8`中，在数据加载模块`DataLoader`之前"
]
},
{
@@ -45,13 +39,7 @@
"source": [
"### 1.2 dataloader 的基本使用\n",
"\n",
- "在`fastNLP 0.8`中，在数据加载模块`DataLoader`之前，还存在其他的一些模块，负责例如对文本数据\n",
- "\n",
- " 进行补零对齐，即 **核对器`collator`模块**，进行分词标注，即 **分词器`tokenizer`模块**\n",
- "\n",
- " 本节将对`fastNLP`中的核对器`collator`等展开介绍，分词器`tokenizer`将在下一节中详细介绍\n",
- "\n",
- "在`fastNLP 0.8`中，**核对器`collator`模块负责文本序列的补零对齐**，通过"
+ "在`fastNLP 0.8`中，在数据加载模块`DataLoader`之前，"
]
},
{
diff --git a/tutorials/fastnlp_tutorial_4.ipynb b/tutorials/fastnlp_tutorial_4.ipynb
index 532118b0..3e148bf3 100644
--- a/tutorials/fastnlp_tutorial_4.ipynb
+++ b/tutorials/fastnlp_tutorial_4.ipynb
@@ -5,21 +5,21 @@
"id": "fdd7ff16",
"metadata": {},
"source": [
- "# T4. model 的搭建与 driver 的概念\n",
+ "# T4. trainer 和 evaluator 的深入介绍（一）\n",
"\n",
- " 1 fastNLP 中预训练模型的使用\n",
+ " 1 fastNLP 结合 pytorch 搭建模型\n",
" \n",
" 1.1 \n",
"\n",
" 1.2 \n",
"\n",
- " 2 fastNLP 中使用 Pytorch 搭建模型\n",
+ " 2 fastNLP 中的 driver 与 device\n",
"\n",
" 2.1 \n",
"\n",
" 2.2 \n",
"\n",
- " 3 fastNLP 中的 driver\n",
+ " 3 fastNLP 中 trainer 的补充介绍\n",
"\n",
" 3.1 \n",
"\n",
@@ -51,7 +51,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
- "version": "3.7.4"
+ "version": "3.7.13"
}
},
"nbformat": 4,
diff --git a/tutorials/fastnlp_tutorial_5.ipynb b/tutorials/fastnlp_tutorial_5.ipynb
new file mode 100644
index 00000000..1e41a36e
--- /dev/null
+++ b/tutorials/fastnlp_tutorial_5.ipynb
@@ -0,0 +1,59 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "fdd7ff16",
+ "metadata": {},
+ "source": [
+ "# T5. fastNLP 与 paddle 或 jittor 的结合\n",
+ "\n",
+ " 1 fastNLP 结合 paddle 训练模型\n",
+ " \n",
+ " 1.1 \n",
+ "\n",
+ " 1.2 \n",
+ "\n",
+ " 2 fastNLP 结合 jittor 训练模型\n",
+ "\n",
+ " 2.1 \n",
+ "\n",
+ " 2.2 \n",
+ "\n",
+ " 3 fastNLP 实现 paddle 与 pytorch 互转\n",
+ "\n",
+ " 3.1 \n",
+ "\n",
+ " 3.2 "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "08752c5a",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.13"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/tutorials/fastnlp_tutorial_6.ipynb b/tutorials/fastnlp_tutorial_6.ipynb
new file mode 100644
index 00000000..bd4b37ed
--- /dev/null
+++ b/tutorials/fastnlp_tutorial_6.ipynb
@@ -0,0 +1,59 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "fdd7ff16",
+ "metadata": {},
+ "source": [
+ "# T6. trainer 和 evaluator 的深入介绍（二）\n",
+ "\n",
+ " 1 fastNLP 中预定义模型 models\n",
+ " \n",
+ " 1.1 \n",
+ "\n",
+ " 1.2 \n",
+ "\n",
+ " 2 fastNLP 中预定义模型 modules\n",
+ " \n",
+ " 2.1 \n",
+ "\n",
+ " 2.2 \n",
+ "\n",
+ " 3 fastNLP 中的更多 metric 类型\n",
+ "\n",
+ " 3.1 \n",
+ "\n",
+ " 3.2 "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "08752c5a",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.13"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/tutorials/fastnlp_tutorial_7.ipynb b/tutorials/fastnlp_tutorial_7.ipynb
new file mode 100644
index 00000000..0a7d6922
--- /dev/null
+++ b/tutorials/fastnlp_tutorial_7.ipynb
@@ -0,0 +1,59 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "fdd7ff16",
+ "metadata": {},
+ "source": [
+ "# T7. callback 自定义训练过程\n",
+ "\n",
+ " 1 \n",
+ " \n",
+ " 1.1 \n",
+ "\n",
+ " 1.2 \n",
+ "\n",
+ " 2 \n",
+ "\n",
+ " 2.1 \n",
+ "\n",
+ " 2.2 \n",
+ "\n",
+ " 3 \n",
+ "\n",
+ " 3.1 \n",
+ "\n",
+ " 3.2 "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "08752c5a",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.13"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/tutorials/fastnlp_tutorial_8.ipynb b/tutorials/fastnlp_tutorial_8.ipynb
new file mode 100644
index 00000000..0664bc41
--- /dev/null
+++ b/tutorials/fastnlp_tutorial_8.ipynb
@@ -0,0 +1,59 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "fdd7ff16",
+ "metadata": {},
+ "source": [
+ "# T8. fastNLP 中的文件读取模块\n",
+ "\n",
+ " 1 fastNLP 中的 EmbedLoader 模块\n",
+ " \n",
+ " 1.1 \n",
+ "\n",
+ " 1.2 \n",
+ "\n",
+ " 2 fastNLP 中的 Loader 模块\n",
+ "\n",
+ " 2.1 \n",
+ "\n",
+ " 2.2 \n",
+ "\n",
+ " 3 fastNLP 中的 Pipe 模块\n",
+ "\n",
+ " 3.1 \n",
+ "\n",
+ " 3.2 "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "08752c5a",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.13"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/tutorials/fastnlp_tutorial_e1.ipynb b/tutorials/fastnlp_tutorial_e1.ipynb
index 6ec04cb4..8897c800 100644
--- a/tutorials/fastnlp_tutorial_e1.ipynb
+++ b/tutorials/fastnlp_tutorial_e1.ipynb
@@ -6,7 +6,7 @@
"source": [
" 从这篇开始，我们将开启**`fastNLP v0.8 tutorial`的`example`系列**，在接下来的\n",
"\n",
- " 每篇`tutorial`里，我们将会介绍`fastNLP v0.8`在一些自然语言处理任务上的应用"
+ " 每篇`tutorial`里，我们将会介绍`fastNLP v0.8`在自然语言处理任务上的应用实例"
]
},
{
@@ -82,9 +82,9 @@
"\n",
" 包含9个数据集，各语料的语言均为英语，涉及多个自然语言理解`NLU`任务，包括\n",
"\n",
- " **`CoLA`**，文本分类任务，预测单句语法正误分类；**`SST2`**，文本分类任务，预测单句情感二分类\n",
+ " **`CoLA`**，文本分类任务，预测单句语法正误分类；**`SST-2`**，文本分类任务，预测单句情感二分类\n",
"\n",
- " **`MRPC`**，句对分类任务，预测句对语义一致性；**`STSB`**，相似度打分任务，预测句对语义相似度回归\n",
+ " **`MRPC`**，句对分类任务，预测句对语义一致性；**`STS-B`**，相似度打分任务，预测句对语义相似度回归\n",
"\n",
" **`QQP`**，句对分类任务，预测问题对语义一致性；**`MNLI`**，文本推理任务，预测句对蕴含/矛盾/中立预测\n",
"\n",
@@ -216,15 +216,15 @@
"\n",
" 即使用较小的、不区分大小写的数据集，**对`bert-base`进行知识蒸馏后的版本**，结构上\n",
"\n",
- " 模型包含1个编码层、6个自注意力层，详解见本篇末尾，更多细节请参考[DistilBert论文](https://arxiv.org/pdf/1910.01108.pdf)\n",
+ " 包含**1个编码层**、**6个自注意力层**，**参数量`66M`**，详解见本篇末尾，更多请参考[DistilBert论文](https://arxiv.org/pdf/1910.01108.pdf)\n",
"\n",
- "首先，通过从`transformers`库中导入`AutoTokenizer`模块，使用`from_pretrained`函数初始化\n",
+ "首先，通过从`transformers`库中导入**`AutoTokenizer`模块**，**使用`from_pretrained`函数初始化**\n",
"\n",
" 此处的`use_fast`表示是否使用`tokenizer`的快速版本；尝试序列化示例数据，检查加载结果\n",
"\n",
- " 需要注意的是，处理后返回的两个键值，`'input_ids'`表示原始文本对应的词素编号序列\n",
+ " 需要注意的是，处理后返回的两个键值，**`'input_ids'`**表示原始文本对应的词素编号序列\n",
"\n",
- " `'attention_mask'`表示自注意力运算时的掩模（标上`0`的部分对应`padding`的内容"
+ " **`'attention_mask'`**表示自注意力运算时的掩模（标上`0`的部分对应`padding`的内容"
]
},
{
diff --git a/tutorials/fastnlp_tutorial_e2.ipynb b/tutorials/fastnlp_tutorial_e2.ipynb
index 93143090..c8f4b7dc 100644
--- a/tutorials/fastnlp_tutorial_e2.ipynb
+++ b/tutorials/fastnlp_tutorial_e2.ipynb
@@ -25,31 +25,53 @@
"\n",
" 将首先简单介绍提示学习模型的研究，以及与`fastNLP v0.8`结合的优势\n",
"\n",
- "**`prompt`**，**提示词、提词器**，最早出自**`PET`**，\n",
+ "**`prompt`**，**提示词**，最早出自论文[Exploiting Cloze Questions for Few Shot TC and NLI](https://arxiv.org/pdf/2001.07676.pdf)中的**`PET`模型**\n",
"\n",
- " \n",
+ " 全称 **`Pattern-Exploiting Training`**，虽然文中并没有提到**`prompt`的说法，但仍视为其开山之作\n",
"\n",
- "**`prompt-based tuning`**，**基于提示的微调**，描述\n",
+ " 其大致思路包括，对于文本分类任务，假定输入文本为，后来被称`prompt`，后来被称`verbalizer`，\n",
"\n",
- " **`prompt-based model`**，**基于提示的模型**\n",
+ " 其主要贡献在于，\n",
"\n",
- "**`prompt-based model`**，**基于提示的模型**，举例\n",
+ "![](\"./figures/E2-fig-pet-model.png\")
\n",
"\n",
- " 案例一：**`P-Tuning v1`**\n",
+ "**`prompt-based tuning`**，**基于提示的微调**，\n",
"\n",
- " 案例二：**`PromptTuning`**\n",
+ " xxxx，更多参考[prompt综述](https://arxiv.org/pdf/2107.13586.pdf)\n",
"\n",
- " 案例三：**`PrefixTuning`**\n",
+ " 以下列举些经典的`prompt-based tuning`案例，简单地介绍下`prompt-based tuning`的脉络\n",
"\n",
- " 案例四：**`SoftPrompt`**\n",
+ " 案例一：**`P-Tuning v1`**，详细内容参考[P-Tuning-v1论文](https://arxiv.org/pdf/2103.10385.pdf)\n",
"\n",
- "使用`fastNLP v0.8`实现`prompt-based model`的优势\n",
+ " 其主要贡献在于，\n",
"\n",
- " \n",
+ " 其方法大致包括，\n",
"\n",
- " 本示例仍使用了`tutorial-E1`的`SST2`数据集，将`bert-base-uncased`作为基础模型\n",
+ " 案例二：**`PromptTuning`**，详细内容参考[PromptTuning论文](https://arxiv.org/pdf/2104.08691.pdf)\n",
"\n",
- " 在后续实现中，意图通过将连续的`prompt`与`model`拼接，解决`SST2`二分类任务"
+ " 其主要贡献在于，\n",
+ "\n",
+ " 其方法大致包括，\n",
+ "\n",
+ " 案例三：**`PrefixTuning`**，详细内容参考[PrefixTuning论文](https://arxiv.org/pdf/2101.00190.pdf)\n",
+ "\n",
+ " 其主要贡献在于，\n",
+ "\n",
+ " 其方法大致包括，\n",
+ "\n",
+ "通过上述介绍可以发现`prompt-based tuning`只是模型微调方式，独立于预训练模型基础`backbone`\n",
+ "\n",
+ " 目前，加载预训练模型的主流方法是使用`transformers`模块，而实现微调的框架则\n",
+ "\n",
+ " 可以是`pytorch`、`paddle`、`jittor`等，而不同框架间又存在不兼容的问题\n",
+ "\n",
+ " 因此，**使用`fastNLP v0.8`实现`prompt-based tuning`**，可以**很好地解决`paddle`等框架**\n",
+ "\n",
+ " **和`transformers`模块之间的桥接**（`transformers`模块基于`pytorch`实现）\n",
+ "\n",
+ "本示例仍使用了`tutorial-E1`的`SST2`数据集、`distilbert-base-uncased`模型（便于比较\n",
+ "\n",
+ " 使用`pytorch`框架，通过将连续的`prompt`与`model`拼接，解决`SST2`二分类任务"
]
},
{
@@ -98,7 +120,7 @@
"print(transformers.__version__)\n",
"\n",
"task = 'sst2'\n",
- "model_checkpoint = 'bert-base-uncased'"
+ "model_checkpoint = 'distilbert-base-uncased' # 'bert-base-uncased'"
]
},
{
@@ -111,20 +133,32 @@
"\n",
" 以下首先简述`P-Tuning v2`的论文原理，并由此引出`fastNLP v0.8`的代码实践\n",
"\n",
- "`P-Tuning v2`出自论文 [Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Tasks](https://arxiv.org/pdf/2110.07602.pdf)\n",
+ "**`P-Tuning v2`**出自论文[Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Tasks](https://arxiv.org/pdf/2110.07602.pdf)\n",
+ "\n",
+ " 其主要贡献在于，**在`PrefixTuning`等深度提示学习基础上**，**提升了其在分类标注等`NLU`任务的表现**\n",
+ "\n",
+ " 并使之在中等规模模型，主要是**参数量在`100M-1B`区间的模型上**，**获得与全参数微调相同的效果**\n",
+ "\n",
+ " 其结构如图所示，通过**在输入序列的分类符`[CLS]`之前**，**加入前缀序列**（**序号对应嵌入是待训练的连续值向量**\n",
+ "\n",
+ " **刺激模型在新任务下**，从`[CLS]`对应位置，**输出符合微调任务的输出**，从而达到适应微调任务的目的\n",
+ "\n",
+ "![](\"./figures/E2-fig-p-tuning-v2-model.png\")
\n",
"\n",
- " 其主要贡献在于，在`PrefixTuning`等深度提示学习基础上，提升了其在分类标注等`NLU`任务的表现\n",
+ "本示例使用`bert-base-uncased`模型，作为`P-Tuning v2`的基础`backbone`，设置`requires_grad=False`\n",
"\n",
- " 并使之在中等规模模型，主要是参数量在`100M-1B`区间的模型上，获得与全参数微调相同的效果\n",
+ " 固定其参数不参与训练，**设置`pre_seq_len`长的`prefix_tokens`作为输入的提示前缀序列**\n",
"\n",
- " 其结构如图所示，\n",
+ " **使用基于`nn.Embedding`的`prefix_encoder`为提示前缀嵌入**，通过`get_prompt`函数获取，再将之\n",
"\n",
- "![](\"./figures/E2-fig-p-tuning-v2.png\")
"
+ " 拼接至批量内每笔数据前得到`inputs_embeds`，同时更新自注意力掩模`attention_mask`\n",
+ "\n",
+ " 将`inputs_embeds`、`attention_mask`和`labels`输入`backbone`，**得到输出包括`loss`和`logits`**"
]
},
{
"cell_type": "code",
- "execution_count": 3,
+ "execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
@@ -178,24 +212,24 @@
"source": [
"接着，通过确定分类数量初始化模型实例，同时调用`torch.optim.AdamW`模块初始化优化器\n",
"\n",
- " 根据`P-Tuning v2`论文：*Generally, simple classification tasks prefer shorter prompts (less than 20)*\n",
+ " 根据`P-Tuning v2`论文：*`Generally, simple classification tasks prefer shorter prompts (less than 20)`*\n",
"\n",
" 此处`pre_seq_len`参数设定为`20`，学习率相应做出调整，其他内容和`tutorial-E1`中的内容一致"
]
},
{
"cell_type": "code",
- "execution_count": 4,
+ "execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
- "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.seq_relationship.weight', 'cls.predictions.transform.dense.weight', 'cls.predictions.bias', 'cls.seq_relationship.bias']\n",
- "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
- "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
- "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased and are newly initialized: ['classifier.bias', 'classifier.weight']\n",
+ "Some weights of the model checkpoint at distilbert-base-uncased were not used when initializing DistilBertForSequenceClassification: ['vocab_layer_norm.bias', 'vocab_layer_norm.weight', 'vocab_projector.weight', 'vocab_transform.bias', 'vocab_transform.weight', 'vocab_projector.bias']\n",
