You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@accumulo.apache.org by kt...@apache.org on 2020/02/26 22:24:50 UTC
[accumulo-website] branch master updated: MASC blog post (#221)
This is an automated email from the ASF dual-hosted git repository.
kturner pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/accumulo-website.git
The following commit(s) were added to refs/heads/master by this push:
new 02c9e04 MASC blog post (#221)
02c9e04 is described below
commit 02c9e047f135e89e650478cd6f6d059ce97c5f7e
Author: Scott Graham <57...@users.noreply.github.com>
AuthorDate: Wed Feb 26 17:24:41 2020 -0500
MASC blog post (#221)
Co-authored-by: Markus Cozowicz <ma...@microsoft.com>
---
_posts/blog/2020-02-26-accumulo-spark-connector.md | 192 ++++++
css/accumulo.scss | 9 +
images/blog/202002_masc/architecture.png | Bin 0 -> 86251 bytes
images/blog/202002_masc/architecture.svg | 713 +++++++++++++++++++++
images/blog/202002_masc/runtime.png | Bin 0 -> 37732 bytes
.../blog/202002_masc/sparkml_vs_mleap_accumulo.png | Bin 0 -> 20964 bytes
6 files changed, 914 insertions(+)
diff --git a/_posts/blog/2020-02-26-accumulo-spark-connector.md b/_posts/blog/2020-02-26-accumulo-spark-connector.md
new file mode 100644
index 0000000..bb25558
--- /dev/null
+++ b/_posts/blog/2020-02-26-accumulo-spark-connector.md
@@ -0,0 +1,192 @@
+---
+title: Microsoft MASC, an Apache Spark connector for Apache Accumulo
+author: Markus Cozowicz, Scott Graham
+---
+
+# Overview
+[MASC](https://github.com/microsoft/masc) provides an Apache Spark native connector for Apache Accumulo to integrate the rich Spark machine learning eco-system with the scalable and secure data storage capabilities of Accumulo.
+
+## Major Features
+- Simplified Spark DataFrame read/write to Accumulo using DataSource v2 API
+- Speedup of 2-5x over existing approaches for pulling key-value data into DataFrame format
+- Scala and Python support without overhead for moving between languages
+- Process streaming data from Accumulo without loading it all into Spark memory
+- Push down filtering with a flexible expression language ([JUEL](http://juel.sourceforge.net/)): user can define logical operators and comparisons to reduce the amount of data returned from Accumulo
+- Column pruning based on selected fields transparently reduces the amount of data returned from Accumulo
+- Server side inference: ML model inference can run on the Accumulo nodes using MLeap to increase the scalability of AI solutions as well as keeping data in Accumulo
+
+## Use-cases
+MASC is advantageous in many use-cases, below we list a few.
+
+**Scenario 1**: A data analyst needs to execute model inference on large amount of data in Accumulo.<br>
+**Benefit**: Instead of transferring all the data to a large Spark cluster to score using a Spark model, the connector exports and runs the model on the Accumulo cluster. This reduces the need for a large Spark cluster as well as the amount of data transferred between systems, and can improve inference speeds (>2x speedups observed).
+
+**Scenario 2**: A data scientist needs to train a Spark model on a large amount of data in Accumulo.<br>
+**Benefit**: Instead of pulling all the data into a large Spark cluster and restructuring the format to use Spark ML Lib tools, the connector streams data into Spark as a DataFrame reducing time to train and Spark cluster size / memory requirements.
+
+**Scenario 3**: A data analyst needs to perform ad hoc analysis on large amounts of data stored in Accumulo.<br>
+**Benefit**: Instead of pulling all the data into a large Spark cluster, the connector prunes rows and columns using pushdown filtering with a flexible expression language.
+
+# Architecture
+The Accumulo-Spark connector is composed of two components:
+
+- Accumulo server-side iterator performs
+ - column pruning
+ - row-based filtering
+ - [MLeap](https://github.com/combust/mleap) ML model inference and
+ - row assembly using [Apache AVRO](https://avro.apache.org/)
+- Spark DataSource V2
+ - determines the number of Spark tasks based on available Accumulo table splits
+ - translates Spark filter conditions into a [JUEL](http://juel.sourceforge.net/) expression
+ - configures the Accumulo iterator
+ - deserializes the AVRO payload
+
+
+
+# Usage
+More detailed documentation on installation and use is available in the
+[Connector documentation](https://github.com/microsoft/masc/blob/master/connector/README.md)
+
+## Dependencies
+- Java 8
+- Spark 2.4.3+
+- Accumulo 2.0.0+
+
+JARs available on Maven Central Repository:
+- [ Spark DataSource](https://search.maven.org/search?q=g:%22com.microsoft.masc%22%20AND%20a:%22microsoft-accumulo-spark-datasource%22)
+
+- [ Accumulo Iterator](https://search.maven.org/search?q=g:%22com.microsoft.masc%22%20AND%20a:%22microsoft-accumulo-spark-iterator%22) - Backend for Spark DataSource
+
+## Example use
+```python
+from configparser import ConfigParser
+from pyspark.sql import types as T
+
+def get_properties(properties_file):
+ """Read Accumulo client properties file"""
+ config = ConfigParser()
+ with open(properties_file) as stream:
+ config.read_string("[top]\n" + stream.read())
+ return dict(config['top'])
+
+properties = get_properties('/opt/muchos/install/accumulo-2.0.0/conf/accumulo-client.properties')
+properties['table'] = 'demo_table' # Define Accumulo table where data will be written
+properties['rowkey'] = 'id' # Identify column to use as the key for Accumulo rows
+
+# define the schema
+schema = T.StructType([
+ T.StructField("sentiment", T.IntegerType(), True),
+ T.StructField("date", T.StringType(), True),
+ T.StructField("query_string", T.StringType(), True),
+ T.StructField("user", T.StringType(), True),
+ T.StructField("text", T.StringType(), True)
+])
+
+# Read from Accumulo
+df = (spark
+ .read
+ .format("com.microsoft.accumulo")
+ .options(**options) # define Accumulo properties
+ .schema(schema)) # define schema for data retrieval
+
+# Write to Accumulo
+properties['table'] = 'output_table'
+
+(df
+ .write
+ .format("com.microsoft.accumulo")
+ .options(**options)
+ .save())
+```
+
+See the [demo notebook](https://github.com/microsoft/masc/blob/master/connector/examples/AccumuloSparkConnector.ipynb) for more examples.
+
+# Computational Performance of AI Scenario
+## Setup
+The benchmark setup used a 1,000-node Accumulo 2.0.0 Cluster (16,000 cores) running and a 256-node Spark 2.4.3 cluster (4,096 cores). All nodes used [Azure D16s_v3](https://docs.microsoft.com/en-us/azure/virtual-machines/windows/sizes-general) (16 cores) virtual machines. [Fluo-muchos](https://github.com/apache/fluo-muchos) was used to handle Accumulo and Spark cluster deployments and configuration.
+
+In all experiments we use the same base dataset which is a collection of Twitter user tweets with labeled sentiment value. This dataset is known as the Sentiment140 dataset ([Go, Bhayani, & Huang, 2009](http://www-nlp.stanford.edu/courses/cs224n/2009/fp/3.pdf)). The training data consist of 1.6M samples of tweets, where each tweet has columns indicating the sentiment label, user, timestamp, query term, and text. The text is limited to 140 characters and the overall uncompressed size of t [...]
+
+| sentiment | id | date | query_string | user | text |
+| --- | --- | --- | --- | --- | --- |
+|0|1467810369|Mon Apr 06 22:19:...| NO_QUERY|_TheSpecialOne_|@switchfoot http:...|
+|0|1467810672|Mon Apr 06 22:19:...| NO_QUERY| scotthamilton|is upset that he ...|
+|0|1467810917|Mon Apr 06 22:19:...| NO_QUERY| mattycus|@Kenichan I dived...|
+
+To evaluate different table sizes and the impact of splitting the following procedure was used to generate the Accumulo tables:
+
+- Prefix id with split keys (e.g. 0000, 0001, ..., 1024)
+- Create Accumulo table and configure splits
+- Upload prefixed data to Accumulo using Spark and the MASC writer
+- Duplicate data using custom Accumulo server-side iterator
+- Validate data partitioning
+
+A common machine learning scenario was evaluated using a sentiment model trained using [SparkML](https://spark.apache.org/docs/latest/ml-guide.html).
+To train the classification model, we generated feature vectors from the text of tweets (text column). We used a feature engineering pipeline (a.k.a. featurizer) that breaks the text into tokens, splitting on whitespaces and discarding any capitalization and non-alphabetical characters. The pipeline consisted of
+
+- Regex Tokenizer
+- Hashing Transformer
+- Logistic Regression
+
+See the [benchmark notebook (Scala)](https://github.com/microsoft/masc/blob/master/connector/examples/AccumuloSparkConnectorBenchmark.ipynb) for more details.
+
+## Results
+The first set of experiments evaluated data transfer efficiency and ML model inference performance. The chart below shows
+
+- Accumulo table split size (1GB, 8GB, 32GB, 64GB)
+- Total table size (1TB, 10TB, 100TB, 1PB)
+- Operations
+ - Count: plain count of the data
+ - Inference: Accumulo server-side inference using MLeap
+ - Transfer: Filtering results for 30% data transfer
+- Time is reported in minutes
+
+Remarks
+- Time is log-scale
+- Inference was run with and without data transfer to isolate server-side performance.
+- The smaller each Accumulo table split is, the more splits we have and thus higher parallelization.
+
+{:.blog-img-center}
+
+The second set of experiments highlights the computational performance improvement of using the server-side inference approach compared to running inference on the Spark cluster.
+
+{:.blog-img-center}
+
+# Learnings
+- Accumulo MLeap Server-side inference vs Spark ML results in a 2x improvement
+- Multi-threading in Spark jobs can be used to fully utilize Accumulo servers
+ - Useful if less Spark cluster has less cores than Accumulo
+ - e.g. 8 threads * 2,048 Spark executor = 16,384 Accumulo threads
+- Unbalanced Accumulo table splits can introduce performance bottlenecks
+
+# Useful links
+- [Complete Jupyter demo notebook (PySpark)](https://github.com/microsoft/masc/blob/master/connector/examples/AccumuloSparkConnector.ipynb) for usage of the Accumulo-Spark connector
