Tensorflow Extended, ML Metadata and Apache Beam on the Cloud
A practical and self-contained example using GCP Dataflow
The fully end to end example that tensorflow extended provides by running tfx template copy taxi $target-dir
produces 17 files scattered in 5 directories.
If you are looking for a smaller, simpler and self contained example
that actually runs on the cloud and not locally, this is what you are looking for. Cloud services setup is also mentioned here.
What’s going to be covered
We are going to generate statistics and a schema for the Chicago taxi trips csv dataset that you can find by running the tfx template copy taxi
command under the data
directory.
Generated artifacts such as data statistics or the schema are going to be viewed from a jupyter notebook, by connecting to the ML Metadata store or just by downloading artifacts from simple file/binary storage.
Full code sample at the bottom of the article
Services Used
The whole pipeline can run on your local machine ( or on different cloud providers/your custom spark clusters as well). This is an example that can be scaled by using bigger datasets. If you wish to understand how this happens transparently, read this article .
Execution Process
- If running locally, code will not be serialised or sent to the cloud (of course). Otherwise, Beam is going to send everything to a staging location (typically bucket storage). Check out cloudpickle to get some intuition on how serialisation is done.
- Your cloud running service of choice (ours is Dataflow) is going to check if all the mentioned resources exist and are accessible (for example, pipeline output, temporary file storage, etc)
- Compute instances are going to be started and your pipeline is going to be executed in a distributed scenario, showing up in the job inspector while it is still running or finished.
It’s a good naming practise to use /temp
or /tmp
for temporary files and /staging
or /binaries
for the staging directory.
The TFX Pipeline
Tensorflow Extended provides it’s custom component wrappers around plain old beam components. They are a bit more federated in the form: artifacts are only produced and consumed. This means that they do not stream all the dataset everytime, they just pass around resource locator strings. Your dataset gets streamed for analysis preprocessing speed reasons and then saved in small chunks as tfrecords
for maximum performance, taking full advantage of the fast storage technology of Storage Buckets.
This is why when you declare custom components
, you declare strongly typed input and output channels (artifact types and names), which get mapped to multiple, tagged input-outputs on the beam side
. You return these with a Dict
. Feel free to look into the source of the default TFX Components for more insights on these
This is why you need to do things like:
example_gen = CsvExampleGen(...)
statistics_gen = StatisticsGen(examples=example_gen.outputs['examples'])
Related Articles
Request for deletion
About
MC.AI – Aggregated news about artificial intelligence
MC.AI collects interesting articles and news about artificial intelligence and related areas. The contributions come from various open sources and are presented here in a collected form.
The copyrights are held by the original authors, the source is indicated with each contribution.
Contributions which should be deleted from this platform can be reported using the appropriate form (within the contribution).
MC.AI is open for direct submissions, we look forward to your contribution!
Search on MC.AI
mc.ai aggregates articles from different sources - copyright remains at original authors
以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持 码农网
猜你喜欢:本站部分资源来源于网络,本站转载出于传递更多信息之目的,版权归原作者或者来源机构所有,如转载稿涉及版权问题,请联系我们。
Java核心技术·卷 I(原书第10版)
[美] 凯.S.霍斯特曼(Cay S. Horstmann) / 周立新 等 / 机械工业出版社 / 2016-9 / CNY 119.00
Java领域最有影响力和价值的著作之一,由拥有20多年教学与研究经验的资深Java技术专家撰写(获Jolt大奖),与《Java编程思想》齐名,10余年全球畅销不衰,广受好评。第10版根据Java SE 8全面更新,同时修正了第9版中的不足,系统全面讲解了Java语言的核 心概念、语法、重要特性和开发方法,包含大量案例,实践性强。 一直以来,《Java核心技术》都被认为是面向高级程序员的经典教......一起来看看 《Java核心技术·卷 I(原书第10版)》 这本书的介绍吧!