+ "- This IS expected if you are initializing DistilBertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
+ "- This IS NOT expected if you are initializing DistilBertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
+ "Some weights of DistilBertForSequenceClassification were not initialized from the model checkpoint at distilbert-base-uncased and are newly initialized: ['pre_classifier.weight', 'classifier.weight', 'pre_classifier.bias', 'classifier.bias']\n",
"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
]
}
@@ -225,7 +259,7 @@
},
{
"cell_type": "code",
- "execution_count": 5,
+ "execution_count": 4,
"metadata": {
"scrolled": false
},
@@ -240,7 +274,7 @@
{
"data": {
"application/vnd.jupyter.widget-view+json": {
- "model_id": "b72eeebd34354a88a99b2e07ec9a86df",
+ "model_id": "21cbd92c3397497d84dc10f017ec96f4",
"version_major": 2,
"version_minor": 0
},
@@ -262,30 +296,17 @@
},
{
"cell_type": "code",
- "execution_count": 6,
+ "execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
- "Loading cached processed dataset at /remote-home/xrliu/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-18ec0e709f05e61e.arrow\n",
- "Loading cached processed dataset at /remote-home/xrliu/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-e2f02ee7442ad73e.arrow\n"
+ "Loading cached processed dataset at /remote-home/xrliu/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-294e481a713c5754.arrow\n",
+ "Loading cached processed dataset at /remote-home/xrliu/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-ed9d9258aaf0fb54.arrow\n",
+ "Loading cached processed dataset at /remote-home/xrliu/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-f44c5576b89f9e6b.arrow\n"
]
- },
- {
- "data": {
- "application/vnd.jupyter.widget-view+json": {
- "model_id": "d15505d825b34f649b719f1ff0d56114",
- "version_major": 2,
- "version_minor": 0
- },
- "text/plain": [
- " 0%| | 0/2 [00:00[22:53:00] INFO Running evaluator sanity check for 2 batches. trainer.py:592\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[2;36m[22:53:00]\u001b[0m\u001b[2;36m \u001b[0m\u001b[34mINFO \u001b[0m Running evaluator sanity check for \u001b[1;36m2\u001b[0m batches. \u001b]8;id=406635;file://../fastNLP/core/controllers/trainer.py\u001b\\\u001b[2mtrainer.py\u001b[0m\u001b]8;;\u001b\\\u001b[2m:\u001b[0m\u001b]8;id=951504;file://../fastNLP/core/controllers/trainer.py#592\u001b\\\u001b[2m592\u001b[0m\u001b]8;;\u001b\\\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "application/vnd.jupyter.widget-view+json": {
+ "model_id": "",
+ "version_major": 2,
+ "version_minor": 0
+ },
+ "text/plain": [
+ "Output()"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n"
+ ],
+ "text/plain": []
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "application/vnd.jupyter.widget-view+json": {
+ "model_id": "",
+ "version_major": 2,
+ "version_minor": 0
+ },
+ "text/plain": [
+ "Output()"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:1, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m1\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.540625,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 173.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.540625\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m173.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:2, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m2\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.5,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 160.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.5\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m160.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:3, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m3\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.509375,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 163.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.509375\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m163.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:4, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m4\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.634375,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 203.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.634375\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m203.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:5, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m5\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.6125,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 196.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.6125\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m196.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:6, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m6\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.675,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 216.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.675\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m216.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:7, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m7\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.64375,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 206.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.64375\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m206.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:8, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m8\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.665625,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 213.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.665625\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m213.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "----------------------------- Eval. results on Epoch:9, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "----------------------------- Eval. results on Epoch:\u001b[1;36m9\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.659375,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 211.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.659375\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m211.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "---------------------------- Eval. results on Epoch:10, Batch:0 -----------------------------\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "---------------------------- Eval. results on Epoch:\u001b[1;36m10\u001b[0m, Batch:\u001b[1;36m0\u001b[0m -----------------------------\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "{\n",
+ " \"acc#acc\": 0.696875,\n",
+ " \"total#acc\": 320.0,\n",
+ " \"correct#acc\": 223.0\n",
+ "}\n",
+ "\n"
+ ],
+ "text/plain": [
+ "\u001b[1m{\u001b[0m\n",
+ " \u001b[1;34m\"acc#acc\"\u001b[0m: \u001b[1;36m0.696875\u001b[0m,\n",
+ " \u001b[1;34m\"total#acc\"\u001b[0m: \u001b[1;36m320.0\u001b[0m,\n",
+ " \u001b[1;34m\"correct#acc\"\u001b[0m: \u001b[1;36m223.0\u001b[0m\n",
+ "\u001b[1m}\u001b[0m\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n"
+ ],
+ "text/plain": []
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n"
+ ],
+ "text/plain": [
+ "\n"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
"source": [
"trainer.run(num_eval_batch_per_dl=10)"
]
@@ -421,14 +976,55 @@
"cell_type": "markdown",
"metadata": {},
"source": [
- " "
+ "可以发现，其效果远远逊色于`fine-tuning`，这是因为`P-Tuning v2`虽然能够适应参数量\n",
+ "\n",
+ " 在`100M-1B`区间的模型，但是，**`distilbert-base`的参数量仅为`66M`**，无法触及其下限\n",
+ "\n",
+ "另一方面，**`fastNLP v0.8`不支持`jupyter`多卡**，所以无法在笔者的电脑/服务器上，完成\n",
+ "\n",
+ " 合适规模模型的学习，例如`110M`的`bert-base`模型，以及`340M`的`bert-large`模型"
]
},
{
"cell_type": "code",
- "execution_count": null,
+ "execution_count": 10,
"metadata": {},
- "outputs": [],
+ "outputs": [
+ {
+ "data": {
+ "application/vnd.jupyter.widget-view+json": {
+ "model_id": "",
+ "version_major": 2,
+ "version_minor": 0
+ },
+ "text/plain": [
+ "Output()"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n"
+ ],
+ "text/plain": []
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/plain": [
+ "{'acc#acc': 0.737385, 'total#acc': 872.0, 'correct#acc': 643.0}"
+ ]
+ },
+ "execution_count": 10,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
"source": [
"trainer.evaluator.run()"
]
diff --git a/tutorials/figures/E2-fig-pet-model.png b/tutorials/figures/E2-fig-pet-model.png
new file mode 100644
index 0000000000000000000000000000000000000000..c3c377c08a0ba3199cca4fb3610aa0ad08ce41fb
GIT binary patch
literal 57162
zcmbrl1yEc~_cs_LcyKa!aF+oB1bwgpf&~r1VQ_bsBsdK25(uurHCV8j;1+`G1a}>r
zy?NiieOq6BTf1Au6m`4$p6=7fPycSjI~6%xEOM+zj~?MF$jhiddh`SW{4g<|10(-T
zo=gG%9=oc`Nj<6@q1*-@o>@vNOFnv36N`Oof(ATeI>~=*~YUoLyU6r#$(nX@Z%Q)j|$DpH=eI^^v)h=C<>GZPupTohL@H`)*UVu(;Y3yKfc2F4E(Lo
zLS^uuN#Z}F{RbZYuOX?AFaqFN5Ok#XsI@^n*rd#y+JtGr`XGKHi9Nn)5l0it
z`(|OO0bHLeK7W1f2H3butrktpI>gl1!4IiH9u(ZMN!Xwl#;03go6lA{hI|z-;sDDF1$7
zc6p;Ay#;UDEIuD?DBI5Kgy(iFkUbHFgcKcJj9hP}0g&1>xIeB}$w&DQ>3Vnup0u`<
z%4etZVA`IOh2D-t2I~k;K4~lJL|(|ABE!WxUHn;LM(E^>kEH2F&zNkFbtvBY2lB8P
z;sB}Bv?Op|F+)ObX_32vx=W&LbXabANx?=okJN+NpU6NOg$cqzB6|KJ3yZUiJkcgO7oo~Y7Q$(%EJ51)_X+;`d2%L72
zBL4{k`eNU{?MtVNUkX`5&RCZA+&%Y2^!j2!ar&<)nVexiLCLS|mj7|1u70^XHDs*0
z2wGn~7&lJ?U)#DlEK_rLo|H*dwaq=Q^4$B3Cr?keTdbhKQCBL0<
z%x~x@t>hUrN+@x}$0OJ3+Vm()+Z&UWN_(eae343z
z2)sg)9>@L7E&hA{wX63i+e&
zMnUWJTPKhqY824OK7PDZZZf2`g1#GdYQhtU~%eIo{wLp90$@q-MnC@G}MjHV^;sPkp^C~k>S
zj8>0MH<{NorD)B_tT8T}L(g~mx325a$aO?RVqY3c>!i%=;u>5X%?n~x)!Rg6vmesR
z0^BPnpjGxDa9bc01^A>Bvu#%?jHp%Rm85iIoK?aq^_qH>~2{q!AwC&|O>1u_NFfA9tk(L*7*9+_M>Gd8xT!D92pGSzf-t
zUJ7NL#@Fd+c_hfkmVz*RKy&=y5*^&fp5`NamzlRS(6>0;2~AAB79(WfWcNY$dA-*0
zbYq{VOZ#q>-<~{w?5J)A8u+Y=T}eNa$?Evy;BKYQ&fI&V#|
zP%cscc^gN)7CBh_&7{GW*<)`;t>{aub1yiOourt56$jq=3qM-iAU2$h#uHOYJ1{uZ
z0{8eJM_gapk^I*%XdT&AraBfU#?nX~o5))OnTbQXtgJyeVk
z%qQA@;d1jAy@%6;vgF%ldpC`5P{B8Z7EAnjHMP1W<=_*&`TQYJLwazX?rR9IzQGX}
zwJ!>rDI
z2%f^Q7{4rv8-^iHqdwIaQ@10EiB1-FB0Bs$()oDWKRT}^DTDf0U}`7#_g1qWgk00S
zC8gd_HyUi=RI+@PsDmhq{CXL$R{cH&Jxec$<$KP}qzwEgh;onSEy9mHAH);P))9;g
z%R!&GoeoWA$z1p8@j;}~;eY17S5L0)-an$X^Wx^j8cDTx>K?)FntE+!4R$)l;2upL
zP?Gb~Ax0UlVMUVoAT-5PDl7Oemz{si5-~gP4fsDRU97gCUt6V`$!>@3iet*FN6JBD
z*xNn%y9r%GYDIfx2M6;A(%Ax)(-TaetU7Ir+Bw|Ni+(fxkY00?DOC0v+OI{h(yHQFqm1wshF1L1Sc2VZf-eHF#Pa
z^MSjVpOC_b|MQ1B@tH)9UKtiQqmP0uojD%{SB10|9Y;TcpFoqMHRyAG(EERTr?>IB
zC|WJ6$8l?+`7t^FPO$fjbUR#C*{5;QH`<9~
zcXc$?-PfWq`yW0BQ0e8L#|4E$*R4_2))P%zXGw3!)5Kz8j9y4ReCny*(odMUNn$@(
z)v_}bqRw%*3W1C&e?H$I-a@Gp(Sl=ahjiV$f`dLu^(4Q@vAwDBl5%wOD
zPjz^!#dPa;qexV&BX_G5z|_Alt=}1)(TGqbLIPq^X=-|r51Cl(GzYGukzMEeMB}9`
zWT=h}uj|vk>W)kX{>bekp>|@UhMHcYoDoDsm=rQ~3#~k-6CwB%0X<0?-#Q+kul8Fr
zP!e4k6|}+p)Kqg*WxchyO~&tifvzKBFS)!)ev$*7RDB3(T4U7?6}zJe>BF8h(FJWV8tp3EJ#`W&McS89lm1$9Cpxco8FyhGh`YWt(DQb-gao&KR=$SJObF
z*gAY0_WaO&v(4qzp}GDFFCogsj~{ZHR$UXw&zFL%uDmC1t9_;L{Dd8-USx|!eThQP
zMe>cV`IftNi1yo~Go5*(t7fFL+Fsd4%-?^SgEZcinJMnjRpNL9-ghSgScUAKAM)tx!ErONz3
z2^X`E3H_4#qprw#bFT+9HiWaG3WI;Zv@lUz29fxhC~U&L>?`?f6!l7C!L?r29F4P!