+- [Complete Jupyter benchmark notebook (Scala)](https://github.com/microsoft/masc/blob/master/connector/examples/AccumuloSparkConnectorBenchmark.ipynb) for usage of the Accumulo-Spark connector
+- GitHub Repository [Microsoft's contributions for Spark with Apache Accumulo](https://github.com/microsoft/masc)
+- [MLeap](https://github.com/combust/mleap) - Scala/Java stand-alone model inference for SparkML-based models
+- [SparkML](https://spark.apache.org/docs/latest/ml-guide.html) - Spark machine learning library
+- MASC Maven artifacts
+ - [Accumulo Iterator - Backend for Spark DataSource](https://mvnrepository.com/artifact/com.microsoft.masc/microsoft-accumulo-spark-iterator)
+ - [Spark DataSource](https://mvnrepository.com/artifact/com.microsoft.masc/microsoft-accumulo-spark-datasource)
+
+# License
+This work is publicly available under the Apache License 2.0 on GitHub under [Microsoft's contributions for Apache Spark with Apache Accumulo](https://github.com/microsoft/masc).
+
+# Contributions
+Feedback, questions, and contributions are welcome!
+
+Thanks to contributions from members on the Azure Global Customer Engineering and Azure Government teams.
+
+- [Anupam Sharma](https://github.com/AnupamMicrosoft)
+- [Arvind Shyamsundar](https://github.com/arvindshmicrosoft)
+- [Billie Rinaldi](https://github.com/billierinaldi)
+- [Chenhui Hu](https://github.com/chenhuims)
+- [Jun-Ki Min](https://github.com/loomlike)
+- [Marc Parisi](https://github.com/phrocker)
+- [Markus Cozowicz](https://github.com/eisber)
+- Pavandeep Kalra
+- [Robert Alexander](https://github.com/roalexan)
+- [Scott Graham](https://github.com/gramhagen)
+- [Tao Wu](https://github.com/wutaomsft)
+
+Special thanks to [Anca Sarb](https://github.com/ancasarb) for promptly assisting with [MLeap performance issues](https://github.com/combust/mleap/issues/633).
diff --git a/css/accumulo.scss b/css/accumulo.scss
index 72f3a2b..17609cb 100644
--- a/css/accumulo.scss
+++ b/css/accumulo.scss
@@ -240,3 +240,12 @@ a.prop {
overflow: hidden!important;
}
}
+
+.blog-img-center {
+ max-width: 95%;
+ margin-top: 10px;
+ margin-bottom: 10px;
+ margin-left: auto;
+ margin-right: auto;
+ display: block;
+}
diff --git a/images/blog/202002_masc/architecture.png b/images/blog/202002_masc/architecture.png
new file mode 100644
index 0000000..89718e2
Binary files /dev/null and b/images/blog/202002_masc/architecture.png differ
diff --git a/images/blog/202002_masc/architecture.svg b/images/blog/202002_masc/architecture.svg
new file mode 100644
index 0000000..80c347f
--- /dev/null
+++ b/images/blog/202002_masc/architecture.svg
@@ -0,0 +1,713 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<!-- Generated by Microsoft Visio, SVG Export architecture.svg Page-1 -->
+<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:ev="http://www.w3.org/2001/xml-events"
+ xmlns:v="http://schemas.microsoft.com/visio/2003/SVGExtensions/" width="5.69012in" height="8.85433in"
+ viewBox="0 0 409.689 637.512" xml:space="preserve" color-interpolation-filters="sRGB" class="st12">
+ <v:documentProperties v:langID="1033" v:viewMarkup="false"/>
+
+ <style type="text/css">
+ <![CDATA[
+ .st1 {fill:#ffffff;stroke:#4672c4;stroke-width:0.75}
+ .st2 {fill:none;stroke:none;stroke-width:0.25}
+ .st3 {fill:#ffffff;stroke:#70ad47;stroke-width:0.75}
+ .st4 {fill:#70ad47;font-family:Calibri;font-size:1.00001em}
+ .st5 {fill:#ffffff;stroke:#ed7d31;stroke-width:0.75}
+ .st6 {fill:#ed7d31;font-family:Calibri;font-size:1.00001em}
+ .st7 {fill:#4672c4;font-family:Calibri;font-size:1.00001em}
+ .st8 {font-size:1em}
+ .st9 {marker-end:url(#mrkr4-35);stroke:#4672c4;stroke-linecap:round;stroke-linejoin:round;stroke-width:1}
+ .st10 {fill:#4672c4;fill-opacity:1;stroke:#4672c4;stroke-opacity:1;stroke-width:0.28409090909091}
+ .st11 {fill:none;stroke:#b5c6e7;stroke-dasharray:5.25,3.75;stroke-width:0.75}
+ .st12 {fill:none;fill-rule:evenodd;font-size:12px;overflow:visible;stroke-linecap:square;stroke-miterlimit:3}
+ ]]>
+ </style>
+
+ <defs id="Markers">
+ <g id="lend4">
+ <path d="M 2 1 L 0 0 L 2 -1 L 2 1 " style="stroke:none"/>
+ </g>
+ <marker id="mrkr4-35" class="st10" v:arrowType="4" v:arrowSize="2" v:setback="7.04" refX="-7.04" orient="auto"
+ markerUnits="strokeWidth" overflow="visible">
+ <use xlink:href="#lend4" transform="scale(-3.52,-3.52) "/>
+ </marker>
+ </defs>
+ <g v:mID="0" v:index="1" v:groupContext="foregroundPage">
+ <title>Page-1</title>
+ <v:pageProperties v:drawingScale="0.0393701" v:pageScale="0.0393701" v:drawingUnits="24" v:shadowOffsetX="8.50394"
+ v:shadowOffsetY="-8.50394"/>
+ <v:layer v:name="Connector" v:index="0"/>
+ <g id="shape1004-1" v:mID="1004" v:groupContext="shape" transform="translate(0.75,-106.441)">
+ <title>Sheet.1004</title>
+ <rect x="0" y="129.543" width="130.394" height="507.968" class="st1"/>
+ </g>
+ <g id="shape1000-3" v:mID="1000" v:groupContext="shape" transform="translate(278.545,-106.441)">
+ <title>Sheet.1000</title>
+ <rect x="0" y="129.543" width="130.394" height="507.968" class="st1"/>
+ </g>
+ <g id="shape1-5" v:mID="1" v:groupContext="shape" transform="translate(327.129,-584.822)">
+ <title>Sheet.1</title>
+ <rect v:rectContext="foreign" x="0" y="616.927" width="38.7" height="20.5851" class="st2"/>
+ <image x="0" y="616.927" width="38.7" height="20.5851" preserveAspectRatio="none" xlink:href="data:image/png;base64,
+ iVBORw0KGgoAAAANSUhEUgAAAXgAAADICAYAAADiIPpnAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAAEnQAABJ0
+ Ad5mH3gAAHdoSURBVHhe7b0HuGRFtTYM/vfj4179vOpFmdM9IF4wYgYlzuneu8/MMCheE2YMqMOcc2Ykmq6io+AVc04XMZJECSKCAUSS
+ oiCCEkRgThpghjwMceL/vqvW2l3dp7vP3ru7z+kz1Ps869kd9l5VtarqrVVhV20VEBAQ0GvYvHyrx0H+Rb9Owuatttqa//OqPwUEBAQE
+ 9DourCP2m95TnDu6pPCSFUvmvHxseM6uNy17yhP1LyH60w7c6v/TrwEBAQEBvQjxyj2yHhsqHDA2WPjB6GDhany+fWy4cDc+j40PFX8/
+ Ptx3zNiyOc/TW7e6sNzc2w8ICAgImEHokMvj+HlscPv/BKGfCDJ/9J7Dipvvhtx1aFXuOWzu5jveV9yMe8ZHhwpHWaNQ7/kHBAQEBMww
+ fHKfGJ67B4j7b/cdPnfzqmVC4hvHhorrQOTr4ME/KtehwvrxoYKQ/WreAy//bh22CSQfEBAQ0EMwUh4ZKr4YxH3j/UfMFWIXMh8sbMJn
+ fq+Xjfjv0ZVLC5vZGOC5k1cdtf3jqafV5GxAQEBAwDTByPjmJds/DSR98Roh9yK9dHjuDYm9Torr6M3f60j+a6IU8MfyAwICAgKmGTYs
+ Q4Cov8zxdvXYN0wm8paynp787cuKm0cHC8OqskZ/QEBAQMA0woZmRofn7g/v+0GOuesYeyMSbyl87k438XoXx/GpN3jxAQEBATMAG5q5
+ 5qDtHz8+VDx3zeFzc5O7J+vceHzx3InD5/4r9YflkwEBAQHTDPOuxwaLB4KYN962VIZnsg7N1IuM26+GJx+GagICAgJmADY0c+Xirf7P
+ 2HDhNDex2rb3brL+zkNlqGZkYmjuLgwnDNUEBAQETBNseGZssG+30aHCan1pab1H0vllsLBpfKiw4d7DZDXOtyVAIHjxAQEBAV3G5s3V
+ zcFAxsuU3G31TC1Z5xQS/K1L3cTrxFDffgyLBB82JgsICAjoImx4hoQ7Plw4iZOrIOXOeO9VkaWW3OIA14v/cfB2/49hhk3JAgICAroI
+ G56ZWDK3CA/7Gg6ldGD1TCOpTrgOFY6SsLklgteDCAgICAjoIJLVM0OFfUG8a/hyEj63u3pmkkA3rxtI8LiuHDvE7TwZJlwDAgICugDx
+ oHUcfGy47yCSO7cZAAGn3JYgo3BcH+KGavp+KJEAwoRrQEBAQIfhEyuI/ZM6Rk5y79gEawPZIGvshwrrR5f0vYZhBy8+ICAgoMPwt/IF
+ wZ/sbSzmE3KnxU24cm38YN+fRg7b6UkMP5B8QEBAQAdhE6xjgzs+eXSocBm3FejSBGuNcNnkxLBNuBaPlrh4w0UBAQEBAW3CvObRob7n
+ gnhHuTkYyDfLEkl643mHdNbrZmS383xXiY/Xo5gtsJVAoYEKCAjoKdgY/OiSYgyiXceXkXBNO8G6aRxeuDzjXorKSvIMZyO3JObwkEQI
+ 6OUJVyFxxI/r93kVUWIXIdF7/+tjAQEBAdMLf/35xFDh3dwaGESb+g1WuXeo8BBkDUke3/OsvFmvyzLXjw0X38C4+PMCvQAhcRI2JeOa
+ fWsE9GtAQEDA9MCf1BwbLnyWB2eDpLn+vSXB63r2jTx/Fde/jA0V347rrbLFQb7dJ9czbOi9/MbFhe0YH5DijJG8eOGwDbc1bkbO9NDH
+ l+5QGB+a87LxpYVXjQ71vXV0sPCWieG5+08Mzn2hTRwbAskHBARMK5ItCkhWQ4UzUu8g6Tz8jW7zsMI51DE+VPwIv4sXn7IHYMJGhUM9
+ crjIYPGjEicd7uDnbkPConcOe/iNXj1I6OhlVMaH+/4b6T0dDdKNiPtaiBw47tIjK5Duwf9/xvXTK0D2+ngg+YCAgOmDkdnI0FPngIyu
+ 4hp4kpUjqpYiQzHcAhj3f4M6Vhz6tO3x21XqiWdehcNntEcwYROuVy7e7f/w2g2QbIXQKU0aEvYmRoeLe6HBWoZ4nQi5nvFkutmYscfC
+ 9fwcnuKKIAo/c8iJadEG7w70jt6nKgPJBwQETA9so6+RoeKLQUJ36xh8miEWWeJIIpsYKiwVZQCefTPETbqm01MV5/XrlsLeG64tPOos
+ oJfOHgv3u2+m86ZlT3kiGxd43+8cH+r7lnrhD5DIuXyUxK5pk7iqrWSiuIHwvw18hktB6fVrMAEBAQHdha364GeOIXOIZCVEyalKvI3F
+ 3kR9eHxp30JRCJBAxwcLp7SxG6VNuK4bXdL3Oupsh+DZgLEXQC9df6oBjxCU5aFLiweCrD83Ply8ENc72dCR0Nkb0XTa3IIjdF0xBG/e
+ j3u9yJYMkEdXQ5/0aoaLFQ06ICAgoHvwhwrgjb+fXqoQVrrxc9kwDI3CxMQhc59JHTct2+X/8rpiyZyXQ89qXd8uQzVTEGGNkAhJrojH
+ pTbhSq+b16mwXIddlNQnDYWwARob3P4/QbSvRFgfh5wzOlgYY2/kXoTJcCVdLi4kczZSaQm9scizRU1T3480KgEBAQHdg+8Zg+SOV1Il
+ qU1F8PxfV9AU/7zqqO0fTx2+l4z/l+t4OiVNj8AXuZ8NzshQ4UOqsuHYdTI5CkJvdpD36OKn93GN/+hQ8UjoPRVyA2Q95xvY02A8SfA6
+ 3MIGaYMROqRR/DKJNgobtAG9gfMdGrWAgICA7sAI/rqhpz4BBHS+TgjSY21JbLhHPFol8BNFGUCyNU+bpIr/rqBXTI/cf76VmIfMZ1T/
+ BJccUqfpJtHLOLprUCZNjnLLhZHhuXuMDBaXIJ7HQ8eVkDUkWBK6DbtIOlzcnJfeIUJvIht1fuNOO5M2ICAgoGtIJliXzNkJBLSCBIgr
+ vdh6cqoX8bCVgJeLMnrSuhLFGg6uC+e9SqZZx+N5hut69ipA9idQH3Hdgbtuox8TjLxzp23Hl+zw/LHBvreNDRe/ND5U/D2eWc3JUD7P
+ Rkb3n6es1wYnTU+lY4K0bJAloIjX+OIddtaoBwQEBHQF8jo9P4wtKewDEnpYJxOnHE4hWXEyFtfNEyBV6qAum7A1gudvuP9UEiyueQh1
+ w63aOEwMuy2FCYbDcfSR4cJ/oRE5dmyo75e4dxSykcMuJHWuXOGwC3WMDRcexXW9DrvUhzEdwnA3cE4CBH/NzUu2f5omJSAgIKDz8Nd9
+ w+t9x0p4u1xFAyJKRfC60uW+kcHintRhjYXB1q9PLJv7cty3ioSLa+qhmkQGE2K8AbIU3vmRo0PF00HaHEd/mL0ONiDcdtgaKHroFH5O
+ 9MykuIZFl39Wh7QCAgICugKfkEcHi5/iZl8gn1RvoJLgZThnsHAzx9qpo57g/QYE93+S99Pj57ONdLYSPLORDQp7DSRxEiXjyyEPvccN
+ u3CCeOa89FYiDY00imhM1SwBAQEB3YGNvxPw3E9JPYziCNT2oPktx7+po57gCfwmq1pWvOdpfMP1Sm1E8qyNlzA5pq5DQxvUQ7fJ0UbP
+ 9JKsd/bq+3tYQRMQENB1GMFPvHvuU7jUkUMo6l23JHh60yR5kvXoYOGbogxoRPCE7XXDTbioP+eEqwnjl2csfyZFGiDt8SwTowQEBAR0
+ E0bIK4eLz4I3vDrDFgVCWPIy0GDxCFEG8AUj/VgDn/jHhgunJee99uZQSjfEveA0XLiA2yDQDv77BwEBAQEdhbwcpGPkI0v6FnK4Q/dX
+ STPcsVEmZOGFjw8WXiX6oMtW0DSCDdWML+3bHc+v1uWYeb34bgtt0JFhH9pV03oPVyqpLZLVRgEBAQEdh+9tg4SWcvIPhE0impLY6OXr
+ 5OZdY8vmPI86mg3PGKRBUVLD821NuHZcXE+Cm4ZxGaXER5dXtivrbc5gbGnfYjEEMJWtAgICAtpC7bBJ8Ut8sxOElMZzlclOWbY4WLja
+ O5hjStKyYYmbls15KnRcpZO0MzGeLmmAsAchk7QkYTZaHD5ivNj45BlCQmNpn9lgiC789jkxABCGZgICAroOI2Tu3wIi+wXXkIOUph4X
+ x//0uqVBGC7+zCZq03qldv/4cOFN0LdJh4Wmw4s3Qpc3WLnen4ROAiap65p+Hjt4LdL1HZDy7/Q32iMV0XvkLg2lrko61Vtl1HCfnICA
+ gICOwgj5dudNX6/jxHxByEiqoYDchbzU4/+MKAPSErx/H3T9pIsTrtTnL6VMDuBg3HXbgkcQ7j9xz5ljg30fnhics4hkPLp0+2dwD3h5
+ MUuHbFKLS8d6JfeLeQAK05p2J8yAgICAtmETrCu9Qz5ASFMNz1A2cuiBnvfocOE9ogxotoKmEWyY4tZlhZdAn+y7jmsnvHjGP3mDlXEk