z-6S`e&0hBmjT~fAo;kHLbsl11WSnDRafGPo%&FL1U>i=%Nv65BP~(L^r3G%r>Yc=(
zPpTQg&GA0+8~DWRL(q>TCs%6W!1oU76MRpP-V6SQo;iGzw#KA2te%&^q|#Z^sC#*|
z8eS;0g8yv!X1c~(%97u*4xM^CRN}yC;}tsrb(6!|)iTk3v=#~9>U+OET^{|EL4%#Jy
z4O`wTSsZ1&!oq7OJcWudIR{v<*IG!a6@o0w1)3BaC-!t~C%s0fj0Z2h`GU{V%7cO3
zFm^|9AEDY#)MR`Sm%`49|uE*RdY^YrQD
zs=AJBeQ5iXfq52|J%k8R$^#%6Py`bJIbDQ{Y53Fp*`_}@Ytau7c
zrZ-K)EzEM*ZkMwy73`w3f45P2}_r26WF;0rFL)ZO8v5bu-2Qf1DIM3!f|V+gtKU7>X*6Z=dW6e#{&g}
zKMCl_i}^8quSL5zU*35Qx}8LUG6W#C9~*A>h6dm)+dCX>L+NpV8bxKEE(Mz>;vKd6
zUYUhxj=JkK*R{24PJ?9kwqxZebnMn64qj8s_YzF%r
zmil8=GlfkgbbqO_Kk>EdFJfx+LZ4KTWefOqtC1k9QNKI!G{?!}ZSH}=c#8&e+uah5
zPMsg|5|0z4`-IGo$DG+gCSt*FN&pfzen?Qc>f^v`mqgB0v?VQlWk@=RN?DB=n_)0W
z|MZfLEFWtC2mZUmuK#nWmBeD@b#az%jAC@OWZJC{ujz`DvA~j}iT2GN=`4woo#m&Z
z{rj+kfgJN#)uSbmFXJ_nH)$@ST0(}QNa%cbo208>4Xfs*ui0An847&)&u1qk(h=nV
zYEtDKgh&osPAp&hp=Z~;qbIy|QYHEITE8543w%)LcM(Qj1zznHYI{INvw5SH$!ow_
z{@l}mxY9Co8*E-&$i2@lEfRrh5+31+TU}y2rz+h}=cAB&E*8|G_mqH&%An2^Njq58
z<0Tn`3l|+eiGr*YNOFjlikw-(got_?q*ZG@wzeJ8>W4%4w$~AZh1uYNplJyj=Y4)l
zD~o0AYbqwFUJlbjkS<;Yj<$_n`JKV$8k{@J+I9@1y>VbZ%B5dQV1-EFX~TA^&L~BW
z%pOhmoT#EGP|w(TtaOAW8|JH17~w^b=kL*j&XkPoaEY(JX{y_`;gmhN)jC`J(%^Bt
z#rz#I^$S-G=dkcWIv?R$@w9rsQccBW;lGU
zp{1GaY~q(4thu~%w%0H-!u2G#MwmL4+#TIFFej!)<0Z5woeH!Br?kid3#$cf
z$A#F#EpdmB_q^ucsyeN`{eH)8xVba4b=SE%LjEh5547lxr84L{+T1z&eU!o_unJR1
ztLNrKeH_GSkP};D`?*XYlI^e5*mj<2>neBu*(Ykw#GUkl8BWFuz6rXgV((>gT2DMe
zg4u=o*>a>k;U)YoVm|yfp99csoahTNXMUHnW=Jf)*q@uq+cc8dOyAT}+|bS&+qSIV
zJdJ8p^k`51o~H^;$|n09B+(Wj*L+$X!LAy}gO>S)ZR4P19-8D7z?+Sns|h{iQ1f|s
zW_MqP$zgbhnVRfd2uQQI$L7`UYlrY`3TV>WXETol=?w1>!<}Crp&---z2D97km|I8
zh%-B6_O@gUuv$}KefKS^ZJG1;mlvb(LDFTjEqxps}U!&#N
zN2qFhOPAeMLnfhT=Og6mf6(|HU_!D7*QCm`u~H_%C&$LJo|VT|uu;Lj61O^-+k6!3
z;`~>F<#+sj>^-O+jeA2te39@dK3q?&EpP(bfn2BMjON*kG#5$H)y&chWbLZI8R_?q
zF*Kss=P6{A0!bHHSn!{Yuf^9YTaI7qU6C=EA~fF>(cIsM`@S`cp8m!Wd`J-)yCIw`
zA!~tVa)rPh{|ZL^dTBT~nHdRZnX<*!re+_deJ59x2AEVm)*^XB?-6-%`5)xj&&Ep+
z$Fd_wWsN5ff!|#^CqunZRbfoYM!jR@;I(8NgaO~tab5I)lI-LKs1acVu6w^vnP1_g
zqq-d|$C>>PYI&a^EhX%TqBCD_o+~XtxXe^UKL|W0aS1-~;TU(m|58>56YqCtDA^O$1fygRDZ)wZ
zIIc6l9KCEwAADl(^n;76)(KDOdT>~USN~!d9(IdUU(D01`*R#PBsPje0lC(GkI1D|
zn`i>h`~|VC4{5H^z2%dYnkN$^Z
z1)cyqqcmF*NhC6?ioM-Y{IdrN0kzoha<~<*xUPJeqOD`ZLgIVfv?7i~9Gy2}5Fj(~
zmM7($>Y7~kos
zenA?Vn#7JRjKlDYLGH^UCg_
z2JB!$^L+))CAIj&KGqt8T_k8i5W@KQve=fWP3RQzZPCLW42i)Onc_(z4Ox7e!$tAd
zR*d}jNjkWeUgK>WzOFNsg!~RgK$E{ypr`P1t}bOm{=eq^jX35_(kO{eh5{
z2I+9iE^hymEI&s7so^oJjOG-1zNpvzBvmwpN3VV^MH=gItud@)Nvli$86{|7bgSA<
zf=Ym-sNbT;VfBg5app*wi$0e=spqPA&s*d}DKmHvldm$KEDDl8fEX;SH
z_J^zVi&e>xPD#Q_PKrncboTQh#_XQ0M*XM1+eqdD>d-n!$eaB{bx?048fMm7a%Yf6
z+x-U{%a;KI@wH2=B++VbmGDKSH1#~b*5V*QA;^pMh;_IX!a}Gpit2r<4IiYXSnjJ<
zB6oE|<8gUCcG^q_5z2q6`-!x*+>t-Kt4^CJHdBh{aQ=7jK=5?P-Ha{G$eqVvwA!TM
zJR1Yvpp@X6#t29&TZzEi3WHiu!@zQja$%ndt-+y+Izouq?UU7ZkCX8
zB%M&n$yitDCKGOVmf{r3j$O0oPphFy7pk-5qW@Y{ELEu2mYnnFQL1*hQI^U#L&fsR1o24->jM&kXG7J2w|DlL^-p
zt0h}@I`#)US9)jP>CvreSA5bO+=Me~YnK8?n2?AYbR){@y6?B{>9b#gkZqwiF^s1!
z5_ukp&f4JidVJ@u7c&CNI@fL1=rD2Oyt8%H?rqk6Z|SM8CKPTItsr(mN5l@}`E$-$1Ro}Yj0k_+&K_0+&8NzX-#Wsp?DdXa7S;a2%T(>gn}q7C
z=L3YxYcQ(Wy5kb*UD^YVdvnuY}e&rL{w0;Jh+$LWq6e=7Qsoy_%}(B?$@##pU&Y
z_o}O{?a86V6O=sVPf;Q~f#hNV5_p?8p3WYFqoCVGIj^P9d`jFgKT0hkH=OFGdNZ7}(!1|{7Dx?wC`{FKfBh^r-TApycguWw|*E>O5oUX=HlsvdvWjRXy1U2)TSz=x)>$^cdxB!LK9^ntFi=>^L>(IR$u
zo-FPJ|XBOE1v-sK^OEtL2EnNVYfmJLA?rQpOIT
zuxXi|BkLT;@yz!|
zXFJE@hTaQ=DTiD^z7}PX2(rm`dUQ{#n!K+0ZY^FK
zHE>~>uf|OA`;F=Tlv^=d%6odvCHk#i$3nKy=i3hiRZ%y@yZ;FzeBu;?-x;V8PG+Y
zA*B)AFh1+ls6O~4ktY)y0nuDw+^e}mS4743;CB%xm(qkfK9B@{na9d3!5E*`E+z*O
zYB~>805SJ)TdAq3#p5TjDv;DhN8^>luocSmZ=Y5llf~fqVr@h(_3*NC)#g3cJOICS
z?MZOpY*{WvoAVRJr|Fy_UEG)nFTnUi|L4}v?}*ql$g-0xO54kKh)c-3LZ*q-{%E90
z<9NQY-0Kw*HvZYc3UxAnxfdo;y)-~RpAy}}oD+kv5L`EWPsTOmooTdGAlZXB-Z&CF
zD%ofnN!qcEcK)K5+yG8C|NVptqPrSURMxp`wlF75u$b+O#qBSQCBbz|g1q*F#fDXN
z%dRtu3qj9z!$y&XF-FD<3EyV?jk!}|u5cC+2f>MB2SBOGoCW`BIs0@YD$U8TUB3^=
zU-Dn!%Q1UO&I>+YLrIT>4^<9wGoy25sri-BTB;8Z8IL!kW!v5G3I{CUN!)%XzuA8p
zul7Jw&}Bjl$N2D)=lE)qXH{x=L>{>Ox@Ko&!jEanS|yCrJT47krKfVP$V}rS3!vtz
z_B>|KnJ`qqT*X`YL#B5##Ic@R|4ZrFf;)Wp<~4_96}x%=is$rGPx>VIt@*f<=dsjB
zbyfZ70r}vjOrpwVOi44`XtQ*)BTW2saXrbFaIdY$H}ly?5n9W?ekOwsp1$n-qr0T2
zKgS2T)=&oWa%Y-rv3pX|UrPp>pB9d6btJwCE?LiGeVc0we&OgYgNXUObO1XzQb@ePC3!wIjox>IAhPdkm1~Hm`3SsB?8q4~DN?0GP>B(3$^n+dG9u~XhVkS!