+ SZNsdThoI7cFhvf+K77Yxf3vuQePREohe9oMFc6TVS9OX9aGZx2Xj0LHjWPDc3alzuC5BwQETBtIxtctdxt2cX8XkJKQIUgpDcnKvbg+
+ Mj7c12/6sqwK8RuDkaHCct3CQHTrNZ244aL13GeGVw67sCdib7lyohRkvxr3Xox7vjrBjc9AuhyW0uANEvfNsAlPkVJyp2SKDxsWfZfg
+ LsRtQHSiMQsrZgICAroKWcUCT5JDBf4brKNDxY9kIVg2AnoYxjg3+6IOX19a2DO3LS5sB71/EQ976n3oZfxfPXTKJg67cMiFpKw9gfsg
+ V+KeE0aGi4PcMviWxU/+dwnUA+3BF7DsgBICut3qHq56yd6jWKcrkdaPDRXfTn1q89Q9m4CAgIDUMBJrNP5LD54HYPDVeW94ZqrhCBnX
+ 1nHzi+7Wl3byELyQnw4TjQ0WD6R+WU44uZHhd9veV3oPDJ+ErsMuD0Ouh5w6Olx4/8qlxfiOJlsBmC2MdH3yhR3eC3K2zcpkzB5h+vFo
+ KiR1vhfgCL74EVUZVswEBAR0FiRNElmjcV96shxW4YTiqNte957bHLmTuNOMNfNs0Q0kV5Da91VtbiKzhoFkPzpYOEV3WpQNw0joJE6S
+ Lr1qhqkTl5SRseHCL3Bdzhet2JNo1MjI8AiP7eMVYejPAt+75uQqwr/fTbwW2ZBYOGlEPH3X2BS/LcoB254hICAgoC0YqTcilduOLGw3
+ MtS3n0wqDhcvBHneTSIjYQqhZVm9wnshbpzZ81TbGIawxmF0SeElIPO76J3Tk2f8KLpUcbUehP15yJs5jn7b4sK/iQIPHNunh+576Y0g
+ 5K7hcpIZjcmo9gaykjt7Fxu1YTqHp2JRZ6MeU0BAQEAqkKBkYpOeqRKVj4klc4vw0F8L0vkKNxHDda1MPoIwZYzZkRPJjEMR6QkeZMaJ
+ S3nLc7B4IMMSsqzzjrOgdpik8AHGD9cHJd5cjz5YeM/40JyXjRy205P0thqwp9LMS28GawhpJ4ajSz6z2UInePXZv9yyuLAjdQZyDwgI
+ yAwhUiP1Bt4phyl4XB1I57sgrb+DgB6lV0rvUtd8U0jqMpat37OK7MmO6/23HNK3O8Nt5Smnha2queag7R8PUn8lpNLsxCMOxfAgESF2
+ HcPPgoTcD5/7r7DRz5WgufQzq010+9/ChNkiDMsEBASkghA6CIwESlJrRGYcqhgb6jsEJHMqyOpmepUkHQ5z6MQphGPKMq6cl9R92SCk
+ Nli4buWy4lzGoVEPIg+akbU2bDWTo3nhxxWE/g0laPZosq6YWa/2fQByAPUxbu30ZAICArZwJF66iv6cgAQ1wheThvoOBbGcDWJayfFq
+ EjrJSleAcJiBww2UTpC6L+vXHM4DsPt+acsLO0XwBD15acwgQuj83CHS9O2JdHyIPRGufEFjlYncYXPZQIzDVCNLioOqsiM9mYCAgC0M
+ RupCbA1Iwr1ZOedlIJcPgIx+NTpUuINETlLneLW8dOQmSoXUcww1pBWGISto+NKQRq+jBN8tSE9IewjyshN6NOqB02b16WwlsoGYzGMM
+ Fj8lyoHZYIOAgIBphHmqjYYmuBpjbGlhn9Gh4tGjg30XgljuJSFxQzCOp9NrJ5HTmxQP1BF8I0LqnLgw5MWi8eGq5+q/ldqLkAZUCXhi
+ 6dwy0nCne7Er83LI5BSrscHij0xnGHcPCAgQGNk08tTHBnd88sqhYjQ61HcsvPTLQCgP0FPk0AtJVfcnp3fOIYXpO1jahbOB2wGwYZHh
+ iaG+/RhnpqNTQyjdAuIoBDyytPAcpOEG70zY9PZTG8g6/OHCBTzi0NcdEBDwGAeJvd7b5Wv6JEsQ+udAIFdAHiGhe+PpFCP1aTviLukd
+ uB6CbOnLHsRDR8mKk9vQCD2X8TcvtldhSxZpZ/SGfq/r1TnMknoYC3mDa3GdrLYZLlw3ccjcZ1Jn8NwDAgKcl+sRIYkBpLE3iPM4XP8K
+ EtlAr5KkLuPpzrMkwU4rqUNIeo7UEQc2MIwTx/q197ASJHnu6JK+15nX3mzlSy/AvGsuqYSdT3ZDK862kPq0t5J1tAOu3LRsX+qU4bUe
+ 77kEBAR0GT6x05uUNd6DxZ+BvNfSmyRxePuvOEJ3Mp2kbssokxON5PAPkuFw4WZuISBLMQ9xR/MZepngfO8aafisTIrSzhlXzEDcZOxg
+ 4VHk3RuoT4bZenzeISAgoMvwyX30kOJeo0N9PwVhbPRev6+SuvMsp5nURWTSlsMP4qWSyIYK18LjPWFsSd9BtlukD+mR9DC507vWj1uN
+ DPUdygbU30BsKnFDMiKyHJLPjg8VD1eVYcVMQMBjHUYCJEOQw5Egi3vWHDHXhmC4lJHEvskjk25LDanTo+WEoW4TzJd1roCn+1UOv9gL
+ TD6YHorOIfTysMzjbNhodGnf65Cuh9iAwc6S7qnEyw9ZDkn7oGH+oigHwrh7QMBjHEbucvTbYOGbJFMKSUaJvYZUuigJqRtZsffAuOC3
+ e/HbRbh+mjsprmqwLQDJjNLL4+w+ZOhEbb9yuLgX0nYbh5rSkrsntJttIPYTO0817DETEPAYhxEMMTZU/DIJVcbYh2Xoo55IuiFG6us5
+ McphF755qi/1rJaj6oaKR8t68MEdn6xRTcCJydlE6j4Yd17HF++wMxqva3Rr4WzkPug2ENNnLxnRfeUDuQcEPMbhL4EcHS4czHF2Dsnk
+ 8CCziK0Kkb3TSeqcJCWp67jzBMI/c3SweMTIYHHPVUdt/3iNYgLZEoCk3sNDL1PByJ07TiLNv2bDSnvo3Ea9zRqKDs+so/1w/ef4ku2f
+ 7+sOCAh4DCMZHlhWnAuCuF6W5XXHczdS56ThBvYQGBZJ3a3I6bsFDcxJY0v7Fq9YVnyRfySdYutkN8ZZTOoGGxdnA4seyvdkvXrO5ZAy
+ JzFYuBsSzlMNCAhwAFkm3js3oNJxbjkco45Ecot6mLLqhj0DEpnsC8PfhwvXIazjuV3wisHCs/34ECQpkroMv2xBhOUPicFjb+c8Vdsd
+ Es+F81QDAgI8GNGQFEASP1HilReGID6R5BM3NiykztU4+I0EdhXH+SeG+16zYrjv6RIRDxxH3xJJ3eCTLxq2xbRPluWQJhzKsfNUR4f6
+ wnmqjyGwAPnSiUpCHfV6felURWwUTicrOXVJGMuXL0+E3/X3rKiP71Sovz9rmPXPt5IpYWTA4/FAGlfom48k4fYJXskdV+4HcymuH+P2
+ BuNLdyhI4B44TMEx9S2V1A3iXWv6RgeLr4B91uqhJpnIHSInVLn86vuWKAds2Cdgy8UMVY6EKNtFt+K/tUfmzcCwt1hyaQR7uebmocIO
+ II3rdYiGB0f7ZJJLQF4bZLJ2sHD/6HDh/fWeJcmI4+wyrv4Ysbt57/DcdxsfLoxpg5qN3Dl8BtF3Ac62M16hO5D7Fo5eqCTtkHxXyV0/
+ Gx63xx6Lnrjvvvs+eZ999vl/+pvhMUPyRrpcVgdS/6t58J0geBV7KWoN5A9c6siDPyRwDyQ+2SdlFi5zTAt/aIbbKHByGTaRuQmIb7OW
+ woaTXj+uN908NHcX6gvkvuVDCs+8efOeGpfj46JS5YSoFH0D1x+Wy/ES/GUVJ3UFOvDAA6XyUyd0fYK6nM7oO1E5+rZco/ib+P0TURTt
+ f+CuB24jD2YnecZJ4gU9eyD+x0O+Rf2I+/Hl8oCsDAAyV/56YoeuPRHGx6D/zKgcXwn910flyhX4/lPI0jiOi3prq7Dkv0ql8h945hix
+ dTn6Lj4fVy6XZf0x4Ieb6IpL8UG49we491vyTH8c619EszCT36H/jXj2+7Q/4i42kryQ/Kh8B/IN1f953LsTn7F8bAQj+OsO3HWbseHC
+ L7iOGsTRuTF41cOJQO+M1VWQEzkGX38gtQxhkOg9MtxSUEPwQ4UTtDHNuikb792oz15sPbAt0V4BVXgEUDly/sDCzfMHFmyuxAOb91u4
+ /2YQwVWo7Nu5/8upW3q7F4S4TxxVHly4YD/ROVCZL1f7TAFZbgKpnDmw74DsAZJluMYIiOGBbM/Zb+GizQOx08vPaESSU3iAZiQ4CT6x
+ IQ27wQ6nIJ5r/fiLRANiLwoI82h9hOE0DMv0gpz3hr5HFsxfuJm2wecRNob8z0+/fXbpi8+X9FUWSPqQX2+Sm1qEByS6kIazmadm95p0
+ QJBPLi5RPFbZt/IsPrPbbrs1fdnFJwYQxnLdc7zTO0FSFz1VeY3eNgbT36+AV/+RiaVzX6DRSCAePSVDnvc4ktUto8Pc46ewRvf2oW18
+ e00lG3RS9oER299eiT5gC4QRDsh1RxDITaz4uG6ISvGj+nmiv39AKlCryl4PIyZ4nW+OyxUhD+rllaRIkd9K8UZeSXQgll/bkEdKkmfl
+ lftAdq+H/iQcyHqSLhuOXXfdVXoHrbxRH/59iP97oesu0eV0b5I0KCkyTIp8jipH8BmNe0uCR6PxFj5numCHn8sNDknaraFk/oCg/8Y8
+ 0XjcW5lXeSH/a2UrC096UuX4cmmIyvE6C7teFs4HwZfiCxCuvKo+lc2MdFYOFvcEaazNSTppRLxPiBz6wRUg9EQ5lgziv2N8uHDS6FDf
+ a+1QCoN49Zx83QK81NoGtfhtbeiyNai4F/baqI3x2TastSXYJ2AyEhICeRyjhL4Jsp4E70grfhAkJyeoG9mkQFJY4NV+2Gs0NpJYTIxU
+ GBavDA9hvYvPtSJJgxHbggULHo94XqThkLySBgoEf5UNnaRpoHxCQ9zfz3iJXo2jhkG5F2HezfCM/NnI8DnV0Sju/E1+p7dvDYQ2Dp/l
+ 74rkWWucYPs9ce+dYiNns2sqlcr2/K8VCVuace8L8fyExv8RhlsnzHfXE4ni/5WHHVrmge/9jQ31/fB+t5yxm2+xuolCR/Yy/kyvnksp
+ 6eGPDReuGIdXv3JZ8UX13jsnZmf7ShsbVllxSPFFsMOd1XXsdTZqJYPOi+dwGhtF6gte/BYIIwZ46M8FYdyqhCNeNis8r/JZPVMgVSvv
+ Ec7WIIvvCWmop47v3wQRvjLqj14LvR9DOHdpeBsckVZO0GeJVhWR/0l8oO/dTofE14UD4lViXl0pVV7G+1IQfJI+N96t5FuOHzbbcJiD
+ 4UHnPA4/MR1oRDh8sxLfK3y2BeEm6YnK0YmV2HnjFOg8RP+qsbHpQjj0+NfzXk3Xzy09aQhebK7pcPap/JW9CKTzNfjvjbgeiO9vwPXN
+ 1mNLCyOH0SWFl4BAVskY7/TsQyNED6Li8M0G2bIAYcu+7kOFO+nVw9N9/c11m4rRW9WtClKV516CH+fRocLncg2LOS9+g/YAzre3foMX
+ 3xhbs4LRu6WwQvnC37QC9prxErIBaX3N80zpydHDW8/P6tF9U28lpvR+jFQWLVr0RBDRRTosQHmoXB6Q02EMIMdvadib3LVylv5FNLWZ
+ kZpOVl6lpAdyl16CxF9JeR0ITMaqWxEhYf8jz56D50Y13g9p/FaBAA+WGxuAk7noKdiLNw1tZD0O6N8WNv8D9Wpj9BCIdZH9t+uuB25D
+ G3LiGTaUyof7P67psTxJtnNtlS4LEz2GwzUsnS+ofEduaI4p89ngEwMPyaB3SAEBddeT96Xq1csQzmp4tnzxSk9luhJe/dFcXljvqc7G
+ 9fPmxd8yXHwW0jZuy1MhtTZpIfTedSx+88hg8R3Ud2E5rKaRCmPkPRVh1MOeJenj64wWqIQg4d2CFO9OPLtSfDfI4Dsggnv4nWQAcjn3
+ gAMOkPWyadKs6SP5Pgt6bqZuJacRDhXwPw6r8ApyPk4JFPfgGsUn83dFMxvxdyXLypGq28K4Gmk4id9NJ+Jvb+rxuYY6/XSRPGVOAA2F
+ 6rwP4bxa/5YGjGk00Z8NTfPVwujv739muRytqNoFn9FQyU1NgDhJTwh5spHPwG5c3WRo1hAmcUEYX1c7b5LnS/Gg/jUJDdI0JWqGaoYL
+ x9GLl33g3WEaQiS65LGrImE4sqdXLxuRcfiGni6+3wE5eWRJ4U2ji5/ep9GddahpUAcLH6MnjnRlX1Ez6I5NxOfLbln85H+nvsfoUI0j
+ 5lbkhkqxLT1WVlROaOH7dvzMScMmFWZr6jQPa5qRhBlFlR+a9+s81sqp7KKDCCZI8PwPBHk90iDL5hhnebAFzE4csoCehyDOayzH58sN
+ Cuici9+uVhIV4kFYx+jfU5KxPB/FN+rzQpYkLoT7WtOp3moy7NPM3p7OXWCLMX1+g+gsx8vlJqBJ+h/XqmwYrBzQ24fuB6hbwinF1wzA
+ 5pxInd8/H+Rf3gW9gZ2RjmczjciPlyAO4vEznYwXbStKXV62tBOu2+D5X2geiF3YQKCc9rnGpryLXSFPkIeb6GwFexOSJDE+3PcFjg+L
+ d+lIPtN67U6INigMl0s3NzIu3PIAZMjfr0RD9InxoTkyfEfMJi/eiPi2Qws7Ip036tLHrC89yctk3KwNOoZ8vY8JsII0qtDz5i3s41AD
+ Ksk7QUjL4R19G5X0dFx/i+tlkD+hEl0OuRSE+Sv8fiLuPQ73vgcVaM+BgQFpLQ0axrQVLsRBKqIS8APWdRfBb/h/W3wW4lUCutuGVlIQ
+ PNMhaQHZvkuer5L3H+AJv40EjOsh0HuBhm3j/g8jnDKfbUGYiX5E9ZOmX8nvH3vttde/Qv8++Mzvm+dXpNG6kC8l8Zkm8U9sj3i91eIk
+ OtGAkPj4nw2X5IWlCXZ5L8MQKVU4pPQA5CaUkX/yWiOl+EZcRyAcdhI7Ik63Ik7P8XU2gqW1v3/BDtBzjRK8DL3BJuOqm+HcIN+R5yyf
+ fMbKSFYYyRN8AxVkutbbHOzRjF5mJ0X2PMdV9mDhmn2ZoBwu3I3ffo142fsSAtp1hpyvVKgZix8sHqFvpubZ6G29evF/WfGep8mkPXTn
+ yvtZA2ZsfQGX4YYoOhhk/V1UhCshd0GEQFhx0ogQRyleDaL7HVdRQN9uql6Qt1JlgV9oEY8zlBhlSSEbKP7Owo3/znW/OY8P399p//Ha
+ DP7/8Jw/pekmiXFsn8sXucplHW2hYfPzRjckEv1AHyUaVi6zEa7Pwf3jPsEjjjIUA+/3pbRzEkYU3wJbP5v/NSL4mjhHlc9qumV1DMI4
+ Uf9q6v2nhNeIxJ/TeFNIuNKQMq62hp/C4SX7LGXH4hTFly3ca6EsB2yVH5ZWpH0PPHuH6nkUtpFhnmo4A7L+PY7i1bh3Vz7TTln0SX58
+ ad9CkM6l9J5JJPSqIRybn3aP3hMS/UOM0wNHovEZ7PsT4rQ340viXF6b9iTfeg1mZ75JjPRcpUSddd5DJqe5vBUN4FHURxvM5pVGrbC1
+ X7BZeVDgF6JC/W9crtC7Ssjaq3RGXKyoJKtHcf+jqCxcVvcof5P/dRUJn+PzOt58Kye74nnxyzVIq1hdM66lD17kaxAnWfrHdeqI4/1c
+ GSI3AYjXZxzBmSdb+aT+RTSNn+l3totP07Fs9Rq9pZGO3DaQ0OQet/ZazsVsQaRJuPB2v6x5IGQF0vzDbru5eQLx4svRH708up/5yP/8
+ /DXUEDwIXePsyLQUfUb/ItrJlyRNiPuZZlvawcpTM9H8kTImcYvik1RVS4K3tKJX8iamRxsVXF0ZNP38ri+F/cl6Ke16r35X/7bFhe1G
+ BwvvB5lcT0+T3rOb/CwK0ZP0PcLppsC7La5j2Pqq/+1jw32HTRw+918Zz8Ve469Dq/ZmcU/CP2wF6TqEPRIdfsrUeLLB1RU1196mO3Ru
+ cac5uQJdLdQghP1BGmegQqw1L13JIvFC9TOJS8hrCtlEksfVnpXKpRV9FSr9sXupVwZ0avOtGhgZoOByFcdvNGy32qQU34I4LMV/78B/
+ 78T305UQ1LuvnGhxSkMqbouC+E/yrLeG3BcNdzXC/RKHvfhcq3RXCau8O8hotcWPV8T5XOTZ2xl/GYcvx9ey8cB1U8yXqnRSUvXXEHUN
+ wVcbJbe+vRx/Xv8ichO8hbFw4cKnQPeVfBuVjRPjD7maDRTtVSMyzFf5I66rpOyh/NBuaHQ+IUqBFnnBuEp8ofsjaiuTCaTrD05/In+H
+ 3k/xfkU7jZlAPEGP6EeXbv8MPYz7Cq7g4NANx4B16ERWvxjpdEG4nHI9x6rVWz0DvYvdNWpbLT/QvW9AoBztg4b+PNYD/SmxZa/BhlNu
+ GC7+B3tKtKn2kBrZoJnQNm4riMHCx0QxsMXs7WPEQSBzd0Ul/xEK/EOs4KwYuvTOJ3S/suQTI3uttFJxy5VL+Aq7RqUlkeaBR5DvQNiy
+ zh1X864tffyN3+2zDVX80V6lbzUOb3Hmahmk5zZnPwnnDtj1f9kzQMX5GnoE/8OxaNj7RfIgMEWjlvzHSVNnr6TRZFrYiMhn+U1/51Xi
+ X/siUU04dQT/vQUDQvCPaLp/ZvHyywkglX6KOCfwbP9i6F5ZtUu0ojKv8jyuUqJ9OTHPRmDvvRc8jV4kvm8PMj7T0stnOE8gSh2aVMJq
+ vBDGd6Wx0t4OiPxw9nIYloWJeG1nK6U6jfoxXY71wuN8O0jlPJIuSUnfgE2WOeLaGbJ3umQ1DcPhLow8nKR64tOB3LvG8uYJsPWHYKc1
+ kjdRJC8B1eV7T4FDKebJozdyECdMKbBrpmWT4sVzHH+wcPPI0oLM72xe3LyezxbUDMnIxF85GjdiR4Uyb70zpN5YjFydNy9j9NUK3CmS
+ t3RyshFhXObIy41/KxkK4Zjwu/72KL1NTsJxpQd1tCD4hPCi/mh/1eFWspQrv2pFIFOl08KslCoRdN6n+ePyBvH34+7Fn7aVHggalTNg
+ A3n9vr7C+iRNb5dkynRrGKtsCA1x3Ib3UhrEt6W3442HvwqNiPRoqB/h/UZuaALaDHG4VPOL8hA9TP7XymaWRk7o45nz3WSz5OdGPF+S
+ m6YRtme7fhVwq9rRwTmLxgaLPwLJ3MsNxXRNN0mZBNXOvjbyMg+usrkWJ1bx+cSxQ+Y8T4Pf6lzvWD8pV6X4IuaL7g20EuX9JfyvlUPT
+ C7AG6h8Hb/f/kEY5mzWrFw9bia31JbXjRDHAXph+nHWQF5T4gaSHlvt4Zq5Wbuddd5fY64WNyXr11NbBu10msQSmIr80MBJDOpcynRSG
+ R5IhedgGVr7oPTZ2S2J5FXXUE6SHpDAwHJlnMA+6VPmG/lUD6poqfd7/fDP2NCU76WEwni7+tXGnaPzVE6/8df4+8+WwiAYVduvl6vEi
+ jVxZJJPAuGp+ROfRk5Y768AJ3ZK+KatoSPRJwxdFR5j9qRsELwcuMI0c/6Y9KNYY4v7dkOZbaUvmA+z6T9uUrZXdvC0OdkE6ZI8hl6bK
+ Ctu6oQm62i0nYZCQ6ifxuFxxdLjvC/SwOU5PosFnkg+9+qxEz/tlCaCcVTpUuIk9hiRM2M2Ii8OisOlxsNH9Vq7EVt7SYLNlL8OGwiaW
+ zn0d0us2FONhKY3t01DYKOhyy5V8K5n6uN8+r7MNCbmzAqCS/Y6Zq5XaebWTCXg6RLx5Ryrojnsk73uZWWGEhrTOFaJjQS7pXjPl6BIU
+ 5uWQY2XYhIU9qnwSNvk04nCTkrt7OaYcHyYKXUWZRAQ+4YCUvqo2dcMCUfQB/k7youi9qcjE837/C3Fx2wZorwPxPxFx/TjjrvH/tHx3
+ q53W8h5paOCJ4/k9qIfh8+rDfhOPuRT/WsfhH3bPiw0uR8+KWxS8jAJd9MS/it9XIrz3iJImdgGS3xCvb2v+qgdfOVL/IpL7zJYoAwfg
+ 3odYNpWAzjfyb5QOg9kMZB7jGenxaPm+CfF9J+K/fxQNvA7hvxqfX+s+u+WgQKp8aQcc35U3SOs8xImhubuAYD4EsrmGJMXJP31hqurV
+ NyAnFbvHeaJ8ZrBwPMf+VX0NYSHtr0R+XMH80Dym0/Cws3N1aW0rO/cKbEUNh57gjZ9JL14nsett1Eqk18OeFLz4ZBdWe3N2tiAhd+7G
+ hwy9WisORZas9YCs0wlCevIHSaxdpctT8ZJnULE/aAVZK/sDqP/JmH89cP/xahtX6EvRV/Svhg2OVQS+ao+Kc44b2pH0PIzK5G/ElRqm
+ k94tKt05Gp+HZMihFF9mb8TWgx437rtWyNQN4WyELQ/kf03iQDtJmmCTBYyzEvEjfJ6fxWZug7E78dujVm54P58D+PykPDJbMVzY9Fwj
+ eMh6NhRyU23+JnGBzd9neSZp9jYCa2VL+w/6D5Y4uvxm+ZaN2PDdVnjJi2j8HXn0Nj4znYRGr5reZz3R375szlNHhwsHg3R+B5F9U3Sc
+ niJ7z+hnExm/Z6OgXujfJpb2vU7VbeUvfdxn/vwCHIGvoTzYjqlmGzpYGyRfo/hks0PWMjtTYM+IVx7nh/Q/pBuRZXv5CY2CDpPdxR1C
+ fb2zAQm5czwZmXmtVlK23NLl7yExT/5OVFIZM81T8cyT6++f/0xU7Bt1LPZhtzSx4cZeCUGhEnxUiU3IjORkhNooLmZb/Mc3VK8lMSk5
+ 3Y40yJr/BsMjLVHVWXkr9NjEsPQoKkpIPux+WSpZii9w6WRFlvh/UG5ySNJp8Bst3LsU4ZjXTJHywfA1DjoMFK+xuYlmvSyzFV84gk39
+ LX/vwX/P53/+szZcRCAeX5I4WCNTjmWtsqJheICfh8dqGmi7hnMt2lg+2E456wToKdZ7iyPv3GnbseHiK+FRngbSWUPyJgHp8A1JXl5e
+ AuHbJmOPQL5yx9BTkyWOF3rpQRrfgLRe5/Ig8dqt7nuT8tW9foBmdu4p2DANG8vxwcIpeQ5gQc9JnpEJ6aHCd0Ux4K+G6llYwR0YGPhP
+ VJyrtOBza9yZGpJpJTJcwzgyriRNxj0LQfqkweGEhQsWCeHq8MMD0CktNHWSGEksvJqdQA7v5L18RtZJl6NxeKs72zO8+jByrZQqezH+
+ jPt+CBOfr7VlkFnib/Fwbw7HVzEOJCWJSym+CP/La/XuNChJ6+PYe+Bv/Iz4nsgDLuwZ6DiFv/NP012PWpKvvBU2+CfTT0IQQc+KV/6m
+ 5efvaPRkt8JmBG9pjvvjfsT7UdpUG54bbWWS2Y6wuDF90K+HfLhwQVCyskPDmtRIEaaL4aLXcxZt4KeBn32xA0eQtzI278dlJkCCItHX
+ j9PrQRffALlPcEkfyZ4eOwnfeZ3FP48Oz91fb9/qO95wDBrXZ9ChQTqdHV39IrFX59mUB2iTKKrIkCLQ1M5tgDoblpV2YRPZaBQrIOz7
+ 9BSsrBOu69X7XzuydK68Wd7zBG+VRlaRRPGvtXJO2qO8lwTEJCtcJK6l+CSreM3IqR4J4VYq/4HC/SmQ1U/ZxUel/xmI4n1GSA2ISQo0
+ 96QBKX4fz/6QXVY2EkbwTUhAnoPuF+HeH4Fgf4B4/4TL8nC/EG8zEmwEuxfp3V3iAcKG3u/heioakflyk4NfAZPPHBaS8N2ReD9FPN5v
+ L/K0IjE/jv39/fS6D4OOs9jIIF+uxferoPeXuKLnX07WUgMNicDCgl12cyQj8wangESO8Bq8JEy7H3F9IsOQNDh7fhvhyTh5Kzva87JP
+ fin6MMJC2iXcHzipqMjxfMhfyaOjvXLVMB3TDY7Tk+wp+pPg1sHCs0E+Hx8b6vs7PXd8vgfy6ephH8v99fdboxxwWfDNrEfaa+EQZaN6
+ b0OXG2xotFU5yYNW+cZxdJmXaMP+vq1gk+/r2615ViOt58tgaFBPNp02zt9z8DMJhf1bSu7WgvsZ3IsihY5eBeKeDDFkIUqg1b09UZmn
+ QJa0dgoMs8Y2XHIIEpwDqTkLNCVmg517EjJOr169/iQYX7pDAV7760d0rJjwJ1HnVSrPQ8N4EuuPeO3OmWtV523u6z40xrKZW5YeZwpI
+ OSYfLdjb9fqIekLXhu1f8u5ZbzZYiR4PiPrOO3N48RCZz+DKmuRoP8SlncanW0gqKryUxcxsbaV7ZUI1jdh4/D0l3Tc8I8E3QyCd1rA5
+ m2a2TuZ0AqYHQn4Nhm+IC5e7HsgunJAvRUPopcg7LV59rw7HNBKdq8HnCRC87MnTqSWSfjlBr+koxM9/a1gmMscGd3xyo3RJekHazdJd
+ D38Lg9HBwjd10pkefKYtDCAyFj82XPj55uXODj23Z7x1O2NuQBXFt5IokeFcQTA5gzMIdNAb4Ak73LzJViXwe3eGfEqxHFyNz9dy2MBP
+ W0qwYNRLGuR5hsj7XCPU6zKZCnmeaQQ+l7zgpI1r4jhkhB+XNM9nvb8e9c+3klkDEp2M03NIA9fTlEA5j8AhJ5srQX3JMseWzHnZfESH
+ PPikrKDheB04gpO4F9hvBMl9fKhwgkwmDxcPGh0uvPTOD2wn5xP7YLrp2U/l3dtY/MhQ8cUg6lvdbpOZl01u5Fux/Dwx3Pca6pMGtlfK
+ ihEgrtx75QydXJS3RtsQdvFEB4dN2OKzUHAlgkzOuHv4f2eGf1zhtBegINHXkZ7tmK4OefEBAbMdCeGgnh/Gg8NJotzwr6YuTS321vO5
+ qGNuAr/9HlrSy+OwD4h9tThqpfiaXat7T8lqofHBvgs3f3hHTnKSVG+H/HZ0sPip8cHCq29avMPO3rxCgmbevf8Z+j7PYRrVm9mL52oc
+ XM/nm8fU1ygeM4GE/GDYxSRfb1e++oxNI+zeyZmY8wfcyhJ8XoPfJiDcrxvdungNf7f/eT+kdbewtcjzOmewEl072/wokHtAgMInYdTB
+ r+owS1ZHTubkZEinHH9f1QFt1bNkKxRw0G7QP6IjCAzvjnnz3OlmBm4RoPu6byYhc5LTTZKK530DCPrk8aHi4ePDff03LZsjq698mHcv
+ vRodx+fvK93RfiOqO+O6+MJG7uPDl80mhvpkCxWGQ/38PGMww+phB9cpSebz3ulFczWLKzhoIKLzuAIC3bgFFW6qFUXP5ktT/B6VosOj
+ cuUc3G8vwuSZzOUz8rq8eO5uK91ktUYHvIqAgC0GVtdZL0DOP9XN4rLOsbHO2bkCnxbFDrmJzIZ3ED+eEuYOXHHDuQxvfVmPgbT6zE3Q
+ ZLknV7wMFx7lBCcJlpOdJHqOh+vyRa4a+sPoYOHrIP+3Tyyd+wL2AKjDB1e9XKe7ZKJ3cAyXTPL9ATQUqbcw4Lp4xMcO6P5DTxzt53u3
+ INz36/CGZCDEz9Q0IuRMrwCF5yq+Ddlq0yxCTxV6FYj+Eq/FTkfyunOlxlleuLBXpoFUR8MFBDyWYESqG6tdyqFY1LmswzOyWk1Etwhp
+ p5fskfsckLodPE8OQMPjVsWBI47gPbZef2LpnEUkdRI6yNSGUtz2CzwNC943h1nYCHDYRJdAbgLRj40O9f0Sn5fzPYCVy4ryzowPPdpv
+ hZ7fmtmLZ7jyJvFg32JVOXNevLXoXIpEUlbjZm3RbfxbvGjoOZO9AQnAQbpfzEg7BV/DTRKt68+/o8M51DkVyfN/G5KZ8N/UrNcdEBDg
+ YPUdPeidUW/+kXOIxp2LwK0poki2ODC9WWHkzu2eQeZnsT6j/le3sEZ44sDpBnzmDdMT52ZrbtVL0wlREv868+5vRWPAN3hJ+PqG7wP4
+ /W94/se4ZymXkN697ClPpH6SM4dbdOI0tRcPkUaG8UJjcvVtRxZkDrB+2ep0gSQoRBj1R6+lMYVclawzSNJl45DLHnssEiMZqVsYDSDE
+ 7xcOeuFaeCiNSJ5hufF9KZzRb/H8i/XxTs3kBwRskbBeLYiZRxPepXUtXY+5KslSZL6JTX15lkhaXXUcEH9fnTWKxUe2jGBY+P83iHxC
+ kiuHi/8BUr5cTptyB5U3IltfhHiV7MW75zAMPXsKt0jGf3fgv8vGhov/MzLUdzg+r/L23q/X10rEi3enRhUP1yjLqhr9OD3whzBgQHup
+ STxjNXBacStXSvEYCo6sidW3IFMnCHGRAsI48W1Qa8mlNbdwXMNjq2Q2RVHls4u0MQHCOuuAgClgQymop69gncrp0BnB31TZt/Is6svq
+ WNU4daXoC9SnCzvqRw/cMZjl+IaXzFuUTJZyUhQe+CluJ8jMwygUkLZ4/urdu8labpnMCVZ43zz0PE3D0UikMdGJ2mtXH5oc0D29iz2M
+ EGHs7WDAK3ViNNvwjDc0Q+9bFDtiz9xaGcm73Q0rf9QGx+Ijrbr8FsVj5XLlTfIQoIVleo3XGpp+d+BFFtFnH8sQ2zWyTUtxKzge67ab
+ Col94NC9x+spZ3bopB6W4stsfyCfsKdCDblHlf/2hmUbcc8Gjeed5fJ8Wzwh6Rgf7jtGx9bpYZNUfZJNL25rAg7DrOOYu42hu7Nwc+uV
+ FTXUw+Efxnfa18WzYvCK1pzLku4XQ2Zvze2c0o3cz4T62hkmqXbbBgZA5LKUEmHIvuauJa/8CgVEdhUkshSsLkMmdbXRbIds+Jz/ktD0
+ FYiZQX16pUzmxJZqO6Yjq0yC1XciLsfL1TOW4VWty2nFbRNcrpxlOn3draD1Q4BGZgn1aR1vNg9ge97wzAfZwtomWkGc7yaBcrULCbWO
+ YNsROZNWib/R/1OLe1ZOyBod7LvQjj2cTi8+KQicKKEx83bXmEF4lluotr0nRV0hPI6FUAviOn739jTvhSEZn0yaAvb4N54disZozsDA
+ wI7coZM79uH7XHpAnGDCbQ0rJSDpTFuBZguYHpd/rdPFVVZcGcVe3XzuTR7HRQptyYn5ZnvcK2aj7STOJhp3E6uzU4ndWwPfGYrk7Nl8
+ SyRR1+3kse+oOmJSePXw84FkDV0PTUHuFOm5C8eU4w/xWds/ZsVQYQFIdK0b6841TNNt2WBLNUeG58oBOtO5ZDIxNjzvQ9XQMu7tGTeN
+ JB48Gwrqa4fgCXse5Metb/8G3XdDt2z9SjhimLpAdQtW+fRrDYR4uAVwf3Qw99BAL4SHkp8Pb+evkBX4fiu+3+6kMo5Cez2ulyCNP+Gc
+ AtJ5MI+1g55Jm3R5FX62IiEv/Z6A6WKDh/SXxLMrRZ+BTU5yk+iyM+VNtBeuKyET+H0F/r8G//0ONj0J10+DNN7FbvxstF0zu3QSRvAI
+ ZxvY69xcb6urA8g6Dx0fF8WuLk5VH2l7sb/0zsvxXea4JbobCPLUHwL+Hp83krxtaeE5IM8VHKbRydN6gp1p2cBVOG4pZ98hjHOjrZ27
+ gprWtJ3umpcB0PN5VZkmw1sheVb2BdcDMAgrpDOBRpWQk8nwKF+KOL4PBf4UR9jxA+wN0cuh0DYicf3+4vo7xO7VXtQa6PozCvQXUcAX
+ 1ROW2mBWET3jXE+w9L5R2feEzY5Cmk+H3W6GPNzQdpBGdqPU2w7kfwV0fhmOyyt5jqgGJ1Dbdb+CpUQzYpftj+N4Z5Yt2GhfXGOkZz+U
+ hwPgPLyKV08WofwtpKC+LCCB4vanqypC0mt1B9c5sPOVtBmuWT14WcHGz4jPu6mvUfzrkPS2EfbueHZU5/vSNC7kIzv79WJ4f4nzyC2P
+ xwYLl+omX3knRLspsppGXsjyD+eejvLnVzZ4jv8jme1eHMo6HkexlS0j/f0Dz6XOdr14hW8ImXjTz9OKRpWwv7//mSDhpSCT85Due0k8
+ Rkh6+g1FTr1JxHU3J0v1f7E9dRlx6e884/RIDuto8AKrsL2MRrYDEe0qpF6OLkTa1lpa2QAaeYgtzC7OBo1tVv0ftnMvxSS2w28I58/I
+ pw+SLDV4InktfgYxKQ74vhNs8xbUw29CLoJ92GN5APIIhGQom4HBRugtI812dTYhUcs9tCHSvUTVElJvLB8q8yrPo26ts3zWbJ5G7GB5
+ Hi0pW+NOUdeTdPKkNMTrb5o3jG9arrGVNDft7TVc9OS5hl1X0mRZqz5dIvMCsgXxYDEZzpqW5ZI+WaICHMMWVQpOeqP7wmdkHTwy8KeW
+ 4R2qRLJKAteZ8LoaVcI9o1LlK6hoY0YkWuApNeTj2dPEt5kv8r/cXyUs0cPKKuTnCvjNUSk62n+JTOM3E7aZChxySWxHcoF3CTc0Ph7p
+ WGW28xrDPLZL/m9qO5Rrtd0ICPFjiFPy5uJM2U6JNgk3nhe/HET9RcTxZstvK1cU/pZWkmdKsexoCCR1xwge6d4XYT3C+9Ve9XZtJbZE
+ chVy86XU14rgrQzgil5DdAnThWeZ183ytJHIShoe+cgiJIq3cmmBh/xJ3cO9vUnR7ogQPD14bkUs0QamneBBGh/WTKNkMbwv/mEbx6jq
+ pFC1gRkhL4131UZRtIeS030e4VKqXlWVcHy75BWny1VA8XaMFPHbdeweI472cknSBe4FWKU2wHYVeNenIg1yIDnLiKav3nb1NsgrzW1X
+ jq9FfN6OaEne1udzt+HbBp93Qd37CuJ0T12Zco0dhWlIKyBfIW2U0ZK+gOSVC9YjqUsoO6/nfSp8zsJNI9Zb/7sNA9Xnt8F+57sqqDu/
+ 0LJr+V2vt5VIo4LnuJBjH+o8TfeNGRvqey/fTnVbA8gwDcfi6c3PONmj8eGB5zZE8wnGVzYfmyZOSwo1yYKZTWPmML4v9gozT8BnJSJm
+ bGglL3yPBIV0J3jsnPC7m5NSmj6rgI5IGtuik2LhyB76RpLsLfGoQI1q04o2jajp8eDz87nSAp7Xg2I7562TcKfTdpQa20lZL1V+aARF
+ TEcD6dsGHvZBcbnyT8tLxI82kThqfPPYxsj3ZtS/ZzMcL8ykDqLcHKZhUrKG49bAl+Pf2wqmRrazcHmlrfWZPD0GGXKSdLn3bIQcbSWN
+ 7uF+LbcfINFz/xi+tITfZE8aWV0zc579BsaFJA8P/i2Mr04QTwvBJy06J2a0Zc2TAfViXbg7K6VKRP1aAKYlUW1Cl+45RP3R25GOa10l
+ lDT55OSneTqFYYunpqsgRkEWb9Yo15DIdELtJiTCpY3oxR2GCjlWtx1079iOSwRL8TXwdKWMEt0kecsXLokFwX6Z8VHSE2LXuNXHN6vI
+ IRy4Xs4lpH64ftoQ/ue9sLOEK70ieTaKT1N1RI0D55dBlIMvKR9QGJ6vbyqR+93z0XcTxwsOo+8Frzik+CKQ+AdBpGeAVEdIqPceNlfE
+ thqwLQqmmew36EHe4xNDc+V84Gndk8Yyna09utC3eZ6Eb+Q8YgXtBhS05zGMmSKetKipAPDs4HXIqfKSDtfosRLWp3MmRYhBViMwflHl
+ Y4i6VLTptrUfHvL7hShLZ7EsacWm3XrRdvqyTnwXerCydzfRDduZTq7lB7meyXDZyEC4i2MnG7x10uhH8c/r58H8dCEOp6hzkClf4EiI
+ 88d8ZSOl6oiEbP16xPkilgNxHLOXAeEhl0eVExF/2eLXT0f9UIdsX7Bk++ePDxcOBqGfAGL9G7cc4DJKTsQq2VJA9rJNQSdfjqoX6t4o
+ WwcPFo7XKE7vfjSWGe5E+fgCLfCdqIwstNaVu3TevIV9DKfVZMxMwi80bo1u5a9drISdFE4uPspKJD2MKP5mo4rQTfjhwFF4C+I0ovlO
+ 4aqOnradNkLr0EAOaTJqSKpdmH1Qx56GOvZrzzadbvRoZ9dooRxI4IClxa7cCBDe8IUaj8ykq3WCBH8U9QF0KoS0aub1Ip7zmrylmi0c
+ 15C4nWJL8en6ImDDMi1j2jy0o8EboiuG+57OE55AsJ8F2V4CuZf7w3CDMl7p6eO39R7Zd9K7X6cNyl08UpDxmbG9aAhk2LFqUDFujcFz
+ CAqBTPyozp9bJvUayfuFBjbgq9P3auEXD9nS0+PiHXhSOUE3eus6yXv6Hwdv7ROIh+tROLt1oifYdSHJKwltQho6SvKmg2cioGyJ5y7h
+ dMc21Ot2dI0q/y0RAIx0LS58kxpxuV4btqzxsDm2TRXdC8ovAxAhevSI3oR7ZHsRPqPPphPyT5U3fsk3wJ3Oqcsyx7ftSD79KQF3nxwZ
+ nFPiUA7kFyDe2/iW6Zoj3Ni9bAs8KJOzfCO2LbLncBDH3knwo4NF2ceemPbdJAEE6ApADMDAeZdONZNqN6tU+aFlUreJJy38eKDQH8O4
+ mkcnDVRtWnpdpCIx/kjL1zRZXbO1l5dPgMf4PdeLkO54L/d4GgrPI1UyWucP1+gGZnmRPIv8aGccOpVYeWU6kIZ3aNAkFCEVI3g3hBbf
+ o+nNGhdbA78WPd0y9anDlqziAo1wPu9uq0fes1OKpkF6IeVy/DuUrTnUmacM01sm2fOkJv0pwcThc/+Vk7MjQ32HcB09SPkmbgzGYRwO
+ 5+gkLUmeG49tAPGT7JsSPgidV/6/kRO7HPdnowHdyTDWtHvvBr+i0qjqZWTKmClEChEzHPq7Tjxp4aX7X9Bl/ToLrlVyL+6zTcSTl3SU
+ og9LQoFOeKM+zHb0rswzZbgMvy4+s0bYMAkpgZw6sTjAbB71y7AV9EvZ6maP0DblYiMlLyB5+c40CMHIECQcOLk3uyNnq3RGI90WnARv
+ 5QHk/nLoHPPeUs3S0EvPRj33ywb2HdiROjvBEyRXO39Vf0rA/8aX7bDz2HDxDaODha+DrP8Kol7PbYO5MkcnaSn07rnskVcZX6fodxFr
+ JOT7cOET1pNo1KOYTlQ9jVL8XiW5xCPokMhSJ+oGySdnOHYi8/LAwuUVPYvvsNBa5aiLd6dEus910ui+Tojzskrxo6iE/yUJ9rrP7cJs
+ x1UaUblyjjoElG6R17TZDmVAzgZGGf0b0rmTn94sMGLt7+/fwRsO6Sa5U4x8b0ecZVtdL+5J3qOOv0vLep46jjDEPlfZNsHecOBz8P/f
+ tTwwrVnyiffaFsR/gS5ZcaLx70i5NciWvSB6Ej6+TtI9MvTUOSDtBaNDfceCqC+C3MehFhI3J0z5maTP/WUoHOaht87/SfDjQ8Xf45kD
+ VN2Mk7vACgIzDQb+k2ZSp8lOSJ6CAvIRCRjotHc5Ffzw6Ll75N6xCigVx3lH7L00flnF/73TpGWHmMsLUfmJqh6mQ45XK8U/13JC6ZTt
+ Gr6ROkl8u3bWdtS1QVei/Kh+/DolEtJAL+ozSrrMC8a1PrxOiq1cu5arwBi+zXfVvdR4tMaJac1iO7GNCyP6LVRVHcM4LuK/SzV89iSy
+ pJV6bUEGX0KTnoHavKPkXg9O0HIIh4TfaHz8pmVPeeL40r7dQfhDIO3vQy6DjOP7WlwfhjwIWQW5At7/8aNL+l7HZ/Tx3iB3BRMnCZQX
+ MJSIYfBOjxfaJA0KfJQcZTVdJF87y1/5JOOi6ewEQUkFUNnE02mon5WpkXDVi4btGgQ+1zkS0MrIpXDR1zXJbdnZnuW15sWVDpSRuvST
+ 5MV2zexn/+nzzu6ds12yUgSOyDsl8Q5J2WkFawS5PBi6JhhfXNstX0zjVPKIy+/KhfULGvx8R3n4trwD4PKNz/nhtBIJR9KD/Fd1W3Er
+ Z/RSkrdUM+YDdRq53wRyl00FNb5dJfd6yGocELKQfZPx8rHBHZ/M4ZyVQ8UXrxws7jk22LfbyOLCc25b7M5dNbCxmKkzWJvCCiYLBTLp
+ J9462SyFII0IybNSozEZlMCBdsgnJZJGDAVpMePivVlZH8csQvuI10JSYAXQfX343/2w5S3wBi9DxToPhHE2r/j9cvw+hqvsh81ur0dY
+ WSteM7HG9GHk7QDTrQ1c5orjN4yozMcqaUkYes0njgxcgzjZdmshE5CrEebFsN3vIBfzu/6+lukjOdTYrjNEv07HkW+wseCU5ZO2FVsh