zWEq$F;Rj2}@g$fG)(hN83azDb9U@V!Q=Vg)pqA3J{k-Rt|a297eQeYPWHdix2{lB`}W6@B``2qK5=Z|G3)&1
zSRRStgA1hvmwssH`)_?liW|_iIPPQTWpAN9JyfoXG$|SVzOg1e++{Mt5o`i1~dA6+aIMR%TQxy403ubCalPxy3
z(gmH(9?xCW)|+)4Q)REqfjFMBTz(a2L(3XFmq53}6DC6+Kv-92dN^O|F>FV;(cc
zKfT!RRYAN=@s8F{O!01lH^8jW-t3XtVfj652;H{sRmf{Wn~F_F0d~l*kDxMMY~GM^
zW*jmfdHxs4zsHbDzo-D11P-|hJz1ot^{|!-i}lT>+sO=
z`mlbo{8~eL8+`d|E&Fy^m$z>f@l&bTZDeqi)6vMfwTRR|`a~C$n(2bVJz$f$wS=}RNh^D&4Kz&o<}U!&%BIX{4uj;$)N7v+aYK74V3b#Kl`VxZk4=2
zq+-jVwIoyLBQq%v(X~3HTTyS83;?W&d!JQ2E=&;mSkEbmGkB1&?k_lJW^;6iZBVYj
zk8V-w{|Gz^*6r=3X=X>&gNrI8;&Vc~1l4-dx(nk#6N!y+3%N?BPpkqLlEIK-5cu+Y9u
z69#}#4X>aK{s0b2-E8cuq~Cmx56ngyxK0M~S>Ocjt@TBPV3_Ox=%`Fsn2?{2ewgWFibkK>Ip{QxD!G
z(i4hHME5#B25ee=0iy~$`J+H9mB#x&m(MFGcs44AqFg>$`0lP4{u}lkZgq39e|s$E
zdqGV1zY=8}`67a7dgIu39Ex8`rBPs?qm{=G8QR};X}P&ofh92hy#&FtT=tF8oY5vhA|@q#x-zww
zz@1$yXuAv;CR*8TY
z60G|dhbHxs@{LQdqwSJE%78Xc*!?xct~?$=6POkFt+==tpAi7Y{MSsk!=*2dwigGB
z>@u3gmGJQ6LGI3@sxPA6wE)Hco;QDaeSNY&1dSC7U0BekEfYQmL%O=U{JK%}v7(=c
zl;*5v|9hwPxbtK=4SaPL!-RA@YBbPZ;|KMdCneIl@l;`*=|CMw&e(-8s
zVqy>}r|!`{EGH)iUn};SJfF{U?A
zX%(1_iwlw|V1MZU?)&4UCDW
zm%{?&^*?87E-WryU91CCH>Sa5>=`@|Fc@52UHwl!BswljX?`<}@PFu9=)mPM^6J8T
zwL3(rz`?2;1inTcj^Pf?csPEZyR-)U
z=g3Hm#3*pku!s8_Ko{UI{!`rmxzN^{^gQpd-(GBO!Y>uc;Drp~fHitY+}8(4Xy1ZF
z{&N=&a?8EjUy0=V6$$8`QCOSn-MJLlq{!JuE9L>+-ZlmS%P>HrV+;EcltERl@pU
z>Z$#}N>@;b{M~px*wk%rCSGE5d$JgC3>E(tEdaF4_`FV3(?<8Zs9Ka>ITsh&JJL7O
zFObFvqxdfdG^)w0D#T3qCio)+U}_c?Ow`_g2-4EiX@!JzfJ6LmgJQit+wN&%-kq*`
zNtYo>0S^azpP1dZxo(b-&;^4{3E?I61v_V>`X64H<+8S&Ot^$ty=f_3U%w*rma5On
zO4`d9OdK4rv&)aF
z5&`n|o!R+qmu%ZO_ge2VkXWPoShfL7dNzu%4m{<)#2R;55Li-A3e(5M&s!I&
zYN=$w>-^B5_uJ%3OV&HdH4UG0bAu+8c)iAx9(e5{%8q0m)3630R(@(nysm(U1V&RW
zbF`C+pQk#0Wf6pMi{H29>F;d>I;^S5zTvY@e#5sBmnWLam!ACQja_G6+p9X8r*3v*
zhO#YbHFftMsd;U+HKvfah8ByS#rT|IWRl)jN$SgBxQlw`X;B!Vx
ztE=K6Rg|}|5k@bi={Ytq9ZdVT1kk-DAK%j9zJGn)uT!qJ5i4omVv~(vZa*+J9kass2DN-p8KM9Qk_Zh
zKyyWFF3V^**WpGAMSiMYX90_4k56*TUDKC5{h`YReIEriE&*cehj}#FmYH{JQ%0k#
z1AU<T!9xfBn5VWN*?L01|5+dUN4ton>7ANGXZ6kn+R)GbK}U`j2H*&L
z>Q7Ge0<{(cx4X}XZc!=c9z~%^nd-`t@bED~z+WQg^Fz52>>eyaB
zij)(G3iL9sqZ*^tjE!Ff-fs+N(zCNG0*3ziPc%X1V&93(Omy&2`~^URqSxZY8Vhg~
z;_0t#NtZ*1(!kQ;R@C+s^pMePu`AR}hkq-uGFm@He(PI>F^;N`F;OGGV>I1HyZ+sA
zIhu8frvk&;)X)j-RH0n&Ea(?w3ZECXev8~B|D*}Dw6u8b)gUefdZ(Lq3P_Z5v7v52
z`f)F|MLT{O*9JZj+5eW-MSMs?FN3M<0=Vgp$0aKJa}Cmxl2gyDrCh7i-A(JzF{R$_
zEw-jQ@_qL_
zz`++S9{U;H7_8;?^NlR5Mq`LN_P<9!MSbo0d-w8$9-PX{F?;0BH>H5HSeKxQO{(zGGpmN7H+i_$}oN^B~FM
z`vujkXFyWnTN8zfx4!MyJLUYp>>1?eKsBY;c2gEiHPM=w
z!`(`8wXE065BJPT00v)kv`9G@+dLjS<6knm_=;~059OMa)y9mCZ%rQiEv1jY)%rWHChp94ETJw?YNkQ(?K
zlT3kp)9YULPv|kz`l!SM
zG2NWpG(0>v)`xKJQ$iJHg9*hywIOpDiz|)W0xGehKiNm}@9QRjsQX>YNR0IwC}Xq1
zy4^B=lZ#&#<9V`2DAdb)S*y(X!ND+CzJ7PG)gq8)ZDDyI41bjj3J!M|3j>>O38zMm
zz4#m7o!#9FtxJW2fy`R=?2VqH5nNvU=AgW)gOD!xZw80|T*XA?F%T2Itxc!>0U>gY
zZC;CC^FqkT$b=25BF0`M2ta^%Cyy!rM=MS?GmS8q4cLm?CqLOOAc4muSj^=H@?s%p
zx`(CjI;)xgn$K^L^Xk{8N}I?Y9T&Y=&V>PeeUKks4U=-lmN_2SpPb&{08`Eiofq0g3qr=?NqPhtLFzn1m!3*kl{n$=R78
z@4M_WJe*G02^>Zy6fWDWaC9Dl~@K1=|(KelJjpPyKK^*-JB
zIWRz&_Z@)`BhY)6YS5D(D;~MGXGi;1{&6CJwJR*l6J#e;NF5kpC+vuC(9nMe>t2vD
zEOZA&(qkBaEk`JqaFj9yl*6~srFkGYjsAqKF<(SH0rkcL)U!~oKW>W%QT~e0DEzmQ
zia4WgB964InF|Iz9XB`LH{FJ{=pR635nK|$LktD-HSoRglJ_PRVAhWnDCpgbcT?|C
zIa|1q0H3v|T8FPE7+#DwIWcV!17D^DV@UqW0cw|1rmuQU4pK6Kphz$q2S@Lus>HBS
zRAy!-J|oWG&Ok^XKMcalu?@wN04%~PsM4VI|i+VKp<6r04O#8@SpxfZ~7A;Q^z25{{fx|C((AfxO{@N4;G>zj=1N>v~RGo+`dd;N)*yQEqV3JMA-2>|m$&1`K4
zjYe&&+9B*KijR09uh|H0o=c$d9bntv-^ygYWeMh>$L$gysWucNGstQD=V?^&(Q&se
z#vq|@doDXozzt+IS=)ZxbzQ6=oLag}q4tOSWwXgsejZ=Ko>*FWb!f
z7l}CyyMD3nEE8+J=Q?n>r^;;mbr}E3`op-$SUCZTipMbwvvCKnxYQIjGbhU^%h<+$
zT(9t#TgUcu(!d@G*A3XeR|O5Ga&J~yAFXr&^^_)qJq$?H!%xH4l5?_|Wy*`W;6J^`
zMS#25Gnh!)^mxKetzI7TbF1#Ib41HA#rJwYyiV^pFt1Smh>4r6D}OccygWJbE~DsJ
zsVyv1Z8}~pj!On3&jBbIPXup(a{g;`vrhO9jWj<56BE;Frka+6laqkPs7ZMgbMcim
zwBW$af0!cwMrW*L2;{EoX0!k*WbB?_XH%II?%@+|wRuzgD;QtMu%Smb^-*yDNjw9)
z1lQ_4HICmWU%bJQG+UZaKdP-{7&hN~`D+#SOy510nmIDmCC=iWUkU#5^0RqzEKw5a
z)0eSFC;FvBZoy%)&WDVZUv+Xlm8Ot?2HpleP@!juITIo(7v$c#PiuUuB{)8MVK-_B
zch7moz}iPopO?HOLT(b#BK3iKEpOd?JX|Tcvq(t+{K=`=jas1Tjv!mRrb6ih#?g@^
z59-5!DfLSCwxSbGVos(;kzhbz>dg)RZWtBjK$V6v5MrL?vO;1g`zhiI6;rVYdVb(X
z6F`5)(cZe>olhG95|Q}{nkvcozvHWwV{shNH%$IHJOumf@WZyQ*r;p#o)|Fc^=LRh
z9fA9oD{S4*;^~<9QzI@oBs*iijr;tR?5#Pe_nqP2)^qb2pu`=bmP2r(=RW)}(Q)Ha
z*!CYFK^?IPa8t&P3fD{JZ!&dVUs#;^mSR3goAEkMp%J}!j`f_2N__ePdK|p8b*V|u
ze{Ug37jjx4eu-{C<4tOCje`guhk_HV|4e)B6L@5YZXHc>imT~OA$7Pkde5h}9y~11
zJ~UH=#Zvu9__RNsXJAy}^j1SZP&x9r8m=02o%b-aoyLFS+IAbxNd2&^sdMF&6jC)GCbUjdX1RO;SN1tWcYbGUDSg-Q}>&K6RP?N(=bzG6)}xtYfA*iIzEp<{9(->
zM<&Y|U-h=8dHu){)c-%?t38Wu=gej%AJlZ(c&Hj}D8zt3?53ci0s-0)#wI48($YY*
zHXu+L_`CNhQ=q}$Y*RrY0`RQ3fPc<`qMV`qjE4P>Bc$qmcAx-+gFMRKqsgi7o{1}4
zB7##9zP4D*FY~;aYr{tDHk9J%@6H9uLBbgmpeQ0!IU^t
zAn@Pri_!em=_XKSTD?1Qk7g5uSPUD%eZ|Ck5Q|IFi1+g=Y9oF+;>RzpM;9Yc@z`<6
z7!BaTFEjF|Q>=2Q^OOi$Z*t#waz#yVuKMWGs#CsP!5w7sdWgMQ-=LdBXD2KAzyO=c}@UnRTT)4-~ecbDXcf
z@H=I&Bq#5Pv3{9I;;a>`82cvjFsJsPI@CH-XyP7kfR*iwsP`7D30wFn`qtIyXLK2X
zE$z2_CXqE#IU8BybzV;;7P(rVe>hU{yGtV*%BI9A-#US6P?E0PIW~5ydihuE6N@h2
z+S=i0P$ww$UN~mD;-;Vp-|p!PD5cs(-^aQfj@L75qiYKKrN^#Bw~v)Qc-K9MvJHg%KFgxpH9B2h(F(}n2bt3ydri>-OC#VTGm?PS#!S1fHNE3n{o
zzDwf)m4$G^ma8&**o*vSdR*iC1+y7V^0^3c{-i_m8b3#}1B`YGh4&UKA#dR&pLv67
z8o2dpztYaBHW5jLmZl)pp-_CL&TJS+h%zl2rI*&#cJ7r`VU6+D1C40@!&mFphuqc^
z%UL8*8%o8}ZSPFHh*Fg#CUy04rjf^SjMB&si{zdqIvXWmd2(H9!!Khx{no1a>>?AWrNF4PM%}t
zb?CRagmkv>xtfLVo&cu>glXx?BIT>8>Yko$Nd!JER5Ra3p<^Tx{+}>UcaEq{daIt7
zR#r>Tq4qYJ9beFCb|bukM*CNnklT1*lS~S_Bwq->6-@o`F5=;Xz{fsYS7)uQ!qQpL
zdXMG%K+(RTU$%VCD&qXw2&8)cf$`ztB-mre)J~}GY0rFG4~!4OY{NrUTRnVPk^3f_
z%CVf^?0tKJ!$ahBi?D#tgJUz_X{yun{8p7NYwtrcl%O8UTl(Pe_2Xh2UNa>&^8A9`
zQ?Q5fyO1SGtwQzKafx4084#)6C9L6tozO5o*bKN+kO&YB^v5RUz^4nQm4uf-rcD0>
z2ibD+9E(0F7M=ZjzDW@L(oX~SR&Y?hsXc9?9hbc3l;0}yedu9W@ohW6pV?#Et3C|Q
zHwOVp8WBf&+nBxjMki*3MjY-OC@Wg=}oT51ft)ke}E
zWKaawBKRb-(|LVWK_kpqAY-;O^#-WKMZA?2m^no_Qpj#sff?YS#agIv@a1VR9I}E&iO)ci531s@83!>w>igrb~_B4Cm=@NV3`F*!z`F3ul
zgDox3`+&PO4&zfK-9NhPSc%-F=o2>GjKDa8n-OX2V
zz!xvy`_0AiL8!o^vetK5JDT4BQ1+(8;&`)(+z?4ZaylV5HhgQ>fMiJa$EVx{cc*g)
z0KRvNpQ}Gxt=IfW`?jdAj4~BFwa=h=bnHphP=~>iYoZTfPc~1`>?G8=Vq#8;B#8tz
zC6?)o8;+Ju;evt3c8ipB;Gm7IJ5
z@@J9hUtw51OS)*wwHd3%{T^)qOU`V9o+nse?tJ3{IrF4Rq?`jEz-R(_Ui%=da3`3Y+D;krQ!PHHJSLz)yWTZPx>L4>X>bcFa93
zq4#=cp5Z4@`EbI-F`WJ5GEw;_~_+QvIR0rqt|1G!;XaN`W
z0#_4P%gR6#G$h=F4#2$r$L-Yg)zPXnPysyw8Xy71?Ov|Ey`fV!((cftup8$0tEmdh
zD`@R*dx45x7
zt`d}sn!bkKN7SM{n+3DRbgl+@RDDHm5*^PM1y#?T3qsDfK8?>GXr*0Kxn8|=Gt-!S
z10CD>7C*D0wS?4={S1I9Zi_7Pj+#!r}H_G2BsEK7GXrmk`e5~P0;>NF>V+*L6y
zBd5Oh+g^HfL9#9sdi#$@Pc6nOrYJ7!N1L{W%ej5dy=baL9WVzyQ@h0YwEB?EsL0kB
zw8iYtXudR28&;_8|BJA*42Y_0+qNyGbayv0q#~h|ba#y~ba#hJ!ywWr+;lh6p~TSL
z2-4EhUEdn-=Xu}v-}gsFahTbA?X|9ToyU1DxgAH$AGd}j7gM@|H^jbG;bl66ftB`M
zmL$UF^HS$*uVNm@FqRqDwV>Yny=6_|oaShpHp=fk+%>J90cJ|GSjuhtE3)9mWmY2|
z!`Ds7o?p}=H;s+t4~GGY(pG4ZGI8j>t5e&_YwD0?)Qj)3+M
z_V!yE44}v%a!I1gq=Oid#bn}ZHx57Cbfn-4(aSc16U{fet(Uogo{?z4Fh7WSknaiJ
zC+C*5VO(M(ru7FEK)1z-3mJKNuqNCq0ILM42)gtP4RwJg*5JZ-At9l_{3RNCnDc5k
z_N~qg8*dvY70lQ;I?%Qw?-}VoKVT2bStYNDqL-v7d!~eYFc-Na8AM14tYebfXv*vCk3O}`1}<2cvV#vo>d4zz`RR3ma_{i}!(IOFB{}9B
zTvqJnuFm!dX@VkR;n=X->x=7?+x>P-@DUlo<20LIsz_s@2|kHf*Uz6RtroSG6KEbD
z9-`yrQOOWAbo3S6pEDGeo12@WvBQMMa49;WFgUj1hTKKd%1W((HV9Wu9T?%y{NiQJOn4;w5PXM^yv4%
z1*z-%_y1t=dS{!%k#GR?2)q_(68OCBnWBrP0ON4<*xBckQrzy(+9UpL(7DwMk$T8X
zk`*0|Q*AZ%x%E#9za5&gDCHMOp(LsU*q(n5Oh^=6#!6M8x)i{XCI?dsMiL+{4*0h~
z3lR<5f=UAD(KXh@e)D3%&w(&03^}aorvaOi1Tl><5Zo}GD331;mYT7^rGr9
zFG+?>RS8!=#wg1#U=*w^q`ZU}Q!
zH6GhJBBVGT7KW2K`nDI*{p(j+>nWju`~94p20KcWP*U1m#i=q7wLsf3L6CK^#K$E`
zcOn7Qeeuas{l10N)~i3SscG1=t>L##^>7QJ?V4&BgwzWQ&o=E&`?G6=*$QbJz7_Yl&*_ZE;*nddK
zrl;gvO0BEEM3fXUz5izCS2-&5)|DXdQ&pW49@X3mN$zzM39XyQcL?Vb+tTNWSFB#U
zmKeVl;jatYt{fzHCLyfHUt1odbDXyi@Zaky!8F1;yY+Fqq$%h3Lu`%$;rwlr+O*6>xs^$(o0_$ngl
zpEx4RwMcVc%4%=-)H!y&bxXOL(3PPi68ncMuW*k$KZV!D{rC*)4ca~sB6@@9rpo^J
zvE-5zm&YJ``Gp@J5Kve&qheb?^Ed4$YI)vzbDDxh?fe*w`<%zH4(8bdyp|LV*Fgin
zxexCcz>a8f`K>o|(-dJVG?0{=+v2U(O(c71uh+*>?}}B%r9|k9#V2;|XjtACUzWc$
z*9tA#dd^+xiRcid)ZDKc3Zl`QmeU~E3*RMd`@*J&DeXrC_AEGuo68X-yrL@6&SvZ+
z{t!nOm$mKve(oVD+xkZZVxVHP=pxGLdrrZ%)8>
z4O*q0JR!@ac&@xP-}7-w{x#b#M}m~p<-UG-NrBYJIOzJ6Zj(DK)_ID+|ItH)XW}J?
z(8mb=Hf#hbB$`?ZAGZ2eJrek<9&raeFutuzv45%^_-$yFPA**J?mDKG7?V;MOkub%
zPD$j1cXaa7_qq;Ts{6=GMGvYTbj{B485xY(m00
zt;}A5((_LgbM2?ilZof+_xo?XVvL9sY~iEaD-O=>h(nh|_mU1bKPV!HslqARncasD
ziW@*=EgM#2^Yf5@8OUhp$`(d!z^C$=+*?TaYKNOzbmIY7W$La*ydUtXx*FwTl~A-p
zBm{^(`v%@Qh
zvVg=Qc)5|~t$%%U6UdZd16t3=EuWQz;MlTl!cC?TZTU=&VO(t$OKulC)x3!|ng^+s
z?v&Tm{RUUwe~9(u3>_dawGY$x{9d96i4~wKzx29h3hJqSnPeD2dw?^y$IeRbMy6kW
zR*5MXlhX<%C#G#vI^QZPJ15--dS*a~Glbw@stlSiQ)PyZE&3BVhh6Nev=a?9CH5yJ
z>YdM_2b38RXt}Vw=_rZW=v%)RH$(@k8e46*PQP=;F~g0uyPdPziCkn|@@&y^eynhP
zrPega1?%V^eS2*$D9YT^aL-V|Y+uN0DCwO8moh~{*x{1LBpMX`WTy(De%Ne)T!yK5
zc<@Y3P0c5ZiMhDA0`i?9zyP-R<6#R2yKarRt!*h5eRXJdgz`SV-`vdaZdSmufO-IA
z_uy)}CanJ{pV61q>m}4SR$Z{Ez0=h|6zui(y|h1NGu=HE&1d1?i4E4`i}c#vp?yi9
zYWpGDXNS}A5)EVe1goOA%PFkMfvg#SstVOf!DI2!Sz63`N$0xgbss*z{Qy4!#roMw
zU;6VOEal#QwZVL^|Nepg3bJ%2l-M}+T=z5dd+Neu
z=3aWuOG<60LlKqwKou&lIK_x!f`*2+zZ#=>$Zvp&iAl|310NtF
z7jO5cTHmbX`Z(TrTh`=9
zj|Luv$3p0H6Nnq!wRzVpOx92v45U^q#~M}oV=Dl
z19C1d&1ly&9GkmLQSjgqG!ImrKuy;LI$YE-lRA9l>#joi#S%){`PB%J=Y}6d25?dO
zT|@lZd2#TpP{NF8NKSXli)xLh(RNA%=^@PXIdN&yR)R!=EDpA~R6_R=2--9GOy!nm+`s
z^>2CGgvgAwco`X&^ld;8y(cGSs}RcBY2BerWGw2)mZ9Rh
z5kDi2G-YqEs*QZ+T&%URc+LTf$^kDgAave09Sy#|el|(O&%~<%>o^CM|9OM7PZspA
z%t^dsO6M9w1J>_J{{pms^+bC&w{^QxMUfkxurRpYh!TuXvZA8mzkFkCEFE|gMvjR5
z@;bF+1__!?72;Gia>=o32%0?RN4H8s$1q0#Vp;+@_A~!;z~ZBf5Z}68q0i+pHVq=U
zf|S-3(`zkG^IX87#alZBCmAUJOuer#guDJ6^}?kIGYjAT4#~!hVx~DpwK_BK+U30s
z@?2_rOl|2+)Ce#S?W&-%6}JT+MsXB*>c6_)O>ep6r}=y?r%pQ~y6D{4uh8s-PBK){
zB=Cce_{Y}M{eI;KznA+#l#`wPkHZSWo5t1I`KA_)rWaxXqYEa^&RkaUy)WhZ5cWz6
zz~FH)P&hycVwIlX%U~QGV?E8rq!3Kx#{haJAq^sw8K64(%30^0x1d%P&4%~-xrqva
zu}EVZ{2kDF2E-#4pq7V+?{vs@i>kWg|Hea6{*+J7XN|B;Q&3O{iHr>XPeC8>Xho-x
znu`l37?V;%s>KB{T>)e5h#253Z)Ghq+ny{@*3v@J)zuYt>rXuDa{egI0M_2v;bZ5U
z%xD_CSiFCZMYejlT!pXq#cmNj$DlgdzcW#7j)>^*mlHhbzzu5%D*(Xc^~Q&agUju*
zqZ6;g9!7kZo26jN0F_J*@KmBa3?^`=+v)|B2_C;}{j}N*vuf`5zCPKPy)U(DZS~_C
zShfMr(-wvJF_6L!G496loeAu9(FeN-dpu=Apfy|mXeb6xc|;Po82}T0&eb}$BZ)eC
z!4UjRmtc6*Ca1%42MfQviF_jG+m|YrtLy74_3LT4C1!vD4$R8NB;~r=sTvaBW}X6N
zX<$}p0|@U_LZ_spybYczP-O(X0uT;z{RLlX%I(|_e)g6aS2msvyJUML>s=r%+#fLP
zod#}3T&Js^HnbP3R(oP@?#hSXZ2tw?lb%0(`n=e2%s3e^Wim2wJNc3?U%s%o*oWu0ro-`mJPxg0F2MZB&Ho(0;NTP<5(&{-;^v<
znHW>KiJfS;NHw-6iZlZ+xo&{8-i9r%|Gy}5xAO+_-H|fLBKS_5ztz|+7-S3r`W{CH
z9s`P#rn6sO_*;TDvpgBG`RHJWtE^xx`^Bqid>$ttpVS!cjCp0tc*VHS(%nn=?={>9
zBaPKIp3SiKpYP{WKcah#%>k?gT?ad?A53{0nhdXAX;)03geM=vEL1f|+yEU1h>1)(^O1ik=$
zJ}LxaQ35viM`d3eG?p_mFrZm_JzC%9e*2vR$gJDRNr|I*Uf0X**K
z`_$>*;NGNfcnfu9+2MyNH2VVaF5e&z{R#90fCOW&j_OY#OBkeuHz7F*4lqll2)?QM#`;jRIY|B}lYq+o$KH*m;pn=5RmDISPPAY3N|pysqA
z!{K3cn-QQ-1H?iyo2u)!Sf-b1KjE+4#cjFvzHX$r@T%G_#wdzxN}Z_U+By1?5I#>j
zz0Ul?EZxZjtrG2DdN%)0xBP%|liw$|i5qOE=qmiRP+^DvV#cHGg&v{W2gG5As~-wO
zFRVkcJUnE+Y@P_p0JQ8y^#ZY-!49UpIf&V8U=$J!dQ_%ma6x>sFiTa-M5rC8&BV;x
z!VXBY#?1=$h>r5Fj(P`+j5TXf1DKOxj;^%D{hP{9@z*>m3uU=ei7FM94HETTe6q40
zo-*~alCqi#%7ks5-7vE=OPmi-1x#O{M>$TPbT6dno0$ME41X=}9R*(InlYm(bMMgS
zT0piM+w=Q+TwE}B0ZiAzkU8uyoLdo~#2vyfTnvBjEJ-UW!`k~3uW#&b9i&6zao5Ol
zncTE|)xjTqxwVx4H2rSCVrM+I{E#{4A~x}u)1O0FEp}n%HxuAg5hIBit5!%a#wG=*
zYw{jboM1K{7~F2cm-krEOOi2R-GJdL?8A`WHxZ%Eq$o~U->GZAu2+2i6PnZ2Hykk%
zL7!l6c!@gli%fn0l_zG`)7`ZpyVhZika*;OC0K>;s~l?FC(8zXqYnaNpX~hhJ(+T(OW2_JU6)Ul*jM=|vVzIU3yM;JZRN0fz}wlnVA_9PeKCdc5nmV
z3lj-AGcY>5OdjSz)^rc0I}AY8j^6LA1N8h&H!6pG8Y{K%R0u_
z??=LQTxV>h?1GU}c}mol0_D`=jZo`zdR{Hzq=FxwZPTN;)+6r<26@y#p{>lCk&ywA
zsf>JTqf4+F2hgJ0e4q3%+|J<)-L(=Y5~0tz%c#p%+h$oZHs|Ma6{62so+`4a7@8S=
ze?r)(VUs_8Z@UZ=lmjdfpG#-pAYI+kaWT>UM75=;?7)mn!cJ4o=Qdd-plK}^-+L0;
z?Dtj~m8*UmyHUX=)aO?qP0H35iN;-1n~`ydUKU^6S7=8HohouxkC>{Qp^!83dWM}$
zOGWi&9IqMpZQ4L!w>gd}T8FJ(gUWL^|M7%ZMtdo`N^G9`4LIDvm!Bq}Q$cl=VRF-C
z(z8S(&OY;Pa!Mw!YLh;K3;+*9v<}%*GKc7D_Tq?kLiKk{bF3;T!Y`wD(!Wp4$7xm<
zS?)Twq{(OK6?c`k1T3IwjHlZu6ruIUhUrI^niPszfiW33)8ygwSheHuvx-W7o`R(@
zfiV^Di^mc<2wt06CBx>KL)Wd*$h-jafz{RR+FEfAK6sW)8MZD*pbhR1HID`Un$)nS
zTkC{Rv+wu1f{r>$-%a&uI9S|2<&U3gr7Bu(v9tDn?{PCDZ1wHOVQNRI1f|LZ5u;#;
z`Z&~@oV5THzOjq=eTx3Zz0zDQ5>*(_Oh3iOlwX}}1G0LSj-DzCiXK?N!Mbj0FBwvz
zPdBysPt5g$NK934yo#3!mn3`;*KuCSojoeon9&5aOarW)sJ_
z!u`hbyWLNjpZm#}vPvObvA?)-`}HQ^)Vk?12JdR7v&R2U>WReBq*-Hmis;IsDdTPp
zzD5c(b^(^rKC9};vAAls`s1#czidC{TiKb3IYKH2Hem)U5@SChsBJB?e)emmT
z?v?xY_irT)?)np*LHD^hL2^gnn0nGF4-H)Fn4lzT(=ZV671$;^;mad_Gu$B{IYqN0
z{tP=NZpb!!O;fKVahS3=SagRW=LLRYxxe>x1eM-;|NR)2-(?Iy4G=tQWZi-;guZ!V
z0)JW1RwD8Od}(=)`*I5%S%@SO8|9kMR$BnYfe*N8miC9#4Vd3bi>e8Y9-3t4hV)Qh
z#@pi=qDYMpS>P~}(8|+Dz>KZIgEnsZ{lop2UX$mi*AQ2ZV0|oJNoJ4OH+E9s-(YmE
z@J&%1i^qPw3k(>_ns=tOriw
z=9H8E-%nkPRZN(Ars2w-*>cK$r<}T^;GIvmuAK~p^RN{>?g(hRkvg
zzW^27CR2>H6SG|}pD&pZM?D95Wj~#+5T#I~8d>K6j0(5DE<@>{j7NJ~#!jXaU%EOU
z4qbis*mW@|_;Hf%(^GRV!;q(4B8(Mkg#(jq(IE8al4L3MwgGp2lq+y9f%T6;^@e
zS4xMA-buV_1cS6-x3o-p`pSf)_s``;1OJm8JyJSX_Ddz-jN{GNWB0?&Q!p&?Qy~yhhi|DJgHzIk|_&i@a+3kb`FNc
z*2AU@BADqY(pReUNqz*s@gN!lYR&KbX&7%yVJboOA2X-#wu6PI1OO*LCw7w~R>tS2
z({O7L_HVZ1=o`SR=g<#L;#N15wd~%8*57~;BMN4^o=e;geGyJb_xtTAq_NmgS4+CU=h88@P%69gtXoB@wi)
zRrJ#Dil-DQIaOnm6KE`7X4BQ2NEwD&lztS&LV9kqA1^n9GD@Eqg3sSJ{up5two2m!