+ rh+vy1P7nEWkbFFYvvR7M3EefBSfbHXZrhZ3rdvW88pWztSukqao8lnVuw3y5seqj7qypNPFmXaO5KhP2YJAy1pDgp1GCEFfWG5O9o2Q
+ NBI56ti0wCsYz4cnsIIZxxUGdRnTCTHy4SEeiZfUTZI33XxzEeHeqyTVzjCUFehNTEtCelF8I9L0/XK5cgj3A+EhFfSoYNNtdUJqW24F
+ S+KQuJSi94MMzsVza6iDurQy5yUFX9bRW4P+Mxp5cymRkBYqId/sNQJuh9xpO3nes90GxPNv9DA5F4Q2iVvkPp3b/dJ+bmvh8hMW4jt/
+ r5Qq8+gg8CUYyHhNHrRvO++NzconmXbCb+gawWxLrxY6Llfiy2cnhM9yYOlKI/stXFTzApLF1+IF+20HG/+Z9yKMbASPdDDfpXyW4nc5
+ ffHXVBclM7nrs7ejXMlBQQDjm5pQpwskeSFuiOznzu8q8t2kV4ndg+eBVN7ITGCG8grxM6gTApKXDH7I1tQS3SB5FGxpuObvM78AMpCX
+ mBg+JEsBr4p6ngmplOIHQdA/h83exvFICTQDmGY2BtDxTeiSJWZKou0RlSMJfl6HeC3Q4Ji/qQui2W6gf+AlIN/btFK2Q+5MU9V25fhO
+ 1O8fopK/igQkgWYEntsFxPYh7u3SMdvheY3fRKXiDpaeqmxaI0rCwnPuCMw8dYf5Vtb3GrhVcCm+Rnp+8L5BqmfXi/2Oe34L8rVdJAnJ
+ Z4s3TxKDvgmNV1b7eD1C2Zd+0LN11vLgyN0d5PMKxg3Yus1lqQFp4BdiLrVjpubMxKnEb8XvR6F5tQY7paeUBb4uVIDvJt1JR9J+fNKK
+ 2EHjzd0QT/c8EB9THxzd4KBo6NqDlZXDR1qh2rE7bSwTY+xVaBCpG1G7T/eW+Y3ajsSQr2HUtKie+7mCCQ2PvOVXB6nsDW1m0sB29Oyp
+ k2F4tssbVz5nW2B/VINoWTbtP/bKWGdyNTTaKLtnQepRtD8asCdpXlD/VDIJFi/ZlhjOiCzZzFr+cb/YtBSvhiNyMn67L2f5TOp8FA3I
+ CXDArNuQcFbDvDYCFeYrmiGUdr2ieqmSPFpzegYaLCpuZzLcSAoV5S2oNBvbHHev7mzHI/j6ozdIIAohHldQKSSfGgJqALunZpMzfgah
+ fgThPdIBorJT6W+3w1j8sFogiT9s506+h1dJfZ7uLCIeoMQlii8AGeu+3g6MUzu28wkC8eVr8g94tquPSxqhvc2Lv9qWBbawXRJfNAjH
+ S6NaHTev191Q9H6xEfL/a9YjaBOJLendQ7/Md2QmeCfS6PF50ZG9LBi5sxfwNsaJCOQ+A/DGbLdhi514vt0keTceV2K4XmXPDWuo5LDm
+ 2pNk/PDTihCUkBy8ROiWt+wIDceIqR1s7Vdq2IJj3jKZyPC9uKSWWtKoDKtqoqVtzXaIw254Xg7owDVPHJi/bo021+aXok/x1X0JBNBw
+ EhJqAzVbFavt8r0ub6LeND97Qx8N42rEz3XhIOdzc5Y1XVFSOcfWl7P+eXUhjdTHjb8JkIZlao+kXOSVHM/bcO8G1MX3aJRaNZgB3YaR
+ Dd9YTLrobPnztf6tREheC/eKUqnyMoarmZ+34ifPIe6fVoKi5Gmg1vM4OVwf8pd3klC64X34RIXP70C4vifvxyuNSKOseXc6VEp8W1Us
+ P02w3Sl8lhU6R6Wukns5vgMNTLINgJatrtoORHIUCZqC8PPkezLZCjt8S9U29DgtXHRMirj3Kl2FkyW/7E3UDaVSLC/KqM685V/gxxX2
+ OM7yEuEwb+rjkFayPmvj9+i9Re/T6ARy7wUYyaOw7YQMsoN6WVnaKSCNRMhA9ScvPPgVNgss3v2cHIziW5Vk8kwW22TwWnhAB4lywPe0
+ u4E6ovooK0gbRGXDNKPMR+ps1TBZ2MiDV+GZdrxgI/eJSqkyX5QDefM0LXz9IOifaBzcS1SN49lMWCaTYRq+vUudjWy3667uhCHY7EU2
+ GY3wUq9A471S9qP4GjYS1NWJMlbXWNuSxm7U32aSvOAYRZUPaFQCufcSrMIM9A+8ABXmRi0krPDdJPnLOWnmh58BSaGGx/2NNiq48zzc
+ a/92QlXXCcpg4fAK0jhHSSO791UdaliPRmqRKHee4STv0CoeyYVDBW0Qgg27wXMfkHX4wLQdLWjk6DbXqtjqn+yNI2yn8zZr0UjJMXgA
+ y1eN7Sw82hfPPKiNYhaHQpa0Iq5nea//t13OjOBlS/ByfL7YwfXAp4PgxbGgwElZLhECpqv+BKSHv1vc3si4lVphOk7yqCAsfDKkAI/j
+ XBSGJzHcLIUiqWxx/NI2vHfpMkvlLkXvF8XAdBfOhKi4jDLvLoDOprqks5oWYJInaunjqiY8k3doKGkYoeeNongayV2RNGBRqZJ/ywC9
+ n2UBZfO91AdMahwtbUjvIXY/rmnziXXItgHoyJvHBiN4NHTbI4wb1A7ZG7rsYrywgfMuEhkgkHvvIlnKhEK8HzLuHo/k/YzthIjHqAX+
+ 2xI6kLJwJBUPDUQy9q6erx9GK/HHXpPlhUa204wkPbRFkp5sRFV9cacUfUfVTRpq8Ahla3iSpzr7u7zwdLUWjRfjCdt9XPXNSJe8Srp8
+ cS9e1UbjuEnG1EvRZ0Sxg2+7hPCR5mNylDnadxPfaIan+yHqUdQ0InlQrbOyGdhduWyQT9ywYCm+lZuuMQ42jBXQu2BhkUIXRdEbkIny
+ qj2unSR5IxP/zbllDJOoJ6V6GAnrwQaycgZ6sr6N6yZ8vc26ZojcBUZUsPk+iJu9QJOlkgrBK+mea0MA9bZMej790ku7Q4kqa89HhtgY
+ jh3KPIO2qzaOUfwjTU/WoTq5X56N4pNVHeHbLvmMRtjetUjfMCJvuDqL5R12kze7NW/aJXg+L3FD2eHLVw8zDG20GselQ2I2duFVPXig
+ pswF9Bh8TwyF8T0sLF0geRMbx72rXB7Yk2FO5Qna/yD3d1KHFLBsZJhsNoVKcTB1daiytQOpFCRK2Pw3YhP1LOvi3kwSggdJ/YWv0lNf
+ HcEn6YPtkv1mrKKmFFstcQ+3EqCuGe6S+71OvpmdpzxUbVeOz0d65EhE33ZW5nbb7YB/8+YtstQH83Yf5Bi+r7NNME8lXzmHxLAk/dNA
+ 8Co2ub8higZk62pN10zWpYCp4Bc+tM6HJ4TYJZJ3k0/RebZ+uhlpePF6HArxSTkqGkWXa0bnmQc6wyRFJBUVBP8RI9+MFXWDW+oZrZg3
+ b56ckeuTlHnZsmV0OfojbYA8zdLz8Ye1viRKgQ4RVW5UCb68E+I31k7vB5+v5D5Cvl6iaruFfbxHy12mCVbVv5LzRtTVoTKXxBH19MM5
+ y007QgfEVlLd0MaiiYDphp9JqNCyc14O72hq0cJI/QhniQZJJIXXYBWtv7//uXjmZl2LnKWiubXICBMeX/2ZkzOJxBOFJ7Y/4vdoDlub
+ N7UK9nkmdfkkZemkB4l78rwFmpAUl6ZS1wwOzXhICP5f6F0zjrAdy1Ta3g9Flkqi/F1vZ4T6tvOWSO6KMMbdeH2GTfpwrzYK1yKec6mr
+ Q7ZL4ohe2dc0jDyryXKL2VrChtOl0Znxhj8gBepIvpvnTtphzn+nl8TwGhUQ+w3xeiufy0GC9rLVJVaRe4OkajzRXRC/23J4okbwd4KA
+ X+DrBJIucxRVPqn5SAJMPwQE/a4SV5JJ3B6pxAnJoYx+2kiOjbgX/6nE1sL/o9FaeCsj8FB5rvF9em8mD14J8ALrIXSi3FkcmQ9I+xmM
+ V47GzSRLeaiXZNgT8VgqkXP5EoZqeh1G8rzCS/i+ViDJVC+D2xUpXCyg8KwPlYAdkkrmVzh0R7+gFSZ9Ya7tKdiJU8nQyEzD0kcCgJ3d
+ lq8Z0ycEH8V3c2dIX6flIYekoNtes89CUDb2/nDUH73K19kDSPKQyxy10U/yO6U4gi/F13G7Yuryy1vVsajQsZDFAXzGe34qkYlp2P7H
+ ZrdONI4Wxz32WPRExP1PmctMY8n7rG1XsZqbnjFePVRGAlrBMor7dqOQ2iHMndzSgIXKvPjLEZ6sjfcrgXk8cpxcOfptDpLyJ3T39XX2
+ AjxC2Ro2PoVxFRunr3A2kTeJ4C2d8E6fh97BCtoZurOMv4sHyoYHeSN79PSQ7Ujuks6a4a30ZXOqMfhEP9J/lDZ0mfVLedWDNBSmPzcs
+ jsiTuQhnVLfbyOV4xZGu9HEbzuUheT5jW2ZcYKdOBZKfJbCM4gsV8IB/pwTLrllnSB6VQComdTZ4G9MjqWfhnpvzELx7Jrpk331fIatM
+ eqvwJYRCIvm+knB2gi/Hd5LIqccIwBpKrnSAnR/VNzfTEoFHUPE3qUfRNkF1CCgfLp0g+D0Q1zXa0OUh+N9zpQx11RG8APnyeSWw1Pli
+ 9cN5t8k+R0m5bgcewe+OMPK9JId06HOrUK9PwbWdpZYSNtMKXcnh+54tA3oZRohxHO+MDLxKSZYTdXla/HrZFJcqtkrjcxIgUE9S3PME
+ 99oOjKlJihVNCT55sQpou5J1CpZONmRI/9mOENITCUQJvnJbf3//MzydCZmAAI9QIqOk1ssK77w7NwneieGFTsJs5yZB9WWfDATPsqGk
+ dLaV8fpyxytse6KW+Swk6ib2eS3Fb/Z1tonkAA2km3sKrbNwNNy0YvMPN/f3z39mVKqcoMd5Uk+eem27Sj7EeDF+veVIBbRCsmVrPE+2
+ CbhFC3wWT7qZsDDZWOhFtoTRIykpzKgk79ICtJkNgvd8K0kmgcyLsgrcK7D4cIgARHJFHoLXivqPBXsveJoohc18MkF+fdUjqLR63dAW
+ PGNOMlJPLw1tEWY77d3dkbXxp7g0Vg+bBkSnlXdcnwQbXJCjvIPwJF/u4RGOvs42Ua0TaHitTkDyEvy11KmnQvlblfj3phXpLaPB/BuK
+ jCyd7LUyE9AcyZ4j4k3rEXTM1LpMzi4gNC2ot6MS7CKhwUvxyRgE/QkWHnpduC8tSdkk4SNoIGTf7w5Vso7B0kjvG/FcmZmkYDuXD5VL
+ bPwTeJylEw3H/wXBn64Elbri8sxeR37x2IAeKNJrr6Sb7eAxPhtxzkTw6iToMr/oC6LQQXRaWnnGLu671t2Xa4nkiB3I0iH7Jb1PEGmy
+ fQLLQU34raX6Fm+58isjYdSRQdrQnY2QfahG6iaec+mOfiCRBHqtzgU0gVYoKWD0qJmZWqnytvhOoEe8bI4DqrcIJL0GhovCbFui5lnD
+ fS9f06euHiOpZB08vDwOQT2kdshE8FqhThWNDskBI/DM/p3kL93vLASl3hiuV/f3L5A9R3rYg38hGqS7MzWOSmAsHyhbHxSFrmyLzoT0
+ 0GOF7swb2yEfleArV3D+ytfZJiR+BBru72m8Mq2B13td7yWK/1fVcQhpG/QKfjpfXkDMXaeTbYRhV3u/hXZNGqaAHobf9Udh+LRWKkrW
+ LmJVqgTP/WmSvdlZ4HiV5WDl+PdCUvAsa55tLUrwldvg5b2IunqN4FXkzeHElko+U4lVVBIJKlMyuQUkBM+3W6HvGiGbbASfbO9sBNVj
+ nlhCxshb7uXj5mdS2s7KnAicFeoBkvywtEL3/rg/z+lbtnfP2R1+czppgKBbtrewclAXfnPxGjfUYdnu10vvi/CfDdXk7Z3b8N6t0Lu7
+ rz9gFsDIQw9tduurXaFJX8h80QInlSiqVE+G2dURvI4P3qCEk2kcVAvqSnp51NUhL6pTqHpjcoCFpE/GhjX+rUXtRpJCPshmVoCQlKUT
+ HaKdcc8/cthu1hA8CPoAxNNNNqpNphTcp/fzTIBXUA9AfULw5siQ/KURyKLb5Z9bOuh5yNbjaAemA3nxJITxdy3f2Zwrr9wgfbJVsqZX
+ 0o7fkiMAIXkdN+0BVn4Vlk7OQhiBlMvzuVTrdi1o+bp1WuBcqx99WAIAPJLaG/dk64I7MYK/A5V4N+rqJQ/eq6w8VWsk83pm2E291jWW
+ Pk+nVCZcuZ1snu0dXAWF99+jQzTSkPED0n5wDhK25aX3oXzJPjEeAVd1V/d6Sd/w6r3yXFT5JPUoRGc7sDhW9pWJ5ZU6kZuZ4MVW5fgR
+ r3FLnAKUmW2jcnSWNvCdqNOfkBAAfwQgoH0kY7xdgFfBKv+TtPjpK1hV3DNSIdBl/Ch1AskYfNkdUpG1AlNsJcMDnBimrh4iqWTSuoag
+ 0ldW8RId+VSusJ0kTadXWXMRPOLiJlnL0TjiJ8cs9tjwFsu1lD/0XvJs1GYviI3BRrJPjNUVv87AA3crkLSM1uloJsnqLXjDg9TTqXpo
+ +Ys4lxEnt/Y/K8Fb2svRPVDzYuqz+OG71Dmen4z7zHHLO1Rjixwegh3kHRcNp+2G7rGOaTllxwoDrs/nOLcWhuwtvlZKeb56MhELghQ6
+ EMxQUoHTVzKKrUVmwyEnvveQByFpY4HnMBfTZys7vPi3Ev9FpK+KRsAqqkfwu+A+O44xkwevlXMtPNwFvs4egaSTgP2ybrjlv+R0OSei
+ qaee5HCFJ1v5WQ5PVogNcXnQttM1nW0iqdfcNI9hafnOTPC2A+n8feYXqM/SDnh2rRwp6XBvuWYNwySZrIcNpCHtkC0ekxCP3ScxneDp
+ VotphWFrVJhfCkllnfChKMGzsIKIJ71Uwx6CnIqTXTfv1b1uqlvdAjPuQVghp2eDONoB2OkrEWymlZteoiwBBZguSZtH8Hyd/e9SybIt
+ 80smITkBTF0eCfQCJC6ME8rMTzOWPdc4okyhXJxheWHpM9tVKpX/gO4/cFUJl4020NNQEA/t/cR3oHGU/Vk61Dh6u49WjshJvtXGrRRf
+ ZuPjft6aPQ7gPvio114Dl61eO+EzdmzhdyUAoIccrVmBScSOTHoSx75QgM9GIZOXDrpg1IQoURE+r4VayMHL4KnFyIrPRdHrqM+vEEjD
+ j93e8bnGA20s+SLz1KwAzxS8fCA5/SJn45W8qIL0yD4xfv6a/Th0Q5LKsUyS4iZavROPeqViGiHRgUG6LlPCykbwjnSSs1LrCR523QUE
+ +k+1QSaC1zJ3IxcIUFen18AjPl/MWd+U4CVNp0GVpNkneCJp5HhmcClerWHlHaqpvnCoh+0oasIMmAwhdj9zeHgGPLr3olLLzoSucLqV
+ KV0ieCl0COO/tVsqGVqXwa0FBU492Ps59kd9iKtUCK3AF+hr1HkKmBtqKMVryuWBAerskDeVF9W3gvOfTlX1iryNrPz8tTC40glp/7kj
+ s8z200nqaAW8WdnnZoZtl8DKPFf4IJ43KwGlsiHsLQ2BPINyKwodRKfZDmTE5Zd2xGGmIRptFC5iL8DX2SaSek5y1nixLKR2DJh2CuOH
+ 8vdlVTeJ4IFq3S5FH85dt6tiu07eCpvIirZecRZ6EvXETmOB2A+Et3sxM0Iz312j+E+20VaHjep5FG1NtMqYIOL+T29feCF4ruBAQZQz
+ WHFfHoJP9qOB/q9RJ9GgQE8LjCB5OAfS9c8c5EGxFSBsEPeivvr0+PkclSrfUPuxcqbvJWg+SuUux0epuqTizwCScC29IM7no3yvVnuk
+ Ix9NF+sJyoQtLyVEv9ku6o9ei/vybBNsSyRPM10dqneSZnk7OXuvRcS2+XDPJvNdxKT6YI0SnSzk/2+0DLUzVKNLb6Pz6HhQd4fssuWA
+ BbveKJwEi8qVc2hEZlxSINEl188syDLJ2CFPwlAl+FJ8upIVMzJLAeC9UiFQiM60tFmXlqcI4f98J+dXJXnxwpbEzcSKECN3VlCm1XaO
+ pDSIcytZzyEr6DgF6qRiNqgoSd6QnDVvKFkrp06SRX+F7YrUNwO2Y1p8AkrGoqXs88zTLCQMUpT7UT+sVwcwDF84PzKoS1clnybpaSxw
+ KNzwj+9QAKKzHVia9QjBf2ieZqsTlnZ+jqK3iGIvzfXwhmrm4dn2tifRhlV0RJWPSQCAbZ72mEYjYgdZ7wmjnQSjPaLjuGZEP9PtjbqL
+ O/3CgelBJePLNCOa+Zm8Ua04k7rLVrBYAfG/Ww2jBSSHJI0IvKrvSQBAhxu7lvCHNuA5fUWXb1KyNlo2nskXdBY6jVJBJlVQKy+4z20X
+ nKOR1PyRSomGQt56JBo0KN1CQuYM0xoX+61crrxD7cH4piZ4tcU9sI0sATV9QEI2KI+fFMdDy2iNjmaiZVT0R5GV56YEmgVmc9nwrxzf
+ nsvpQfzUVg+h3vaLYpfmZvFL4g4OcUd4treqxvU+IzTK/W51lqarbfvMViQbRxnw/fnsdiOz1mgB9Im9viBaFxMkX3m3Pt8JYkuGiFDx
+ PyoZn6WSmXgFjl4C9fkZjji/VQtyQjY5xYiR57K+g7qB6VhOmrxAQshQVpTYKtekMecj6L1b3Jvlp4ULouFmXDdpWck60Uqx9cz3mcer
+ YSZk2CU8ztLAoToQzGHWi0jS5h86nb7s2e6lN5o+s6VXHrbmFrriFGQr08mQjpWzTpUxy+e6lVfZylC1cbuDL0tRn9e4NYTFnw4inr/I
+ s0nW3iCFz+iB3ZW/Ik2yQKBDnDTrUNOqySv7pehTMNAqGlkyqjmx+2KnJl3tFeh2utlJxUPG7I7ue9418IyzDc8kPYw6QvyA6ua9eQqU
+ L26oJopvBentQ/0svJ2qgPVgobXKo8MyX2aetUHuPtHK6VSt4m6Vhl4vyknmHSXrxJWhiG+2zpfDvf30dRoad23ky08Cuf8OhHAhFxF4
+ /2/FVTBaPqwu1Me7XqTMKcFcyH2OqMfsaGWP4eD/X+WY3LeX69Z2+uU6iyMIvrp9Qvb8tNVXf+cEta+3Fcze3Po4jpIXrHKVJXvfQ+ti
+ R7dymE1IyF02jALRoQDLPuwZiF0EBaGhQW2vlyxgRluBwOe50JdrskdFSJvPowLbJF4twZcqX2nTY6gXa+xuAMnL6/1EpyohwYLq61vQ
+ v+AZJFimsw1yry7ta7JypgFYhqTSoHE5TCtlWiKsFyFGlxfRJfCo5XARoKaH0i6YHl/fQP/AC0jui/bbH/GuHvyt4DGHZ0r507KkcW0l
+ STrYC1I9kwgeZXsO5x0kvRmWlyJ/bZvlZLVIh+YsmJfCCXDyjvYchSy9C5d2lv9S/Gs7xSoNwQMJH8Fux6rN85YlSvUlRN3sTQk+CWeL
+ BwrZtkj8e2HEa2hQRxDpib1OvLWo1bFuZO42aVrO+oon3f4ovsgqF/TmyWjzpMbnq1fIuJi3ALAC/7SNJZKNJKnguN6EHo2MAxJeGnMV
+ snobEbDT65CG6zU8SpYKmQjsK2uroesqNqzU7dmpKSw+Olk9ofmV15be+vHKFfBQZUkrwbikiU8TyNLRetuVy5W3IcxRhsd4g+j9Q9Pd
+ kuDsB5VX01CqJPvAG8lZGtiw4N7RHPZSB0KGHzr55mZSJtnQabyyEqzfuH1f1SVpnwp2H98noWOnjlI7jpcO1cQTKJ8voO4O2Wp2gCf0
+ IANv3G/hIhqBhqTnl9eYFHuFeiMy2PZ8EbByUZiJJlbp6gsASIvLx2yHQqkwXhhpJSlsqLg1R/X5BQn/XazhdIrgKQzbCtd97B0t0q66
+ QrxSLWytvAoZw/fuTYDvXL73v0ImVhHy2Ykia/lJ8mjw5a3VtN6On3fIc9tTP1cjQ9GG3JFExO1gK4dw+EmDSBo5Dbep3SDSkPNefpZf
+ FexZgcR+yPCMvCX9dSvB+vv7d4CXvULzMVWazBGR3kxUOYJ6FBIHsxfiUMF9WY+IpNiiht/YNsF+HrSBxEawxy/VC2dasvBBssKHQ72q
+ zspSKlhakAcDCP/BHPbxxWts419aOeqQvWYHmBFeAW6H3E2M5FkIf0wi0qCmBO7dnd4DdNgujewV5CWt5MUHVCZZyWAZ6xUi7rJ4g4aV