zT%LGQU~CEbQt~GlEO$=+1Hl8|1O=&ckgD9FUV0xBKk3w#4g
zHmeb$jVQL4SXeb^H?rH50pdYkM=LJ>G#OGBr>bjB_(4HBd2A4
zJ9ZMPLk&#no1>r;M;LP%v(WWzHfJ>v!9EjSW!UA%*fIsgC>uiZ_xRwlYO-6mw
zZCtH5qaO4m6c1=~%&E$x!j0*+2bGAgqQ9^!&NFs;fR8IJ1
zg5~=$E&28`Z+)%en@d8xl~$GZqE;XTX9EwhG}IW*!^$P;l10^cui!^d=7O?Ks7*+6
zOzObGmy>6@AB8s!KW(^J%37~~`qb~egm(N?)@U$nXcf1yvs9JT=%T^Guz;zTdzYRM
zbeJf#yuSNpu$LLPKB6x4``=gJYB%NIbmF$;8o-XK|3%0fSFAyL2a!#AE1(i46T|uj_ly
zzY2ZIT7SMWEX}$jREg_V)<~g`Oy@Xc3e$<>23_t
zW7|?<)X+Mm4s0og3K<6SQF?3+4J{q>;Tq#29fQEFf+4Fy&`6;#i!1c*mJbu}MA0)}
zr<6RUhsjl_L9O^*^^fNzmZSsVPz;{A8OT#^zQ#6QDsnf1oO`o>gAA26*H;w<1Xv!a!HD74KZSWl?gfUzP
zKJ22P3iBHfo4)14Z*Oqv=nI1_baXAqeg8C3i>-Ze?fA-|Ef;l#bJ5QZW$y^*mcN(e
zqfwW}uoxN1?dN6QvZj;AlJNNtMG8h;TLvl+No<4Qa?2B=>*HP7a_ixHrtAxd@6g|{
zp2F7h%VOF#af2+(&3U5GZ;FpUmZu$ZP}K5>J^WJzjESMs@!=utJ*xvf!T4DNbKm5N^Lr?24>
ze_}QTWnr)Zy3%<1TDvs)08!THBfaYjEAQ4HxM5}D(kv9I+JVyPt)Jcv^pU0&B0`iu
zZB%g|%x7nzsKRogp4k|vm_{CJX2SxKMrgAyAK3KQ5B<8f@NV6@-$v%W=y1*BIzSP4
zmR>IY5X2gpbw97GM=7hXdH9gh%*aDQe+
zCGC@|#cSg_>uiJ_wtQ;%rsxQ&)PAClxlo)NgDtVb>Jm_Aggjdc`CBzc7=UuagwZ(6
zAXqqdBWrdqK6hJ_nsiWBL^MIVMoQ!IHulliuFnro_fgL}W``&PlZH1!fS>e!Q~X&+
zu*gp_28VVFDFoBuv1pUk51&(l7e>p2i)bd)PyG!M!rjk
z|BdEke7o~VFTebQ!~9Q1*DN!kS-BnRQoiYari#7n7U1%&cq$GDHFM)g-}ca9yrn<;
zBr7O4aZ*auNqLEB;{{)o`Ps_sddL!wY-UbnU{+9wsZ+t!LgA!N`;*)YMvB%hB@)zg
z>#aZB0FVcNS^2^cbo&hZHy=sOkP=5j!UP(L^ot;pkkV4E)7GHtg`5K~zqn4l9mw4P
z7wmX;+9^ayxP{@B4$S$d!giMABqyKBk}mm}MU<4;VrUE86A
z5$!}3B`=zy1s%HWTCobgdIEuXcVns-<5{wASRD1-NjQ0y*=8S6XpLfy@-!q{gzua+
zG9r$~Qj@aUc;Vfo2Ja;XX%@>jQ31>JX#q~71!*7YI67n3)GO1c27jIG^WCz2C63Q7
z@lqI|q<|U1J1N@}i))D)jO#pXS6lk?oKfP&f~jQssbBK~o?^ty(5+Uq&i(TeT9QHl
zAmgs>j}@7Z-@cn0UI_T$NlOe{rT8$A&bxNU5pVZ`EqtPns5$&1C1EiucgJoXk8;of
zt4AKKtGN@JdE(`_IGH;l34)<63N#=KcAZ-{pE04Z%?_u`
zvfJz=QlAozGo|7yw2bC8_8Q5vm?>}7uDX4xMc*vSzZ-)W5ycRd2^pnR;gLj0Hp7)m
z=hX!5MlhP)7$^20<$Q)L4|@2LrB#@R?iDZ7|Foh5@6`LTh}(OHZZexE6dd6bV?^wN
zOMyZ@X&+ym)z+Bv-_IjwAz$}?*4i>b@K96mO*erBDQyj{uaXWGN1(n^BDG0@%ne-m
zWhuMSg#>}S!zHu7bWJH6z`!46%8N?Jfwv?7X)w_^>&VHf3aCR9}CaHF!$3F}YrpyP3
zajOU@VKWuwmF_+fBR!nrWarp+_hs5(@$Av_gIX%rI;kr|XnL@;@g<#%EfuuT4>%QHYQ*^;(cX-8gXR^L0P
z7iYR9k0%O1S#x{_4i%7EJC**22m9RYlgv7Krs%|i~4S0Q-1OG6hxgrlCbTrN*`B;;)
zpai7Xlfz;7rh#gKaOGWKZHZj^YR6!-JbaM&(vjqM|265)8vWo%@8lM}WKPMs)`$F`
z*|>#%+dwR{C@I4QD^Mt_i38?OP0FH-B9$fIKB1uVT9F%OR&LK?Ti437?2N8n1M&5N
z21McO0JA>yvs@)KIsfk3hq-Yw^7MS?=UXCP7dFY~yMc-=JE3kZiT-^{uG1%7_na-S
zFmzf+$o+kZB1Ui}VdHhw=ci*+2g{m%-`@`CF?=PZjr*C$WvrQ-;nSYG!Fj7Spn5oA
zp1%6;K9z}{6Z`(UkGs0T)w}u4OYqE$lrDu`uB#<`dWQEH!cts()lUHc%dUs0=5IQcV%{b%Y(2W
z0?Lx_v*Ri6iR7O`6f$v8$0`DkZKtD?}(&e
zD+?J2ezp@V+wrPHAbikpxrlwlBj{PCrS67i#HsXWpgYQ!!ssMm**z|B(95~N$xsjOk
z;bm16{1Tv|&LFcy*$)@mCB-R7Fk~2}FW;^eXrFeKZ1nM6A3x!u!n0>GGsT!rf{6IU
zGx{R|H@oZti(NkW3nI?#_P!cq+>(9PVDasd9w_n&guVhfQW-TccbLV+AHg}ifeTc`
zH~jQ}=BJ7lf24fnPP{IB46>pzNcr?M*4*h)5#5*iE&99zGhse7rtnCe0hEH<~R|}@g-b&sJsNbf8w+HSkF9w-WWJ|`u0!k5}nt*N_M+Jx5YozM-N`lOfyHM|`Xe2IJE2@1CDA7k{^D;gEs!@<+0ExA(7aBP_shx045;p;w|{3co{
z=~(voA)?5)^Ux03UQ*spkvw)*eM`unI=T>{du-a{{<@-?6
zNFjf-bl-SYt7_Yc2l7chIK98pjDtf+*DIESqc$$Qoo;>DbX?j#MENalcwBtZ;k54Z
z3B$JX2B9e)mY4()Bmrk>MnbAPdZy+Fhnhqp4(O#Le=d1hPaecaoIgl+^?Nw9pK|J$
zK3}ANFrz6n5;JC&?$+M;)!911c<|4$BWre(xGcA
zO;dY)kStA!$Zt3;7T_me6^F`XSx0O|4#^Ga^n!0UcLI51;eTG=}b;m9cC%IT@&%t3EpcNL9e4=el#
z$ClRRQUCz#E!*#=GbWOjY0B+A_0%2$Z1V^Sr}*~xXK^`cZ5;_okU}1Pk@lx~z88O;;2
zd>A(EbuWZkzWPa*Oqns79!5M@HT)cZRWtDc_My%IAoX@>%C=0XO%7PJ(c(SDfU9_5O72tGYmk$_l`Yld6Ri|Kd-%=Tc>mU
zvz09Z>lju98)uMq_ro-xLJ+fJFTVzF8+#z-fx0}l7}yu~82)CW75JJ?Oz--27RRc!
z3YRrGWp)_ZRZM&p9hgJ)TOk(Nc>xXXD3C~?7ic-`d}GNhz8cuNKla3o*USp9{chjx
ze8v49U4ykUKC<0{u!i^IKNk`4(sOVLFngeBHv;3Ky&<6ux}9a@D+;xF{3DhmbGh(K
z98*8X;?(z^xHWU-`fIE-fgmXbdC;F0?A{*237-fg&c23wmrSL9O{e#*T6J?N=pm}m
zHU;a%kfy%YAvi8^XOGr(J{d>0(#*_yrQ(;%otTJT`?p=LJ}|&#Wd53-)8U
z|8ZI|3FA%@Eq%hYHRRYU_Y43q_$PT8JfeVy8%^MdQ1*3b$xq^BpkLRR6A7n6K^{r^
zZ)*y);11a(;M&BG8?FRVfmY|A9f7UuJLlT#V1SxR{BJK2d=tiG7L6$rTq2MjXMQ1=
z3MB@h8GyEksixN_rG-%_aGx0R)VNK?r4JoCYWt~%BtKW?FbuoANC)ULL><$o;l0_SMGIqsvY(e$IW74>UIF%O?f0N%jk{Q;D~VO^ji(5*)t9+B{ZhBFtDC5^m#7_Qbd)Bt;w@uKZqM7PaCykA@2aM
zJ1C^xmc$E`VnB$~8PE>4>4<5K;UPMPvYX-a*3zTcx(-@UPN&$k^q}ctBIS0&4Vhl~
zRi|8(v5%B8Ri}%cw+rJajo8`wey*aO89b^VcAqkqi9E}}&MeOm`r9k!3dqnl{Zfg>
zjg3nfn7=9|o+m-gz@5x57nGw#_nVH29cp@bAwC;W(%s*JJFBLz;WLSEG*u*h)77H{
zIcaS!Mc%yW$}J;)-Dn=;FvX?-Llr%p4q}D*Q;OVG22#05O77VL59Z9Nf@c*&NyOKg
zI5#3hkgT?#EN#r<>2H=|4l@-nEJMFOxaW`F^Fxh+jW?
ziIrdCL_Y^rRF_lrA6YRq?WDP9D(S|ArlAH2rLmgDw=Jg0pOR>_4GK40Hbos~9GI5n
z>I-NrV8;1lC&Z7nABWV_`mTTzk8mhDloPlzl(}G|4i3Mi4S>RrK9X$6Pw$gVQ$PKr
z>x6z3qNaQo%qSFSK2V;9gPbLKDmV+KKE4jS>?%0FH{^)XS^6%)=y%AoiX<9vTrSGb
zlEnTf=MFh#=_m_l>M@z|hAhxOo7qnS4gIOGb=%K{f=v4<-GAKH)je=$8sw(TURfDU
zWm)C?&Oy^uDi6(CBeUr?*1NT5A%UTDuRY}=`=s0T_8oUkR(^I
zmLnzA5ir#hCnz8{o;A)PEI`x~wV9s9=LuZk#P{Myv%u!^|HItS+|*g8R@izaiv`T_*XOq!#-@l;VcuF*G6-u%?3|d
zWl-frHS~h$f(AkCBeE|uz4xK#Hztt)$cu4QPI^4yl@EOM{&Y9PCgY8H&=aGaTiHY)Zy9gU}1q|LrQ!y#)b=ZJSX~Ju0YiK2cdB
zcn7r&;ip%BB;KNysM(nRUv1TBW9%d%s-r`kO{@GKDZecUiB)^guV>Zjj|S_&qdFp#
zLACi&gCZcZM)V?h?Hgk=X$r)MyJGd^?3F9@?{+%tO2annvZ{yT8vsIp@<=n@*46d&
z841suJcX|TDAf2`3Q?8JMh#e=1|R?}uwm>Z#~iK|+1$lJXb|T{{ja8~47TH8mzrrF
z`G0qvv_n4NHB}G4r7`z2FHWHvam58*96tk4UVC(@{?x*1r){Onn&;{_7I5TB)$pWXz2kgd6L3kLGDD;eR?B(cfXueh{jzuz8y}
zoy_;m&>9gU^u2d`h+wVc86c=VD)l68W}dA4&!+*Ay?;gmAdX=!SiBN)S3V`@%>^j!
zZ-88ClT2~TB4SKw|<7P0@Gx)`2ApLCY7?0yOxKlX`%xtA0+Sr5J0MN
zT3VK8W?kAx4HsP|)H>WA67`w`vu}|F)o6X5OGQdt
zX@{q*R!a;EXjTVY5j-`BBZR2^E2KuJc?|$5k_Gd{)6_HNaS_^F-)$m@mVqh|mcaBW
zGZR?2&dqfMmI`|4cGB`SjV}vR4CES;4e*}%6}zK*3Pn2}Ig?HZ%oWpd7vUuVHX-1%
zpUp-VKZo*K4~nNFWK)cC->9p0{H<@MukzOV>b64*9DQc-yxzaZB^d9i|0}U3zLM#Y
zCuo+m2_xom!Tn>=AZ$4Lm-kF8E$iRiNNIwgNuz@pH9lr`jwLf(t_oh8@xb|iv>!OT
z8uH!YmC09-8tF$zSaC^8cu0g@5*in@@>BLX^sbf4?s)PRS^LhQxnN|ZCc~BAnzm%#
z?id;%qqK(<-<=G-7}V^;>IJ>HL>_>+M&?FD!TN1}5k2mGfJ_^Utln1(FnG4;6%3*z
z1M@e(_bY{b6?rgpMa5Wu>mjW|d#%Cw>BZHEfMkFLa6O#SUD5CB@5|Mn=m~!QG0A3x
zV4|8(3;~&fSdK_289d6Db%^?X#D-xQ2<$`oM70XWMTPM=?TzcvUjZ12lhXOjt<1Z_
zR%5Q;05cyAFC8(pro0nM%vz~<_=+TVcp`mF
z*L>0ZhJ--i^6LJL_YIf_@G?gkg;WrCHW6S({G0n6(%@h+NG##C=Cb%1D~`|lQmAMM
zke(MN^o9QH1&+v&)td;zya#;kJ6nRJEy9GRsl^ovV6w({G2^atRCQ^*OXceh>-)Yr
z5_~Q&ujLrUFO2{g0k+rHdqoyGmn0*pd_dpLuei;k*WOjGbh<(qvr)ITU(T65CDYX7
z4x@NjUSD_T2TV1^O#Hn{tX=
zV_MI0LqEGyH#0xKQz1@<|J*xpjk#>KBF{RL7?CWFc%z0q>U-3m0L}4$z|@ms|Dlt3
zMOs-PSLyc&WVLohHA@zJQiu`-SworOI1|
zRvCAb>06`2ZCE>hoMC!Wli+KJYDj#$<3FowMphUVX%rsy(^#741Mh)Q-!%n=U;dsB
zn0>l@sAy*I)`$zr0zk)F`%qn+gtBAx;54kQ8PywDH{t+bli5t{pToo6@+hL#v0wYb
z&V(&pd+2~NUHE!)^WVEWDzt;>BJs;@C{C3l;sg1$8I%sj69ATUDSr9dFjYDq-hNmV=4k53;&c=#
z8|GLcn4YYWW7>Tbkk|oVTbs$lU$Z{3ly8Fv3o?I|?mw@%W(S{c@2=yEPuOAGP9bYJ
zE2djmiOfKd+uB;3A;{}R9EQsM-E0_}nzF2JZS|67)&BA<5(Q$?R?PFaSl`+m6<7Hw
zD+I|wKgMVeN;w^&BdijQdKnW;QKVSS$6#aDmSQlZv0nHfa1xq3;*F$ndy
zqX;RFkhjI?$-@{R%YAh~Nq)B0F5RBzE!Ng|TVd;$K__IPB?s+?{C{>G);fdYSYj5*
zi%*H2VFzYngrn2Yr;_0Y|qrnj=_MFQjM&@l9{i-~z~XgJb$QJq>&
z`ZjD^HZotpy1j-bDj~dC_&lr?8R#4IOroc+GiZ5Q36DF~T8i}@dLLUk^WBbJM-Xdg
zP_OsrU#fbPx~Caj{pY=Xop8)XS8%nINVpNmyP8pnY^`MQ@MU+OHgiADwn27Z6@Vj$
zkXOvH`Qmp96SK14inOSRxUu+R{+(f6M`6p@D{A@5stB`Yx=+LSVD9MEFCK?w>p#ID
zqOuH29842pC6kc4U-O+-_u>*+15v}HWD_`+K%Q$yz0-O?$flBxPMCU;#!>>*7n>%x
z6PVU3Ob1BBC1w1YnS9>7k_?cktFI4D&OF`yiT)02wrzdd$M-%j17CZ7OF0FtDSU^v
z3`C?85sUaXExQT0Y^kmOPF7g<1xdUlyq1r+xw-Y4Ej<^GkUns~Gh9Xm-~%T|BAo4V
zUz#%FfOY;B2UJfPz(!!rie(BfmDIzg38<(95G9YrXuw@b%kAa%{vp@rPJH!`CyC{r
zwH9MfxRMU!nae^0(FqS-?tn`z5G4Z{SsG`^AY~hpI6bB%{>bK6kqxLh0dEUVcz<=B
z2|jodFu*K>l;=4+o~-zv5%psb1Xn-uSXv+`h8B+%{EjHB~Aap!5`=GJGdW6`DFi8V2-g!AaEh
zkl8A;NGxzL1we}I;MQU`z(>Ub!kZos2_9kGUHL4BGRo0FQiy3Zpr&8s=!sv^(+8G0
z^~JLzKmC7J4NmS3UwbyYkL){AQjt4t)^#1P+WB)%Z@!|e&@wZl`dpo^`E|+Ven}Jp
zCYzR0N}prwu<)YN6Bu0Z_I$&!xm5+-ty052prQf}A|R!uCGY)7LSa2$Y=k#}}Z
z&Ju7?tZz4%kI+zC&Uv3N1$+B^O*L=kasiBxqmD^r){WZ{jxKim@N?jx~0dPiaaA{bWgN`bit1RE489r9-I`Q2VHwzcsz
znqesmS@+yoflNu9pM9ZJTViV+){|r3CjlfB{{lKVRRI!OB}F=pLAt=c%U2P<7A$HY
zJYi?>SqJ-wx|w@62)_;a;i@4#%Lj)awqGWM))DTEZ|a8hn+dsU6X(qSDd_nInuJ^@6Rpfij1TOhlh^
z731AQ%#!kj1k)m{Z!i!H%E($pY7KHMDnQI^hWIQv7^RhS^}MG95=1J0ILIg6v|a1RN#M
zHZ!9gLh_W2A3k9NGW@|4^;o8*^r3N>oNfc(JADWoh&GSP9Q(4_@;|XqV4>%&_BEozoBEd$(bMgEA3~|r$i59Puto8^%+ksHSF?xd@rU(!^4xDlRLQI!#l?VzrgL=s6(I8Xn#lyXM@ctZQYHn
zx%Fd9(msJ^Mjc3zPse|;eR3p}MN$NamNh9P4~EXZ=C^}6Q!*LFyGTTQK(9Q#N^r}w
zPXJ;}Jfhx+cn@GQO{yNs7WNJHx?T6%6voA!35*Z606>N^Yn{G#85}b=YyPp@+h(~M?>Q2Os=5I9
zJtX4)F2)m@UL88NLK`s7_zG_%=$%FQ6ruwVE7gW5KFU0EE1F>gzdP`;4n!
zWYHQ++1mUX@0&Ej;CEQ`n&YZRfrHBm{1|9;gJ*2{na3xSu*Z}NNJHfz1Hv1Rbd`sp
zW4{3wd4LTbsW4R8(GlD=&DVAfKsN~l1g4*|05a{^>Y+5SX)vNHF<>U`GVHc4OM==r
zkEKvlkK&UuXuOE&XSJkUMdz0wdJ!c2@
z6R80)vsVnY&RlT#KCo0cv_^L;#%~=k0L<^UiU8PK-}&X)cVN{_5ui&%6PJ((QR~||
z{?qn~-uE9@8=En<;r!{`nZl+Y8G%E;2{^r;S0_Ln4p5z|0Ylx1l;;ogZ=CGgXqKHi
zzh_|j%bx-v31=jzrj#X%67Cs*H{O>Lg#B#+;q*0f_F@UuJ^K2iLWqo{=Ix9Y#t@aZ
za`Y7$zz%{qYy5-X{b*@GdC!v1x^Zv2^GOs3#!fm$w>iK9)mT^kjaDsUpCBQRd26$~rPF@i
z(zjv82M43A)F1j4v(Q6RO8T%rw*20-N`e*!Xp-vCdIEo06Y;PrDOMnOt!$xKZIren
zJM(a%#QgcbAYujaCJy|?1M#rz4MuwG4LYP7w_6YGgn0xAFz#&3Thvn7^fygkxra?5
zCBNc^$Hw{r1b799Sn_O%Dz%@fkvIup)*qol0Ql{kX=46)V%H4HZ!XjgmmzoSH{kxS
z8qO3KuZ6%_%RfvE;d)D~Fgw@b#Wk&_jy#lHH2OoKhk3|DMfXDT6ZmWsEtomgd}Swk
zs@qrVGzrM0;ra3o@8HP#rqTQvf;bHH^z?qUI9m>&WmLA{kXAlLaJ!nHZ4NxT)@4@#
zR-UQqA%wsHGlqj?0h5Bu@!UpZPTSCzmP9K271(^RcfSO57M!BghFo1K-QGM}KYDu#
zUaxf^H+@m|7~s64V4!|y@NDTYxK%GsAnt>eyoUiuS}nL33tz7|OVn&;_MBb{#&l@U
z8CM>`AG)-l#^UbWu&8S!5VN%%56(rE*WRxXLF(3Gxt_BTHAUE-L?944Q$#MQc5r%N
z@A(EgbH1ZjuB~#dN^#ZrYjU@PUI1=d4nm`o&_i`vtTWG>Z#Dagy@4o>3fS21nx;v|
z0bIQ5U?>+!kb8CC9VWrT;^Y8g3(`jPFzR6>VvzkWtcn#L*>GpN%}2(}hHt?==UZAa
zeV;`qFo*RVFZ*Fp4Fuky*ck-r4ii?ag5Rv!>qp|mZ}fW0^iEn*TU3*W|16H(lukn-
zX?HW8D)L)tfj7y#65&Upqv6~H&5>mVOfm1AaJ}P-eT*rjpWA{cq75?45D!&^nY?lD
zBtO#ckKMpSUYI0&8!y`lQ^)D
z48vh-nr4**b)o*9n^o__!sL1^q1!3NajkRVT7RhR)7{0Rk+YmEB{hyg
zM5U@ZB9@3}#7HaW8OS1-K&a-mG?u&_Yrl-VJ2gACEW{)Bf$9X$uXOOp>DOeuwhgX6
z@Vl2``};lgg*u7w>U6krh8CoVuNhA+J`x`QsoHo|y+w(PM_o%5{?DVO2chst2mwG8YCgc#A(Z^$1dlryE$94cKor4{!u
z9NDLfEld*}3B0fC?$P_Hn%=}a(LRz$S7%=6t7EI=N=h#*yf$5xE6y7V_t^
zFVN6Te)R!|U@x{lhg_faY#x5k6BY=6RttCe7^o67_AScjctNFk?T)~o%rKPo<~U{N
zs^DQ`F8TZdL&UH3c~BM)Sk{_fruE$e$^}ERcDTtR`TI1^eYNT{;DblPg|$!7`+|x(0!e1VuIP{V=FBfH!RVc=cHhBu^Y1QiU<97%+xEdY4;&
z!eP|MnW+qP>@0%N1g-Kp*v!iowz!P8^eP&c11s_B(S`!wB^%%%pM2d~aC9T{Tj3$@
zIwlNx+k4<@0o93_Dy<{Q&#{lH;7PKJb=7U24C0E&z@t-@B)oChT=;@oy9DP>nZv-c
zkE(S-G?vvy9!k1`fsu<)*_luVK9_ACC8n;XQKl5f3`NmRm1&vZ@_%_C2e?D%lJ*f&
zUivNEe^$$}rJO>-hj0Ay$IWbB{pAqoy0kR?8+X9J<(
z_RbDqieNH2C#xoZ%|~&<0&%L7dR;tL@?bf`B;p)c-gSsEgD|b`MY*D-2K|8DuDfyB
zavjsqbD@OSRHG4SZpm$bDHGsV8Lt302403&wkK;#}E$jaC_Ksg)H}VtZY4o5PjpdXSdA>FE@SZvKv;}I!gflKr
z!~fKEI-k5zSFYLZX5VMEyP4u9nndjokA`mCSO1IM)3#%(w8e$Th*erR+Cng#Hw?+7
z-pMXJU$o&N&d+iDx^qyl)}v`iDO&peXZr@e6NXrW?QZ-=z|#YQad&8!W370#3Qar7
z?&d>se$9%46(YwU0(bG_UPG%d&>ijQ?R}x$6q&*e{+?60ilK0d@G?Mv&=kh?#pc(JCk8eJQpcS73n}N@hFEk_2>}RM3+4=8^Rw(F0E3_vto|r%9qI8
zG-qRSx18*#9kO0*7?Y99o}Z=_enA`$(r7bRa3Sz^y_d5R;I%!2I{|fjlbMMCkn>^l
zv>RDZvr`PKYPVKPqjtLp-PL>EBT#t2^8C7Wf?BLrD{;j&+Mq+k4hTp}gaOW*G;XT|
z6XhB}P3j_}(8AmtN#-9}3W;4f@pheSG{Bh@?2k1IIKS=M8Nyfc8u`;myu2lejKP18
zrN+qAR7?{KV;zpfl%I?pnmaHP=SRGQbQU=jMO)E}dzr5S&F;76j@Fn-1##6sl6_T=
zGydSg*6JK5dSryY~Ic?|e1kjmuLiHlb$t;eq|i8>>ZcbKF#&UKU=mH
zm_iEJKr|lRtn38Qk4?rZ%&(R=E#5MDUmnENlWsbGyCffD~Yv*_qNU`K2ofYypy
z&C1126GlcxaYR$-}Ds>y)Sd>$ROPH9PUUQX
z!tCo^eKYp~Znb0%KV_NlFesi$0Z1y0BiZetE1P;iNi>ZOl7ZFuR~4vjF@vbgp8(pU
z-l>MZYbsnDcG+~Gb3OF+N&l@56pT!VfSr~-P>Q8V)`6A``S3Z6#sC&B>EkmyAmm0&
zDR??=JBcKsE}~UPL6p(LP-w)rVU|r-hiI|LzcAy8I0CxhvS{_@&^)zvN4^!bV`NXN
z56eI2I1N|IPqXcsiJ9we^9}GKYG~M3qRzx*vqWPZU&ME80@R;I1bL1y-=
zpg$4r1XQqMuyAmUZa=F(HrQe^kETX6;?aypvhw8t#3`i#b5Paao{$Ez>V+{);vChBts#k4nQ2h2fhKmYbBM&o
zi>S`NMESkZR%%+`aebWuy3~NlJW)M3vL$MI-dgvx314COm0C4@@l%c*!C(nT6XqSE
zXS;l4eTIuHW|6$LHIpdx8sMqQvz*L+6pyp-yrTnz{ur%efIA2FsYUDBPvAUTIJqD4
z;|j0q0Pu<=5hli%y}%m4eVR5HVqn_9^qU1}JDYNhb%4x_z%mA7=xf7;Ba6Wu7h+iA
zA6!&D9xUQ_;z^sycu4lCA!(l1dLgTYr)kbQ9i|l8lYSr@XTn@O19f*fp1X7kpepin
zabsh{!Od;3RJ|Ceyud?=c#r`SPUmF($Dg1*=~7QuL&LJyA{`XCRVU2nl#)6g$edj>
z{uS^9=(}4otoW1CRUC$G3@KBqFdB#v-Z-9aBkRdlea81V!GLb>3ph-oH`_3T0mQRn
zX}kjic-m&XAi&5V@Y(C~(^x+nVAEa%x&QXPEt>mrBg_R7i;Nzf#UaeM=CCvE_bu%}
zH=f^V(D8RApMZa$sv9|=!ou`J8nUzNxxZkpU2=nQd!IcRoo3yL{u8xZe@&g=!Ikv6
zLGUZHYELVdG~>-W=05`kHBTC!%)grtcuyRw*F?R{sr00MR1@52w>Qd}1b@_H)I<45
z0)?a!hx7B?OnGkzG3u9syn*YTz+dm~0U9g0s+x|5cNdN5gI_ggYJ?S?OJTruZT$Y`
zA{IKVJzj0Dx29{jWA8gr=9w;!%fF9(SUFn>5Ot3pw6F5p9QsYgDLuRHx{iM6SG$q?