+ d+y4lYgtpJCW4gtAVq9HIfOJ3iCNDomFws/NKkVlXuWFsOsxIJ5xR4LSDWXjlDfvpPLQBiC4ZM/3LJWAceYVdv4v6Mo3OVcrSQPp0lc5
+ C7oXKlnXwBps33b4eRLxyxxFHO8N230JumWOSW33SHItxQfw3l11aBFh7oHf78xCMhz/VX0o/+U3Ug/A+EicLF9RFt7E++Te9A0z7aK9
+ w+hE6lGkJtBmsHi5sxEqV3jesx9+S0Faqo1bKRoSxQ5Z4pfci/J4nNbNRHcOkZEFjdPRqppo22Y9DavA7AajgOXdNa6ZyEy/tpxjIO3P
+ IJyI62tR6LdFsOJdua5Y9GwWdtxzAp6rTu46gshLWsnLR3CUquthtRBb2mVYIetJPdlFiM5VSoQBokeaj6DdmX7GoxVoJ46xg3wOFI8d
+ xM5CrwWfcW4n3kKkjBuI79y8y1zNniRRkMM5mtZONJh+OXoQcjrKyjs43JYmjrjnSZDdkbalkDPw/N3UpbZbj+8kDetp3Y4GQPZ1MU+P
+ G3lZHHBNZ2folPtlDXxZJqq9xjppeFgGNFzmQapyXvMSkTdPAkxq0LLC8hBx3gVh2Ju72fLQ0s709LuDsYGkcUsLi8vChQufgjL/Jz3H
+ oB0+cIsHwHM82Yy605SfWQ0zouv+x2d0sFKaCPmwoEgGwbiQv4CgzmJXHnIK/j8fv93CguFVvHZJiyIvzyBdV+2ju9l5GZqcGq9r4NkY
+ SOGs09Ex0YopB0YwXtqgrELhvQB2+BpI60gQ+EG4vppeJCrHGxD3JfjvWMTrJ7jvOtwvROyRQjsFnsJnxU4I53o2tM4m+Qq+PQc9PInf
+ vPj2h72UhKlPj3oDucU3Is4/hY2Ogc0Oob0k3Ch6LRsA/P5ByLdQ1i7E/bKem896++FXyzmImDbA5xuQhp00LXRC5EUkPpuJ4HGf5S/0
+ PId6rK7ZlYBD8wUJ16UvbT4m81xI56HU4zUebcHyj70chHOPpiEzwetzDyAvas4PyIjk3QSkcz/os15WXl6Q+qJl/c9IqyydbNQj3JKQ
+ tKxI8FvVgO10hRqJkAh1cmkXDdxINGwWpnZJi2Ie1yZ6vUwfwMJiXkTiTaAQvr1L6W4mLKBCeiRqS79WCv73ECrJg7hKga65J0ruaWc4
+ xpfksBIuTaM9lIAyeVsGvyJzfJtxZoXvoF2lfNAuZhPGX1+KeUTt9rCRTIN7aLdJ5Qv6hODRIFyMevAkxt8qvjd8aUScPNdCbJng1d62
+ GELsRqK0FfT9JMfkvvNEcQXxvd7X2S6SNHP7hCpRZyN4pF2fG7VVUH6jlgV+eQIpf555pGUpX9l3+Wcr/b6pqnPHb1bACseCBQuehsRf
+ SSPi2r7XNVmYKaycXA7GQkOiErKK3S56/NzouUwiBQAZKelokok1RBRV/lsrI+PXCdJMJ66w0c7r9LM0cqwcJvyO3327dcRGKkbuD0XR
+ gOyuCZDYE9vkgZWngf6B50K3vXnc6fJkjZyQdRO7+fclNm4i5tmd4nl00sixodI0GLnUP9tIRF9Urvyqfj8U0y/j3FF8kbxQ5vK2Xkcz
+ sXy7D86JvGfRKS/U4gg7LKUN1Y5ZCJ5lVRu36I+2CVo7BGrPCj+h96/8lLXR8QV1ScrIRjRkb6du5YNcTs1sgOd1RYdr4ZkOsuuWftv8
+ /3Ku7We66j0cv8Ahzd/Qbj/JoNtpbia2SohxYOE16UacqM9I4hHYxt64rWn42kB1D5dS/C6GmYMosgjTM9lu+qILpNEzvlRXWUTxFyUF
+ CqYDXv3vxFYuf9LpQxycZx79QFUlZc7KYn9//zNApLa0NUsD6PKOw5o6pNYhgk8IDvE6jg2lhNe6YawXSTvTBB3Jnk/tEDyQOB3osbyS
+ +rU8Mc/rw08rWv6jFcgPWQCS0oaMSzMh/M/1aPVfd2GFDte5yBjZVRFGzOJV9ISwJ6DEtcq8m0aFy35jBUalPl27ySSGNBV4NgvTJ4Wb
+ +UsCFoMAbVbCGvgNBQjya5onlHYqZbdE8tyRptv50GxBDxT/ZT10ehPnW2TfpnJ8LPUQptOu3LIa996hRJql8dPeQXw56ut2vs42kZAP
+ eh4n6qQmbZOlTvDeDeIwdXAIxH8eNv2SpD99g9tI+JzYEeWTh7HIfIvxYAvQRizbtoWKidnOll8n5V8hS531vxkh+SRQm1RSQ/RihWwm
+ dvwbvNLKmzQ5liE1sAKz114Ln4L7L9UKk6WSzT5xFcJWi6ytlCtvE2MAfgXqFDyn4QmoROcmlTKbRzgdYm87cr7mIMbZ3gqFk7AryE5W
+ LPE+75lWIqTDsohyeAj1KKQcmq114jDr5L4SKMtr5RzqITqxU6I1ysw3hHG+1oms9Z/xm7Spn9/g54WVJ26DgrS7t3+z9XzqRdLm8in+
+ oAQCNIlrHlK2Zxo927Y9MsMMKIdfRPEvO/xmZ7fFyH0TKmWy9rZZwbJKxkkgPJPpIIdZKvIugVaKlezqiiGAbpC7wcoUN/FCuJcnJN9b
+ trZJS3+TM/HoUJb8F5HSOQBInzYYILnoFdQDsByyoieVnb0n1UtJ64lWCdQ7mg7IQ0A1sHIgW5aAQLVOZHJ6WMb0yq0BpLFUtB0/ADpc
+ fYZd+RKkNcztOGbGG/dxCTd1qx38+CYcwnDRGHyaPT3UoUNRPD4A+R/8LstBoWM+ehgfRfnZXR5QPVxyi/vfwSFwvjjH3zrR6GWGVUhc
+ 90SiV7mWOPMRbNMmWqBsPHkDDCinShGtiMuMq1uiugqc3bMUb0qvjf6fcVH7JKtOUDi5PMwKX1fJ3WBhcFdBEMcV2siY7SbFebqFw3ru
+ hZ5oHLaR8VibGEWllJVlGYnEPyjFTg4TG/iVGmX1aC23Qtr67FSygauBXHmNPqGqSCJtE6jVfbd9QtJryUae1rjxCojiJttM54FfXiM7
+ YMbV23bqoC0T/gMncqnbbEHUhhl9GPefj7JyBZ919Sn+DcrJkfJ/FJ92wCtfRV2/8HWgAdiH9qTALLJtyQycdyBICiBa4PcyEVpYe9GT
+ l5dUlDDuhxGTcxinIK6kQuCZV+DZPGvgpUBpxW9ne4BuCeNHYpeDj108ox+g0MnaXyCZCJ0GJNslo3DvjEogW8Gq7Ugg7VTO9iVZA49u
+ v74vYS85sevO8p/kc6PnJ4stkeQeRDtTj6W/hizK0be17GZxEmyCEfGOFlPPFGU9NWySkcQM/Q/krPfWuN3Jt62pr1PxM1g8UZb5Bro7
+ 6L09fvL3DfqKBAI0ijd+E1IGN74GzyU9YYsTytBZ+H0tCP6fMKPtgw+eqXwWYazBf5wb3J+/zRTB1yQMET2Wa9e1UPUGyTsyTbxSfL4J
+ Bl+kUU5ToBJiQwYdYhUG16wkLd6UVmbKzJOVE8ZDKpraZ4Lp1CSbfTriUWVAQvLc+wjEeQrjp2RAgptJb94m284FaTyBcbS4ovxXJ4jT
+ N+K2iutSO8LSvDn/ivDOzjEMakMK6+rJpV1U41Z5K/Rv0nCyefDIR7XXtQv6F8hQRIr6mBWJc1KRt9+TIbGscfXFhns4b+LP3zWsJ7iH
+ 8ycjIOuK/iSA8/JrlJkzQfTn4PNX+Zs4NSgL7HHgyt1l/4u/dyrfcsEym0DEvuBVRnlhiYaYASF5SmVgIXLxqZxVqbgDBYg0Xql/DyrZ
+ Meyea5pSkTPvVVv8A5l5FESWunkVop2CllcYdyF2FlQlmPUoZCejQMkePMQMkbshIXleUTf+G3GtHibj8nbaiR5hu5VX3ilcBpSPM7KW
+ D4gclsJnoULKWgOC3w73uTmJbOXFJsnviOe5LRU6RBQsExJX9Aw+wPzQPMmaH26FT8Q9l1xj6XNJp+A3GrDz8RKma4DbcbCS5afGKfVx
+ N68bBP9q3D9h3ng1X+OLkfYf4feD2SPUjQzfA53cmuRV+I97YgnBz5gHb/ATh0h+HJHTbowYkpWxHWOmFy88ZoBWxjHE6TCLo2b4lORO
+ +IUDJHMCzx2F7rTdZOnOcQkZwv8NdXA9M7rbJ+I/N5bnKoYRfbdtxHjLSzxsYJQwaLOLvJeXiJrD02cSteVqoIyKcCFtpkTK+DOvu207
+ 6uYQn8wvLVywCGXK7etidnJ76kSXqE398sFrI7H/3CoX9eAIv2Hjlfvo4L48e70Ywd/Q7luidUgafZTrL2kYVvf88FsJ0y6NG9J+kqqr
+ ye9Owho2esjoSV8v8ygZ9tRvIIy/LkGt/KxK2n55dZ9jbiUiczaV/fjdygzSfRHs99N58xY9Fboux/9H4vsZdATx7IvjKL7b9ueZcYIn
+ ahPHrlu0ggZQb1XIBZUySyFIJaqTBT/x2HVd7l0cu+Rbkhoti2Nqr9QqBCfSYHy3fC8lwWu83CqGUuWHolCBlvm1iBv3Pdnk2UgKPaRT
+ hCWFEEKd4q0zLpIGKdzRhSh877ITiQhN70x57Q3hkxJXbaECHIb0XGuNlF++IJ2wnb1IRn26bcZA1XZwGJB/cvasVbz58+cX8IyckaDE
+ wfhMJYyrG4fWNfWELWO0dMvmfuX4XuYfrpKXqQTxkPigy0+7UVeHGu6kfIDcTtQwzPaN4zJZpFzKs+kPi2kHSY+Qb6QibCs3D3vx8aU+
+ vpPuQX5IPjNfwHeyzw9hNq4l+Hh0EsGz/pXjX8jnUvQJ3HM/8uxGzu2gfO2GcnY3npGhtZ4geMLPIPf6eeUHiPSDrBhaEKwwWEHNWhl5
+ v09cFMksr7KvBBkfDyPJy0uKSQeEp4E9I68+l+OrNQ3p4q0eDZ9BRh5DPX4jyEqHDHwH7vs17nuA93lpkDRqIXKVB/pINvpfVVw4UgBr
+ 7sd/Yhc0dtTLgohCNc5KyQamAbF3ovJ3C3IAh36mHecijz8Iu/6ZaWf66m2HClJjO/29qe28+/k9KVMU+Z0b35WiT9mEIGGeobyIhHLO
+ cxIsH9MIPdgF8/fb7L1nQPI0cUSA3pW7b2FDHa3EnToV/5p6FG3nsdUJd4RgfDnDyJJmX/ZzvaFJe+B3A379h8N1wkLYvVGcsgjTvXDB
+ flLfOZmq6omkvKKuvQ55sIZEz+9Vgo+vhFzKzzxpi/roRPJ73C+T13zXQnTmOemua2ACfE8BkVwEUjkVEb7TCgKvJBwlLEdMzvNhBasV
+ /u7+I2lVK58Sl+pag8p+MSrg0TDsizVogRo6l1dq6ejv72c3eZUSiMQhhSTL1GCDQVGoDY1PVuwdoBDsD/L9OtJ5DZ572K8w/Cw6nAdH
+ 8UlKfuN/vIc9F/85/Ecb3uy6kZVDuaxNgxUwLn7B73Uwrn7Zmjdv3lNhu7ewhwTb3Yi0rq+3QVrb2TMUfsd/a2G3v6JMfQdhvAF5ZquK
+ DIlX2E9nxh2m8hPk4w9QFn+MytpU+L/EmfMeck3mPixtSXnVoamfUTeeOwX6UZdaS/U+2RdfVtAoctUDH2Z/2bqb5zGX4p8jHSfVx6GV
+ SPwisdXPQGZ7i2IXt7bj1wpW7+I4LiIOX0Yen4/8OBfxOA/yK17x+y/w29n1gv/O0ftqpRT/Ena+BPd8znOaEocS9udhMd/weEnSiIbt
+ Y7Dfh/mZ9yLcY3GvjNOzPLARQjxfyu8+X/QM6iK1NSK9DxKxHAa5AIa9HZ9rKhWFFbJeGt2D5x8pl6MVNC51opAs2HfffWUVgkG9q4QM
+ coAZIc8z7gjzYSEK9cxTiM22r8PzMlmilUMymJlan3H4Ds+08mrJ7HLEZVTXQu6ASNj19uB3JTDG6X7ISsiVqDynQccx3BYXOp9jhc3A
+ cOt/m01g/H2iJ/DbLrDdm9jlh+3OQfqvV9tN2mnTtx3+Z4N9H2w2huf+iOuPWPGgaz8Sgao3mGfWVSKqg5WZvNItUHe7caNMK/xyQ0Lm
+ mQbI0ydQ2Kt2w7HlbetFf5f76oXPcasKO/ilARqlc6r0T7ttMoPGhAFqSGzRokVPLJfn7w6v9l2ohMeRjECal6GS3QAZRWt4G76vxufb
+ UeHGcc8/8RndmQor7bdQgT/ArguHgEDiyTCDwg5crqn8OZFkACr7GxEH5wlmIHghEBlLmy8vDNWTEkGita6+D9htW3hJO8KD2xMNxGtB
+ OosRjyNhg4/CW/g4PQBcOUyxlF18EhLuexHfLmykz8LpxKvqvQKWrfryRbDi4vedKqXKXrQdyst7aTvY6yNiO+npcQ/4yjDueyvfKsR9
+ u3LCS1XUwAun9ytdwJRoVA87iHbLyOwrY81IjMB/27D1g7f09HJ5wfM5qURSw/eXk7Bw3Zlj4Gwl9ZFJoO4uVMCkEJAcuMafxA2ySE3w
+ 2tsYMU9wioL1OC8dHQF1Ueds9tbTgHbtsO3EUaDgcxoykPXWuWTqBje3bs33rhEG484wGoWdVqBmpghN4t5Jgc5GabG1+PX/8Tc/7/17
+ mj3T80hOq1eD5EGiQyt0V4ygBhbAC/8iu/XqvSfjty1El0iC4EvxnxBPWefr65wCMr5rJMPPXob76eXnZFzfv1//eyxCygdXHtTZrr7C
+ yEldZjv/fv7nbgkICGgHyUn2VsH43YTfTfh9Ols1j4y3jsrRqVkJnuPiSvCnK2kAqQm+GZj2RhIwNRrZjRIQEPBYgxG8m4SJL1ayTkXw
+ Oukpa9wjb88KoF2CDwgICAhoF0bwPCcTxH6LjqenWyKp4/TyTFT5gCh03mIg+ICAgICZhhE8X2wBWdup8akJXlbc4HODl1gCAgICAmYQ
+ ibctyw91LbV55lOKEjzkwRigHoD6AsEHBAQEzDASbxsE/24h63KONfDl+M5yufwc6vEmbQMCAgICZhAJGfOFIh1/T0vwnIS1NfA3cA0/
+ 9QSCDwgICOgNVAmep+m43Sll6aOSeCvx18BfYC9oBYIPCAgI6AHY24Vcfx+X41+QrJXcU7/kxCWSePbHohAIBB8QEBDQAzAylh3z7NR4
+ NzyTag08RdbAR5XPiEIgEHxAQEBAD8DIuOwO653QfWhSLZGMS24Yh5OsZe8gACAQfEBAQMBMwwhediMsx/fmXAO/CQT/elHorcoJCAgI
+ CJg5JCexyza97kASEnYmgoc8WEIDIRrDGviAgICAnkBC8PDAh+m9yxr4DATvPP7KbdzqmHqqm40FBAQEBMwkEk+7XI6P0fXsnFylJ79R
+ JltbCe4Tgi/Ff7EDJALBBwQEBPQGkslQLnOU5Y5uyCWV0IvnAcloHH5TLpe3pZ5A8AEBAQG9AR2eKf9LVI7Ok+WO9cMwU4hbIhl/VbQB
+ geADAgICegM2RMODwnfjZmPl8sAAz+2cSkp6xXMLF/Qv2EH1EGGCNSAgIGALQyD2gICAgB5FrkOFdUgmGcsPCAgICAgICAgImGZstdX/
+ Dy5zyZJv/hArAAAAAElFTkSuQmCC"/>
+ <rect v:rectContext="foreign" x="0" y="616.927" width="38.7" height="20.5851" class="st2"/>
+ </g>
+ <g id="shape1001-9" v:mID="1001" v:groupContext="shape" transform="translate(289.786,-381.889)">
+ <title>Sheet.1001</title>
+ <desc>Spark ML Pipeline</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="56.6929" cy="615.392" width="113.39" height="44.2404"/>
+ <rect x="0" y="593.271" width="113.386" height="44.2404" class="st3"/>
+ <text x="13.22" y="618.99" class="st4" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Spark ML Pipeline</text> </g>
+ <g id="shape1002-12" v:mID="1002" v:groupContext="shape" transform="translate(9.25393,-458.504)">
+ <title>Sheet.1002</title>
+ <desc>AvroIterator</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197" v:verticalAlign="0"/>
+ <v:textRect cx="56.6929" cy="589.059" width="113.39" height="96.9051"/>
+ <rect x="0" y="540.607" width="113.386" height="96.9051" class="st5"/>
+ <text x="26.52" y="555.41" class="st6" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>AvroIterator</text> </g>
+ <g id="shape1003-15" v:mID="1003" v:groupContext="shape" transform="translate(46.1043,-458.504)">
+ <title>Sheet.1003</title>
+ <desc>Pushdown Filter</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="38.2677" cy="615.392" width="76.54" height="44.2404"/>
+ <rect x="0" y="593.271" width="76.5354" height="44.2404" class="st1"/>
+ <text x="12.76" y="611.79" class="st7" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Pushdown<v:newlineChar/><tspan
+ x="25.67" dy="1.2em" class="st8">Filter</tspan></text> </g>
+ <g id="shape2-19" v:mID="2" v:groupContext="shape" transform="translate(9.02896,-576.316)">
+ <title>Sheet.2</title>
+ <rect v:rectContext="foreign" x="0" y="608.42" width="113.836" height="29.0914" class="st2"/>
+ <image x="0" y="608.42" width="113.836" height="29.0914" preserveAspectRatio="none" xlink:href="data:image/png;base64,
+ iVBORw0KGgoAAAANSUhEUgAAAWgAAABcCAYAAABDeS/aAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAACxIAAAsS
+ AdLdfvwAACOESURBVHhe7Z1ptB1F1Ybvt5wAZRJUEFRAEHWhOC0FFXXpDycUFMURJ1BwWEsFFAcg4ICoIBASBGVQEX4oaoAgksQQQEIC
+ EQcIEiBgCIMjmdAIBPo7T51+uzrndvep7q6+94a736xandune9fuXbve3lVdw0jSER5++OFkyZIlyd///vf0TDn++9//JosXL07uv//+
+ 9Ew83HfffclNN92UPPTQQ+mZctx5553J7bffnv4VD4888khyyy23JPfee296phxr1651tlizZk16Jh5WrFjhZD/44IPpmXLcddddydKl
+ S9O/4uLWW29N7r7n7vSvcjzwwANO31WrVqVn4gGZyCaPYbj77rudzl0AG2PrYaDM0JcyjI01a1Y72fjeMODD+DI+HRvUPergMFCXqdPU
+ 7diAg7AFnDQMcBscB9d1hegE/UjvH8Dx58yZ4wwJdD4PFfK//vWv5Ne//nVG5jEKXzIo9N/85jfJf/7zH/d3lexFi65Lrr766ojO15eD
+ Q82dOze58cYb3d9FUJ44Hba455573N8xbbFs2bLk0ksvzci/Svb111+fXHXVVdGdb926dcm8efOSP/3pT+mZ0ZBekCi2EIHFtAUykS3y
+ r5KNruiM7jGBbbExti6D9KLMKDvKEMS0Bb6GLUR4VbLxYXzZBzzt9QDkSd1btGhRemY0pBd1mTqtYCqmLeAgbAEngSLZ4jK4DY7TS76I
+ 49qiM4LmjY9T37zkZvd3FUH/+9//TmbNuiz5xz/+4f6OafC//vWvzoghBE1FWbBgQZT888CZr7zyyuxlVQTlSYQ0a9as5G9/+5v7O6Yt
+ li9fnsyePTtrqVTJhpTmz5/fCUH/7ne/S2644Yb0zGhIr9WrVztbEMGCmLZAJrLJA1TJRld07oKgsXHIy4oyo+woQxDTFvgatlB0XiUb
+ H8aXQ1qkdUCe1L2QlxV1mTpN3QYxbQEHwUVwEiiSLS5b0uM2OE4t0g2OoC+//PLk5pvDCPqyy7ojaBx7vAn6iiuuCCZobKHukJi2oPlI
+ RVxz//AI+o9//GNnBE3UGELQRLddE/Sq1cMjaHRF5y4I+ur5Vztbl0F6EUGjr7oAYtoCX8PnQgkaXx5vgqZOd0XQ2CKEoOE2OM4IugEk
+ wwjayzCC9jKMoL0MI2gvwwg6Bz28EbTXywja62UE7fUygvZ6GUG3gBH0aBhBexhBexhBe5CnEfRoGEHnYATtYQTtYQTtYQTdO5dyWRFB
+ T5kypTQBjkUjVRi6qGvyMILOwQjawwjawwjawwi6dy7lsmERdBHhcu6MM84YVa/OOeccI2hQJdsI2sMI2sMI2sMIuncu5bKmBM0Y7nyZ
+ M9nl4osvNoIGVbKNoD2MoD2MoD2MoHvnUi5rStBMCJo+fXomm4iafI2ge6iSbQTtYQTtYQTtYQTdO5dyWVOCBueff76bUo6Mc889150z
+ gu6hSrYRtIcRtIcRtIcRdO9cymVtCPq2225Lvv/97yenn356ts6LEXQPVbKNoD2MoD2MoD2MoHvnUi5rQ9Dg1FNPdUl5GEH3UCXbCNrD
+ CNrDCNrDCLp3LuWytgQ9CCPoHqpkG0F7GEF7GEF7GEH3zqVcZgTdApJhBO1lGEF7GUbQXoYRtJcRk6BjwAg6ByNoDyNoDyNoDyPo3rmU
+ y4ygW0AyjKC9jIyge5UdVMmGNFhAfbwJGlt0RdDIJg9QJbtTgu7Z2Ai6n6cR9GhEJ2jUBCjNYtbMkgFVBM3AbZyvK4JmcW9tYVMl+w9/
+ +EOyYGE3BB26YP/KlSudLR7tC/aH7C4zURbsR9fJsGA/vgeqZOPDXS7YTx0sg/SiLne7YP+syt1lxGVwW37Bfn6JjZSg4wnWA/3vf/9z
+ DhVSEdlm5pJLLukkamS8IVsFrV4zvCIuXLjQVUSRUls9dD9b4vz2t78NqohstYMtutjm6Y477nDb+axcNbwisqALFXHdun5FbKuH7lfL
+ KiRSopJgi2V3xt/mCZnIDtnmCV3Xi5Ra6qH7sS02DtnmCfKk7ChD0FYHIBn4Grao2uZJwIfx5Wybp5Z66H7qHHWPOlgGXUtdpk5Tt0Fb
+ HYBkwEHYImT7PbgNjoPrQAw9PPqyOoig+0DpmTNn9h5ieFOWt9aMGTOy6CAm/vKXv7h57iHNet7gOF9cQ/dJicoVStDYQtFBTLBi1kUX
+ XRTUrL/uuutcJBE7gibyYi2CEIKmyX3hhRd2snktMpEd0qxHV3SOHTViW2yMrcsgvSgzyo4yjA187cILZwQTNL7so8Y4IE/qHi3YMkgv
+ 6jJ1mrodG3AQ9S+kNQ+3wXEi6C4wggI4XqzEm5XE2y20AuCo9GvRH0fzhfuLZNdJyKAgKfRrr702zakY0o23J4VDRIEDttWD+5HD5pbI
+ rerXEmg+EUnQ5xnTFjSRiQIXLLgmzakY0o0IAp3p84xpC8gAuf/85z9dPlW2AHQB0Iyk3zGWLZAlf6uCdMu/NGPaAtsiN7Q7i+iSDVsp
+ y1i2wMfwNSL5Sr9Mf6PFoZdmTFtQ57BF6GbJtDro5qCO93V4cJTsOkm2kL9VBSbSjVYNL01mAj7wYJ/3imQ3TeQzwhuZQqeAcNo2CRkQ
+ Im/YGb038m1L+02QKuhh9XGMNyPNhiL5oQkDs/EjjgQphfS5CjRbcBSaUG3ssp4tevJCIp98wWMDCp8Iq0h+nYQMbMHzhLQkBCKUWLaY
+ O9fbQh+OqyD96CPGjjFtgSwqd0hLQqC/Ed15hva2mOtsijzWYxgG6UfZ4c+UZUxbcAwJpAQISbagbGLY4le/+pX7GDssf/X/8pKdd8W8
+ zBbU+aI8QhP1De6hpRTS7SUQfKFDW1vkk+wCZ4wQNRKq0/m/+KbFzVPP0ZDBF2maQXrIOli7dq1T6ve//72T5XTqyS3MryilOnBEBsbj
+ rVYXNG/4WPHnP/+5vg6kVA/sQFLkXAc0m6gIsWxBxNPEFkS6sWyBb6gZXQfoTYuM58g/V2F+RSl3DzKwa5NmKeUo/25jC2yJTdWKqIMH
+ e5EaZdnWFiRkUN+od3VB/ZZ/x7CF+nzrgMibOh7LFryEiaLrgpeb/LuRLQYSMm7ocTLc3Fkf9MTC8DehwWDYkDA56nRnBB3SPBhEk3uq
+ gLyi4X3D0IUeddG/J54ezhKN9YiHpjo0KccybNC2SP/FwwZui4mgR2Qd8pgkEbTBYDBseDCCNhgMhgkKI2iDwWCYoJhwBE1/TqzUBkXy
+ YqVQFN0bK1WhzrV10VR27+pR98ZOgyi6Jmaqg6b3hSBUdv66YdeOJQb1apMmGhxBFynaNjVB7Flr4OFH6slso38oQuR3YYsQFOkWyx5N
+ ZY+VLZRPrOcdBvILyav4mrHRUWhadl2ii/yb+hq6xE6g0whamYRARMo9jIlkwPiRRx6ZfO1rX8u2hilN005Npp46NTnmmGPcotcMOs+P
+ tw01el5fJrcwoYLdd7/85S8nJ554YjJt2rTi/Ick7jvppJOSmTMvdmM3QZVt8voyJnv27FnJUUcdlRx77LGF8kMSOkydOtU9j/ZA67mB
+ OwrSiXHHv/zlL5Ozzz671qD9Kuh+5CGXiQkalx1qC8YNMxmAMqasi54zJGGDb3/728mXvvSl5Mwzz3RjrIt0uW/FfW76/3HHHefsf8op
+ pzTyAe7Bf/CjU6ae4sbJanVFUPX8+o3xvshiEoMwWH5Nwaw8fKNs3RzpQFmwTgW7UIesVdEl8n7B+HTsAlccffTR7lmGlRO/f/3rX0++
+ +tWvukk3zObUs9R5pq6ff4QB+ytWrnAL6DCDrU1iZhbGqjtJRQ/JdM9vfOMbyZZbbpmMjIy0SrvssouriKFTivOFwyyzvffeu1Bu28RG
+ kVWQHgxSP/7445OnPe1phXLapOc///nZ+hZFdvnud7+bXXvwwQdng/eH2bAMug9SOvDAAzPZvLQ8RsvWfZDBCSeckDzjGc/I7o2dDjro
+ oGxJAGYvnnXWWcnuu+9eeG2M9I53vMPNFhPRFNlW55hi/oIXvCC7lxm3oGl5AN17zTXXZHI32XjjjKSLZP/gBz/Irn3ve9/r6jxoo0cT
+ KD94hjr+nOc8J9OraXrSk57kuGfZsv7CXHWBLnAfHDjIi7VTj4vhZLh5hLc50xSZLknBt0nI4A3L1EfWOdBSjlWQg+IYu+22W2awrbba
+ KnnmM5/pKuV2220XlLiexL2S89a3vnXoGro6z3KSJ598cnbvJptsksndfvvtR+VXJ+2www5O5h577FH60pAtiNxf/vKXZ3o0sUVR4hl2
+ 2mknJxMCAmq5KG9eks997nOzfDlqYaN81FIHuo9oFHlPecpT3PElL3lJtg7FYFeU7iG6fd3rXueuJz35yU+OZgvkbLvttslGG2+Uyf/x
+ j3+cfPrTn87+3nzzzV1ebX1A+fH/Jz7xiZl8ojk96yh/SG3yox/9yF274447uuMnP/lJt1YDKPPpUBDVI3PnnXd2R1oJQHJ1pEX66le/
+ 2l2zzTbbuKOi+aZ+0QTSZ/ldy5N9993X6UEiqKtbTrpe/kjCDsxuBCHdo0y/Z/0OOA/ui8WjcDLcPAKJsjYui5TESKwGRXOMufE0R7Om
+ Y0GEpIKFQF/4whc6A2HcrbfeOjNY04TRkcX/3/KWtwStNfDDH/7QXU/+RK6Pfexj15MZI9E017rCeV1kC2z4ile8wl2LA8WwRVHSurvS
+ QflTfs973vPcNZtuuqk7Mg0WNK2Iuk/RGqTH8aUvfWm2xGxetv5P947IGVvkK1LMxHMOtlSwe1f5jfzfiMtPZauX5aBviiDoEuK6pz/9
+ 6e54yCGHRCNounmQqbpCgAIkV0eCChG0WrjUcdDUL+pCuhCl0gJBh1h+QVnIBrvuumu2rGvRs4nL6K6kFYQd4DzqziAfNk1wMtzcWR80
+ bxbeKigOBh1Jf2OAKVOOdobBQIou9txzT9eM+vCHP5wcfvjhQ9Ohhx2afPCDH3T3vPjFL3YyqHgyOk5eBOlBBM91pDwpvvOd70ze8573
+ uKjqC1/4QnLY4YcV5l+VuO+zn/1scvrppxcu0JP//7e+9S2XL45Hs4v/E03zXB/60IeSww6rn78S937xi190RDkIOSJORhcI+aosYkXQ
+ ImgR/8te9rICgva2oAuE6/K2gNSxxQEHHNDIFtxDedB1s//++ydve9vbnFzptdlmm2WJc0RUlP/73//+5HOf+5y7t0huWVJ++A/57bff
+ fll+IpanPvWp2WJaeV8QQZ9zzjnuOhF0FxG06klZBA1B77XXXu6aLbbYwh1Z2Ac09Yum0AsLneVLtL4ppw984APJoYceWlgWg+kjH/mI
+ 8yU9Fy1m2YHvHHquQRvrb7gNjtMCZF2gk1EcgAVY+NBXtp2PHp4mNQbZaKONsiY1EYI+8hFtFuVRlBSZIlMVT5HRq171qsJV7aQHjsl1
+ KiASTRZQR4eqJOT/D/Q3fa004clbLwmcSB9kYuhRBtkhH0GLoGNH0CK/oghaOtKfp2a3iOx973ufiy5AK1v0/pEfMjieddaZTj7E8/jH
+ Pz4jICquVt6LYXtkAFZNQ/7jHve4zN94eYO8jcsi6JgEXSeC3muvfgQ9HgQtPfiOIUJV3aYbU+un1yknlQf9vp/61KecLDgIoub/1AVQ
+ 9nxwGxynxaaK8miTQPQIuifaHVkflaUR5eA6L0gB+lryxn7DG96QdUe0AZ399C8hk0rHkXVvgfLWkaiWiIxriNY4nnfeee632FCeeegc
+ fU/kLXJ+5StfmS0cHhNFOsgJJwpBz5t3ubtGZPCiF70oI+cuwAgA8tlm237/Kkl9kV3gpz/9qctD/kbkN9i6mngEPX4RtPTgQ+4TnvAE
+ lz+JbwjaWaUNiILf9KY3OZnygQsuuMD9NmhjcRncBsdVdeO2RWcETf9MCEF/87hvOmPssGP/IxpND8DvSqEgD13PF9DXvva1Tqacj+E3
+ QA6lI8v78Tt9o3pRaJcLvWUH9a+P8vulM0N+yFsvFl4aoIkt6kK2mCgEzRAorpEt9nvXftlvMWyh8nwo3dJLH+JEmPSzKlBAt/71zfNU
+ fvInNmUgHz64yR4a/ihbGEF7KA8+qub1fc1rXtOr67kItkY5uevTBOj2QuaznvUsd/z85z/vzg9CZSmCzobOBuZbB+NO0IokVOgM5QFN
+ H1Zy6c7QG1GFedppp7nfsgqQHlmYnt833WzTrJtFe6ONhfNJZ8ZwkrdI6WMf+1j2W9fQc04UgmbsMdeIMOkDjkVKeShffSCWr2yx5RZB
+ 2x7VhfK78qorXT60ltSkViSoa4ygPZQHdTiv7xvf+MZaG3JU4TOf+YyTqfrH9xowirvSvycFQa+5f03yla98xUWufEij7xFQIE2SIhTG
+ 7vKBEUNrWJC6Lbguf+SLrd6a9IVzZHcHoGu6hGxB5SBvOd8RRxzhzo8F9JwThaAZL841DIfiyAe67JoBX2oDyWR4XT4/hiPqO0jbip+H
+ 8sOm5CMbM7RxsM/TCNpDeVCHyVt1mjquzRfa2oIRVsiULZjMBAblTgqCBnpwSJUvojJ0rALXpAj6qTiGELT6tyYCQfOFfayg55yoBE1z
+ M7umwJeaQjIHCZoPlAoW2lb8PJTfIEEzcsYIuhzKY5CgP/rRjxpB10EdggZlRsUhGN7FVjJsRROayI++PAiWIXIYmmFMHM8//3wnO6sA
+ 6fGOO26fkARN5Rkr6DmNoPv5PfvZzx5TgsbmRtDlUB6DBM0oJyPoGqhL0CBfwAuvXeimXO6zzz7OSKTNt9jcOQUf8eokuisYyqSB9fR3
+ g6wCpMeJGkEbQY8jQe9sBJ0/GkF7TCqC1sNj5O985zvOMEp8QJEzNE2PecxjspEZLJACsgqQHo2g/XMaQRtBA8nV0QjaY9IQtB6ccYQa
+ waDZPCQmKBhBjw30nEbQRtBAcnU0gvaYdAT9i1/8whmEj3maSZdPdFfgFP2uC45VyV9rXRzh0HMaQRtBA8nV0QjaY1IQtB6aZfpYdwOD
+ 6GMejnjRRRe5D4RMp0RG0QfBosS19pGwPvScRtBG0EBydYSgtViSEXT/70lB0KyqhjE0QYRol1lWMWDD7MKh5xxLgi5aLEm2MILuHceA
+ oAcXSzr5lA0rgrZhdjVRl6AxAsbQjDFmzwmMj6Zg6iSbqNIMek5IQqvZaQW5rgia9aCNoOsT9HgtNyqCHo/lRpXHIEHbRJWaCCVoYXD+
+ OwZvAxmz0VTvTW2qN9ONyXvjjTfOVpFruzC77ps9Z7aTpy4nknaw0DV6XiNof/ze977nrlEQw1rIMUiJe9XKlM/hg/oNSAcWqlK56aP7
+ zJkz3W+6pksoD5vq3RJScjhB9/9mvVyMIYL++Mc/7s5jbKVQkIeux4EfNYslpc9VxxZ1IVvQf0/eRM+KUtTlVNcWuk73sYED8uhy4gXA
+ /7VMpPLXMxpB+6O2IBNBv/nNb14vgq6jn67XPawxjkzVP3Wr6XfpwALy0lWRfH45XtC+XMrvlx7FiyX5l1UdHQZtMakWSwpdbnTmJTOd
+ MUSMr3/962250XFYblTniEZ4SZK/7EcTOwbYoDUv98CDDswWO1f+OhpB+3zZPo5rtKgSe/A13TsvD+qZghhFx3ycB4PlAQmy8QDXqPyI
+ 7LtAkb11blIuN4oCMRMIXbCfNzPGyC/Yf/Ahh7jRHYC3c1EeRUlvcmS+/e1vd7JE/E0W7GetapDpwL9cfnWTkP8/0N9lC/bT/wfq2KIs
+ VUG20Oag2EKVgZ0jQLAt0t9VJuzijRwiZ9kYUgTKF3APMIL2+RJAaJcgESndHPg5qOMXKg/ql7o35Gt8G9AekVwrSJ+f/exn7jpspJ1M
+ GBkluVzXy2W9/OomIf9/oL+LFuxnSzvtO9rEFizYr30oH/UL9gv1trya4gxCxZXDQqja8ipkmym2udGWV3x8QgZOJDIoiwKlR9mWV2xR
+ tN6WVw22WeK+NltesdFsrC2v0IWPdUWQHgxXVP+zjiSGLbJtE7tPVNlC+XAd12u4I0kEw1oXRb6h/2MrrluPoNOPZj33dccYUOUbJOjO
+ F0u6fn2CLlosKY9836vIET/HL/D70G2eqE/cQ/1ChiaG8X98TygqE/TT7uIiRxLLMlDOfLxsU0dYsZCoXDsIldldH03ztmBPU56r7pZX
+ 1G26SZCRDx4mzJZXtmmsx+CmsXl5sRJfivXmzusiW2DD8dg0VtDfP/nJT9x1RPTYQt1ETRMThujTVgtBk4bK8h+MoMd6uVG6EcYighbB
+ FEXQQHkT8aoJjl/kX5xNE76lOsKONVUvCL0c1RLixUJ5KtqMmT7xiU+sF5UK+j9BTrZp7PZxbEHkLFtMqE1j2dqbpnys7cL5cMBbhX4z
+ MhgGGYAIlo0fZbCtt9ra9XXhjFSakMT1JHWVkNivTM2fsoqm8xCnhrqRcD7JpfAG86uTdtihv2MM0bC6LAb1kS3o22KjWOnB89S1RVHi
+ GVjnGJnaSVoVT5BOHBW1kYgu+DhUxxZcx/X8X5EiiS4U/+zr20DnzzjjDHctkTZHIuhMtw4IGnuQD8TMEZt3uR40KzWSj15YkAJ7aYJB
+ UlAZ8UFVJE1iuBv2dWWyXViZ4ENcnye13Xff3Y1kAoP+UIRzzz03u5euSfqB6/hFWeLjHN0sBARqXQ3aQmWx/K7lyb777pvpgS30bKF6
+ FNmClpO2OguxBdHz/PnzHefBfbF4FE6Gm0fo/F+xckWyctVK1xfTJvFmI+pQ/3EoZHQclJXsNM6yTdpll12S6dOnl5LhILLK3zvOmTMn
+ 2XvvvQvltk1EhlWQHowPPv744zuJ5GlOL1261OVTZJf8OT6CKFppm9797ne7yQ15Ww9C5/jwo5caiUgFFN3TBpLHS5yZjcpPI35AzBeC
+ 8qPeaaNSEt0L+q0oP/3GS4MgQi/aNokIGh/Ti2EY8raHlNjlpkhu20QrU8Scz1PQOXiGOq6XaptEdyLc0/TjK7rAfXBgETfWSj0uhpPx
+ kc76oEGRccugtxX30AdF5/uRRx7pNvNkaE1VYhEkjhQsfUezZ8/Ooh8w+BYuQ15fPigSyeIADD068cQTs3zqJu476aST3LjRbEhOhW3y
+ +jKKg+c56qijkmOPPbZQfkhCB0iH59Hed1XEk9ePDzO33HKL6xfGFieccMJQW0ybPs1dx/VcS4TG5CEhxDfQc2pPFq2xLpGPUBn7SndH
+ FUG0hWSu6lVE8vr5z39e2KQfRN4v8G+a1vjElGOmuLI9dVpxWWSp9zv1iXrFzuLq6wWhz5m/Dp150dMnzIQXVqNsW0ewRci45rwtIEbG
+ 6vNsU6Yc7Wwx1D97v7PvJUNbZ8yY4T6OKr9QW4A61zZBJ6M4SE0QSqR1ENJMyaON/qEIkd+FLeoCPevabxh4rq7t2wRFOnWpZ9P8YpcH
+ aOJrXftniC26KJ+mz4UusRPoNIJugiJFm6Y2KJIXK4Wi6N5YqQ6K7m+a6qDpfU2Rz28s8mye3/r3tU1NUSQrVqqDovubpomGCUfQBoPB
+ YOjDCNpgMBgmKIygDQaDYYKiM4Ju0p8Tuw/I9StVjFQoQxd61EX/nnh69C2x4dqiie5lcNIal0k8NNIh/RcPG7gtJoIekXXIY1JE0HEd
+ ekOH2cLwaMDk8OMRJkTccOMNbslNZq60SchgARG2qqo7WQUwrpLxtkyDRdZNNxXnU5X0HMhgNpLGHdcBY4+ZCs2MoqZ24T7sQNKU4Tpg
+ kDrjgL0t6uuRtwXjVTXtvg4YbxvLFvhGfnx6KNCbiSs8h+QN5jEs5W2BXbFvXVCO8u82tsCW2FSTqOoAf6Ys29qChAzqm8Zg1wH1W/4d
+ wxb5MdmhwBbU8Xa26N+HDNagyY/VDwXLSMi/m9oin5DBhiFw8wgFxCwtBnozy6tNQgbrbzBNkamPmq1WBTUPeMjLZl3mBtAz1bFIfmhi
+ AD+y0IH/11nQmynnDFy/9NJLW9klbwvkYedhkH7MJmJyCks/trUFCRnYgucZXN6zCkwuiWaLud4WWoK2CtKP5QKwY0xbIIvZokULV5WB
+ yosNeQaepY0tsCU2xRZUxmGQfpQdMzvRI6YtOIasVSPcetut3ha9sollC0hpWP5qDTN5ChnowfRo6vmg/Dpp1uxZjnuQpeAyxBa8IGLY
+ Ip9kFzhjBCVY/DtWItpRxIPiIQXP4HAUY8YYbzDuL5JdJyEDh8ZorCFbBenGGwtHYeorb+e2enC/3vLIVSRdZQsckAXyWbs6pi14SVG5