zkC%_kHD@$sUh8Axgij#q
z>o^EXxpSA%=2D%Hyg@p*G+{QsCj_^1trmTLNBEr3zz1?V;o6*D5GjD%BfLhv;QFk6
z{C<%Ky9xru7sqjCD>fiD+l|P0b=w3$+`{+%)0@QOl}vt~P4tx0(dq{r~dou0g{r
zUn%I6Eo5n8pO0j;qmLfA+SJ4vjGN$tBSQ|2FEc!rP^u008+WLlKpGdP2woL6l^IG=
z)qd>HRJ)S6Hp0m7xN~k{-Pan6eVl~U5&1bJ`6>&GP__)G39H*$_0!!=$)puQ#z
zEPUWi;&s2u3Rph6)t!s!?5k?=CDXS3{8rYJW<{SSjs{qj*Jxpph++?JkAZ~|7qG8i
z-YF61Q`|^_#u_GJ-0j0MkQv{Y3vZd*whltE=6Q9Gh=;RIEJ>i1zqyk9NDrE>`@Z`7
z4}FSX`REw3v2+Ip7^U6Gfg+nL3}`EU0$ebn#s*gbSe2;SRHiR;3|(|I*v^0^<$|%tt=bHXGJb
zR9{iB=8_gWMQx2}Yd{T>8X)ZzI?jaQt?s#EnqQCIl39SR8Q+fnNPnDTqB1q({x<}2
z9Q?bAjfyp<@WPYHoXXgt^*1Erc)51-7pmeKABN~i*dN@%5|-Rg#udEPgFFKS%fBoN
zLy-TJKb>9J#>4Krcvmd<&f4$LK+w^vb+m#4ANh<|dn%j1>0QKju>o6&iIGKG)c~t{
z8he~2=mg}P!6zM1_XhCy!E|vk;K+WMdAP^nz8i11@T61)5pl4?iJgbWUIQ|YG|#>1
zk|uq;oVW>qEaaJ{Hj_P#+!?^{n&I{0lD-~!e}Y=
zi@KjsareP!X(f9@o7uba6TnOLbrnMqqD{Gk`LZ4YV;fb)qI8BW(cvRIbzHZmtlDAz
zFC~6jpJ_v6NC(zrc)QuefJQ_)AX@|^5IszBGx2F80hgugZf8b8O8>2+#
zS_LV&QoWsPyyqz%)*-GCXv8N-5>aT$Qw!owumOr-F3k7v_E7QZz!2aPIQ_fdU^th!
zn5E;`9Y;kL##!Bao(TJV1W0CV0>`T2=UaU#A`%Pjh7q%-DwIKc&`UJ(2Fo%gZPmVVyMvuzqz}7Q&vDyVhPnyrSS=I
zfr^F2b{iH94WA)%**|gYA4=_mr}=>1@@~KQ|B)vMyI4TAK6Zz`p={y5g6BO||NUUx
z-{b4B+oty9aQ@F|W2h&a#&5uFK7YY?Uk0jLVGMrbT%{iL
zklc4)39=-BX&OTsmfHIZc59>syn>NXsbp5|;j1z23i%_5o9{k85ZKqH^*c8Q?0oU%
z@tCcSw?X*tR8e08eG*pteV;u;1{he_@ctku
z;Q<)gW`sU<-X9$150Z*Yya51L%$l<;{k(KFQRtaC1Y29geA46)qpDRn;cs-jTFlh5
zul!N9o^i9CU`*TKbDGmYt~GLS2|z4|_L)!mpxCJ8f@eXHeMs5ce~9j(Qm4V68@`Gt
zfPC_EA7kO;?XU0pX>FY=zkRx_$QnJpyYB{Br1GwBbW^+EOD-shJdr^rCc9-f5V(wX
zK0sHSu!gE8HnL&iO>T=KZXI8IScKhS*_fcF<%Np&B%-|uWe6KX2Ry8&4BGtdLVW9~
zM>A)|PmY-v{J~0LMMf#(Sp3?u-_ZwYDfjFHk7nz1LW$8y%;*%qfF!w%w$_?UE~
zGCOV~iO1@xwv_xwW)TT6t4Z1q)AV6;$3J9}alAnjF4f{3b+Ytcu5_KB=!Q9s>f8)GLU($!+N4I98L#MgPl$py!q6F&^pFORM%7QB1_$^F)$OXp
zeYH$TEGOXYbi)p)s0kp(Ke>h$PnUt>W?Azoj2il{v(w6rl5Q$?RiJd}@ZO(zI%Xm>q~
z$4OnI=m5RoK9n1lwcOA(ixn)pvHzrU1AP)Q9U6{EhWue*Mqarx~pO+WjK}kVcI=u|l
z*y~8VY1?%w=5QvoFd{h~d-R{Y5J7rzZXSbeh`NF@y8
zp1R8-Y>wZHJzaK8{eg*{@5a#pn&p{|c!p7};A+tn;I5bYr(~)g^)DjD#K73SY_GSy
zf7nOD?-h|Wb$;h7s3oj?FScsHBdu+wTZk}JT+TZ5EN=vC>7u8eo#4-
zQ~+1k$St%|fH{?9p){}$bYgsa8Oab?Vkts>mnP6pd$_HF9Gt(lgzpec0cJE=
zIewoBwr-cEliKH%&n#=ga9qOf>w&9n`^6MAmjpgd_DGq&^QM2MTGIY9)+z%vlN}L4YQkAF*<^>&^e!Y+hmPd0@0`)X+Z#qHghYPkHYDDx
z!{4=crI8^-JbS^Xp7(*_`?d-G=|FvNsi?z4a`6-tS9&FU6xJUehwA9)ShFEgUWJ>wM_KGD^?8YQGMx1EpNu$iJ^>SA)k
z45fXQ2~4?_?~K_beyvTwyRozPFLJv<-)(C(KKY&frV`}A?p&pJiA1dh`EPm4Mkf?8}h@JP9*6zOf5%H)<8w%iQPegpUrax$+%tzXa`sT4ei`l95
zE2{`SuSLdo3yx4?`(!o9MYe)ReNQO(A32G=&F(QC;(@gY=pVk+({)B7i>dOdlQ0)g
z=HP=eN-wH*>TuoDuKb)mGWZl}v!mC+iSZo9YxFnFe{+}S&si9goi
z-`-aYA-%4lZUXXN35h@PqYZ1wi!n$!?U4vx3*Fbxsi*OMMhh6aq4|I^sy+Q(kB;!R
zX8ZY_IDe%I>12y>A6!24s6G%(
z(s~mMj?JB?-r8_2Z_rBcM3f*uew&f4Y*O<>a3v;m-OB~j=MUX_wXncm>j{sKkS84a
z?cw13iyv0C;(KWQ$om3MlxZdwY1u|pL0a?fpeq_HM7#?uubkXNeV1UVn#eM0x4#?4K6$JuZ{T=uT2JRnOx4j(G}fHud`?E
z)%K6ABmw7F5`>HQFf*^P^SncxU?l&ZpS##E0P(;E$!@$E|sc3&+d4m$QHPss(BX+*^
zN7b*nhA2}U0P-Ab%0l}6
zsoBjBc|Ppn69T#wDK(cd+8`Lbbn2YZS``OsfQg9z-#{u&H1wmIVQ|h?^m!j2=8t3T
zO?&6!c%9u*J54lvh_+ra=QosQZudjWl@@-9<6fUFYb
z5iB|ad4Dbi>Bf>UH5+BCk-O;V%NqbQf7v$2thUTsI_nzNUqYJTv*~b@s^AkjgB-%-C`bwB
z6pn6)E~&2AnwkF1q9KT&n=9PF4JV)}0Qg?KKg`mV@gax3NjERSf@vHJmpzGnHx%->
z27S9A$JlXaB7^o~`s0@&&RYwO+}k2i{LTlf(Bd1bVi(OWna*$s}e}EGN
zWBTsDIdr_7jAx!lR;$!%1OudiS`>!3Mk>1%j6*|YGgOCN)X0*)z_yg^MO_f?m
z&J$i-yfrD{9r`k9CmA&cZgp{D5gqz6^2y(bk-UJ)n=pbT8$IXLwE^7@f2MW3+
zdq{YA&eg#tVm}w&1db)m>#lC8@7HFO7T9pgT@r|F#=xPI9#W4*!t2%e$1aT*e*Gs%
zpMc9zK?!>AQ?DtYnqElbOa0svp*l%#xmkZtlH)7k*u_dO4dcc@`<=lEsq+#1nGwZD
zbaKk0IgE7m^S5q){e*V>6Rs)j4kv1SULl};ayR=#hu0+>D+9LkyhE^MGP(2ii@MkH
z-3PoWopWjS1z{eT&O@z-b0Zc|M_py}4oOdVn0EPdmGAvTNsHuRaS*eBHZ*wo^j
zKwz2uYy6p?*NYq&;-bE8<>!_F2%Rm3jrAp7egc&Bwk8%0a`MDJN3TU`wn$v_!xuB>
z#3V6x_CDp_^e>lN&EA=ePt9+H`KqwKhj{_e`CZ$;wM3kvE@*ePc0&9>#G5weqe9D~
zjpn7J{U)~YLG_NDAretzHkDlZ(C1ZrUiW8GQRyT~!D$@9kLKlm*%rtGiyk%felH_N
ziuyGqqgitRa>m`{!A_Z^1v?g%=M(d&dmsIZbKRaGV%z!bzPJ;Nf3Nq=qSHHWq_7{j
z+R^)QhXU~aExS@aJBe%eD~qI-?4nU}^t#nfXN(<@)j7w}yPaqd{chD9ky+YPk_XGn
z`r5v&nTgW-Jwkn*Jdn9_b?N9w?}h3+bc1voykrF92lT3=)R5C(-Zq$&fbiA%zc2j!
z{t@QJOHF47$Pw{mF0RLLx#Tp`e`NGO*`~VRTV6z^f%H4}=z(k^_XaP28q?=x!uK(F
zOSEG*{O0)_arK)oNyNUh+wE`4am%-W=nIm!^3vs_umMpAOaRt|NAXF{>At0Ut;d09
ze+mcqGqg+ip2pzOfePQ=DQr0=`Sv^rgTvZ}#wlrtIpE|l^)})cZ^!Y_MV~%aE*oEZ
zschi#t+pdMz47CuJ&x|BuN!XTfEM-Hp~cwaA>qov{7Jvz5kZfqAqOXSb}Wm7;9%zi
z1DVGr3~EjT?2jS;Q1?~xfaG1fvfwyDu}ObGYTEy74fODYf4I_yGAjLoa}w()_Pr)0
z!N9jH%FzK=7jXvti9QS8`d)_M53BSiMq-XnwcYs_W)Qhi+@tw)?tTkcsk^x@A?%_cDEDZR5L9T$|fKYZA`9nXezFJDmudjEv)1DkfIq#!jVfYeh
z!P8=