+ r1lQvJKdIN2IaNCZKdAxbcFsPeSGTsFnSjGVjwoZyxaq3MNmKEo3on50RveYtsC2yC1a4rMI7PJDxaUsY9kCH8PXIIVKv0x/g7yo1wRf
+ MW1BncMWLBQEhtli0aJF7iVLHY9pC/lb1YQV6UYgxQuOFhnr38fQI5/Ip7M+aJqQTG0mIi2DHpQuBQpHq2nFBFEgb8aQyHHBggWO0Ic5
+ R13ggLxhaQaVQXnmySA2eCPjUCGRIzvKEFnFnjGG4zGNn8WCyiC9eLmLDGIDmaEBBLqiM7rHBLbFxtq9pwjSizKj7EJaYnWBr2ELdT9V
+ 2QIfxpfx6ZggT+rewoUL0jOjIb2oy9RpLfAUE3AQ9U8bZVTZAm6D45p0l4UiJehyJepCD4TSNNNDCJpojZWgtIFolVFCIRlE8jSh1qzp
+ r6xXJZsIhehVpNRWD93PmxXnCyVobKEFbNrqACSDJRSpXCzGAqpkE6EQVa1bF2dzUt1PxSaSDyFoojVssezO/gI2bXUAkoFMZIc0Z9EV
+ nUVKbfXQ/dgWG2PrMuhaojXKTstgttUBSAa+hi1CCRpfxqdBWz10P3WOuqe9QIuga1f36jJ1WjuptNUBSAYchC3UJ14lG26D40TQMfTw
+ 6MvqIILuC8aZaSrQdweKRlLogagkRBIhb61QSAbRAU0hdf5XyeZjBVF0XEP3o0YqYlV/o/KkImKL0GZvCCSD6ACHCumTpyLS3IsdQbOk
+ KxUx5MVN/zO20M4hMW2BTGRrSdwq2eiKzlrHOxawLTYOeXFTZpSdWpkxbYGvYQt8D1TJxofx5ditCfKk7lEHyyC9qMvUabUyY9oCDsIW
+ VS9ucRncBsf51kR7PQYRnaClvCKlsj0JgR6evlk657siaBybfkdQJZtIqSuCps8zhKDdx9KeLbpoTdDnifOFdPfwRZo+2i4Imj5PPgiV
+ QXrRrMcWXRE0skO6e9AVnbsgaGyMrcsgvSgzym7Y2uZ1IBn4GrYI6e7Bh/Hlrgg6pGVFXaZOd0XQ2KLqe5G4THsSZi2rAo5rCyPoHCYN
+ Qd8fRtBdRdB1CBp9u4ygV60eZ4KebwQNyNMIejSMoHMwgvYwgvYwgvYwgu6dS7nMCLoFJMMI2sswgvYyjKC9DCNoL8MIOgc9vBG018sI
+ 2utlBO31MoL2ehlBt4AR9GgYQXsYQXsYQXuQpxH0aBhB52AE7WEE7WEE7WEE3TuXcpkRdAtIhhG0l2EE7WUYQXsZRtBehhF0Dnp4I2iv
+ lxG018sI2utlBO31MoJuASPo0TCC9jCC9jCC9iBPI+jRMILOwQjawwjawwjawwi6dy7lMiPoFpAMI2gvwwjayzCC9jKMoL0MI+gc9PBG
+ 0F4vI2ivlxG018sI2utlBN0CRtCjYQTtYQTtYQTtQZ5G0KNhBJ2DEbSHEbSHEbSHEXTvXMplRtAtIBlG0F6GEbSXYQTtZRhBexmTiqBZ
+ zPrmJYEEPas7gmZx7/Em6NAF+6kkVMRH+4L9IQQ9URbsR9fJsGB/KEF3uWB/KEF3uWA/XBRC0Et63JZfsH+DImi2xMGIIqUqgmarHbbz
+ CdlmJhSSwYat7CcXQtBsPcQC6jHy76MvB2dms8+QXUTYyQFbhG6eGQLJWLZsWbr91/AImopC1NgFQePUIaREBI0tutj+C5nIJg9QJRtd
+ 0bkLgsbGIaREmVF2lCGIaQt8DVuEbP+FD+PLnqDb6wHIk7oXsv0XdZk6Td0GMW0BB2GLqu2/xGVwGxyXbf8VyRZ5RCdoAedjSxiRbhXY
+ wmbx4sVZZBcTOB2GDHnj03xUoccEhcxmn+q2qMLatWudLUSiMUGEhGy98asAgXWxWStgF2RFxVXA8dFXJBoTyES2KlcV0BWduwA21guo
+ CpQZ+irKjQl8Ddn43jDgw/hyDFIcBHVPXThVoC5Tp/VCiQk4CFtoi7wqwG1wXOwgxiNJ/h/SrklBEPZHsAAAAABJRU5ErkJggg=="/>
+ <rect v:rectContext="foreign" x="0" y="608.42" width="113.836" height="29.0914" class="st2"/>
+ </g>
+ <g id="shape1007-23" v:mID="1007" v:groupContext="shape" transform="translate(289.786,-511.169)">
+ <title>Sheet.1007</title>
+ <desc>Accumulo DataSourceV2</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="56.6929" cy="615.392" width="113.39" height="44.2404"/>
+ <rect x="0" y="593.271" width="113.386" height="44.2404" class="st5"/>
+ <text x="32.51" y="611.79" class="st6" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Accumulo <tspan
+ x="22.11" dy="1.2em" class="st8">DataSourceV2</tspan></text> </g>
+ <g id="shape1008-27" v:mID="1008" v:groupContext="shape" transform="translate(289.786,-446.529)">
+ <title>Sheet.1008</title>
+ <desc>Spark Dataframe</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="56.6929" cy="615.392" width="113.39" height="44.2404"/>
+ <rect x="0" y="593.271" width="113.386" height="44.2404" class="st3"/>
+ <text x="15.71" y="618.99" class="st4" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Spark Dataframe</text> </g>
+ <g id="shape1010-30" v:mID="1010" v:groupContext="shape" v:layerMember="0" transform="translate(122.64,-526.202)">
+ <title>Dynamic connector.1010</title>
+ <path d="M0 630.43 L160.11 630.43" class="st9"/>
+ </g>
+ <g id="shape1011-36" v:mID="1011" v:groupContext="shape" v:layerMember="0" transform="translate(339.393,-511.169)">
+ <title>Dynamic connector.1011</title>
+ <path d="M7.09 637.51 L7.09 650.87" class="st9"/>
+ </g>
+ <g id="shape1012-41" v:mID="1012" v:groupContext="shape" v:layerMember="0" transform="translate(339.393,-446.529)">
+ <title>Dynamic connector.1012</title>
+ <path d="M7.09 637.51 L7.09 650.87" class="st9"/>
+ </g>
+ <g id="shape1014-46" v:mID="1014" v:groupContext="shape" transform="translate(289.786,-317.249)">
+ <title>Sheet.1014</title>
+ <desc>MLeap Bundle</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="56.6929" cy="615.392" width="113.39" height="44.2404"/>
+ <rect x="0" y="593.271" width="113.386" height="44.2404" class="st3"/>
+ <text x="21.59" y="618.99" class="st4" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>MLeap Bundle</text> </g>
+ <g id="shape1020-49" v:mID="1020" v:groupContext="shape" transform="translate(289.786,-235.706)">
+ <title>Sheet.1020</title>
+ <desc>Accumulo DataSourceV2</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="56.6929" cy="615.392" width="113.39" height="44.2404"/>
+ <rect x="0" y="593.271" width="113.386" height="44.2404" class="st5"/>
+ <text x="32.51" y="611.79" class="st6" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Accumulo <tspan
+ x="22.11" dy="1.2em" class="st8">DataSourceV2</tspan></text> </g>
+ <g id="shape1021-53" v:mID="1021" v:groupContext="shape" transform="translate(9.25393,-115.512)">
+ <title>Sheet.1021</title>
+ <desc>AvroIterator</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197" v:verticalAlign="0"/>
+ <v:textRect cx="56.6929" cy="555.295" width="113.39" height="164.434"/>
+ <rect x="0" y="473.077" width="113.386" height="164.434" class="st5"/>
+ <text x="26.52" y="487.88" class="st6" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>AvroIterator</text> </g>
+ <g id="shape1022-56" v:mID="1022" v:groupContext="shape" transform="translate(46.1043,-203.386)">
+ <title>Sheet.1022</title>
+ <desc>Pushdown Filter</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="38.2677" cy="615.392" width="76.54" height="44.2404"/>
+ <rect x="0" y="593.271" width="76.5354" height="44.2404" class="st1"/>
+ <text x="12.76" y="611.79" class="st7" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Pushdown<v:newlineChar/><tspan
+ x="25.67" dy="1.2em" class="st8">Filter</tspan></text> </g>
+ <g id="shape1023-60" v:mID="1023" v:groupContext="shape" transform="translate(46.1043,-159.145)">
+ <title>Sheet.1023</title>
+ <desc>MLeap Execution</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="38.2677" cy="615.392" width="76.54" height="44.2404"/>
+ <rect x="0" y="593.271" width="76.5354" height="44.2404" class="st1"/>
+ <text x="21.6" y="611.79" class="st7" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>MLeap<v:newlineChar/><tspan
+ x="14.36" dy="1.2em" class="st8">Execution</tspan></text> </g>
+ <g id="shape1024-64" v:mID="1024" v:groupContext="shape" transform="translate(46.1043,-115.512)">
+ <title>Sheet.1024</title>
+ <desc>Pushdown Score Filter</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="38.2677" cy="615.392" width="76.54" height="44.2404"/>
+ <rect x="0" y="593.271" width="76.5354" height="44.2404" class="st1"/>
+ <text x="12.76" y="611.79" class="st7" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Pushdown<v:newlineChar/><tspan
+ x="10.78" dy="1.2em" class="st8">Score Filter</tspan></text> </g>
+ <g id="shape1025-68" v:mID="1025" v:groupContext="shape" v:layerMember="0" transform="translate(339.393,-381.889)">
+ <title>Dynamic connector.1025</title>
+ <path d="M7.09 637.51 L7.09 650.87" class="st9"/>
+ </g>
+ <g id="shape1026-73" v:mID="1026" v:groupContext="shape" v:layerMember="0" transform="translate(339.393,-317.249)">
+ <title>Dynamic connector.1026</title>
+ <path d="M7.09 637.51 L7.09 667.77" class="st9"/>
+ </g>
+ <g id="shape1027-78" v:mID="1027" v:groupContext="shape" v:layerMember="0" transform="translate(122.64,-250.739)">
+ <title>Dynamic connector.1027</title>
+ <path d="M0 630.43 L160.11 630.43" class="st9"/>
+ </g>
+ <g id="shape1028-83" v:mID="1028" v:groupContext="shape" transform="translate(287.049,-115.512)">
+ <title>Sheet.1028</title>
+ <desc>Spark Dataframe</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="56.6929" cy="615.392" width="113.39" height="44.2404"/>
+ <rect x="0" y="593.271" width="113.386" height="44.2404" class="st3"/>
+ <text x="15.71" y="618.99" class="st4" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Spark Dataframe</text> </g>
+ <g id="shape1030-86" v:mID="1030" v:groupContext="shape" v:layerMember="0" transform="translate(338.024,-235.706)">
+ <title>Dynamic connector.1030</title>
+ <path d="M7.09 637.51 L7.09 706.43" class="st9"/>
+ </g>
+ <g id="shape1032-91" v:mID="1032" v:groupContext="shape" transform="translate(167.222,-540.551)">
+ <title>Sheet.1032</title>
+ <desc>Training</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="43.2913" cy="627.512" width="86.59" height="20"/>
+ <rect x="0" y="617.512" width="86.5827" height="20" class="st2"/>
+ <text x="23.52" y="631.11" class="st7" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Training</text> </g>
+ <g id="shape1031-94" v:mID="1031" v:groupContext="shape" transform="translate(2.16732,-302.598)">
+ <title>Sheet.1031</title>
+ <path d="M0 637.51 L403.94 637.51 L403.94 371.62 L0 371.62 L0 637.51 Z" class="st11"/>
+ </g>
+ <g id="shape1033-96" v:mID="1033" v:groupContext="shape" transform="translate(165.159,-265.748)">
+ <title>Sheet.1033</title>
+ <desc>Inference</desc>
+ <v:textBlock v:margins="rect(4,4,4,4)" v:tabSpace="42.5197"/>
+ <v:textRect cx="43.2913" cy="627.512" width="86.59" height="20"/>
+ <rect x="0" y="617.512" width="86.5827" height="20" class="st2"/>
+ <text x="20.06" y="631.11" class="st7" v:langID="1033"><v:paragraph v:horizAlign="1"/><v:tabList/>Inference</text> </g>
+ <g id="shape1034-99" v:mID="1034" v:groupContext="shape" transform="translate(2.16732,-111.26)">
+ <title>Sheet.1034</title>
+ <path d="M0 637.51 L403.94 637.51 L403.94 456.09 L0 456.09 L0 637.51 Z" class="st11"/>
+ </g>
+ </g>
+</svg>
diff --git a/images/blog/202002_masc/runtime.png b/images/blog/202002_masc/runtime.png
new file mode 100644
index 0000000..06c0e6f
Binary files /dev/null and b/images/blog/202002_masc/runtime.png differ
diff --git a/images/blog/202002_masc/sparkml_vs_mleap_accumulo.png b/images/blog/202002_masc/sparkml_vs_mleap_accumulo.png
new file mode 100644
index 0000000..2a2dafb
Binary files /dev/null and b/images/blog/202002_masc/sparkml_vs_mleap_accumulo.png differ