TensorFlow Turns 5 – Five Reasons Why It Is The Most Popular ML Framework

This month, Tensorflow, the open source machine learning and deep learning framework from Google, has turned five. 

TensorFlow has become the most popular tool and framework for machine learning in a short span of time. It enjoys tremendous popularity among ML engineers and developers. 

According to the Hacker News Hiring Trends, May 2020, TensorFlow jobs are in great demand. 

Stack Overflow points us to a similar trend where TensorFlow is growing at an extraordinary pace.

Here are five reasons behind TensorFlow’s popularity:

1. The most ubiquitous AI platform available for developers

TensorFlow is the only framework available for running machine learning models from the cloud to the tiniest microcontroller device. Models trained with TensorFlow can be optimized for CPU and GPU. From x86 to ARM64, NVIDIA GPU to Google TPU, the models support diverse architectures.

With TensorFlow Lite, the same models can target mobile phones, IoT devices, and edge computing environments. This makes it possible to train the model once and deploy it to an Android phone, Raspberry Pi, Jetson Nano, EdgeTPU or even an ESP32 microcontroller.

TensorFlow.js is a JavaScript-based framework to run machine learning models within the browser. Any modern browser can run the TensorFlow model with no changes to the code. 

The ubiquity of TensorFlow is one of the reasons why ML practitioners prefer it. 

2. TensorFlow is a part of mainstream public cloud managed ML PaaS 

TensorFlow is an integral part of public cloud platforms. It is powering the APIs for computer vision, natural language processing, personalization and recommendation services. Mainstream machine learning platform as a service (PaaS) offerings have extensive support for TensorFlow. Amazon SageMaker, Azure ML, Google AI Platform and IBM Watson Machine Learning have tight integration of TensorFlow with their platforms. 

At re:Invent 2019, Andy Jassy, CEO of AWS, claimed that 85% of TensorFlow runs on AWS, which indicates its popularity. 

3. Keras + TensorFlow is a powerful combination

For an average ML practitioner, the initial versions of TensorFlow were hard to learn and implement. It came across as a scientific computing toolkit aligned with research projects dealing with extreme parallelism and high-performance computing. 

With TensorFlow 2.0, the toolkit embraced the popular Keras framework, known for its simplicity and intuitive approach. This was expected as François Chollet, the founder of Keras, joined Google just after TensorFlow’s launch. 

The lethal combination of TensorFlow and Keras delivers the power and simplicity for building sophisticated deep learning models. 

4. Extensive support for tooling and integration 

TensorFlow is more than just a machine learning framework or a toolkit. It is essentially a platform to manage the entire lifecycle of AI applications. Its integration with Python IDEs such as PyCharm made it accessible to a large number of developers. Tools such as TensorBoard help engineers get insights into the training process, which helps them in tuning the model. 

NVIDIA and Intel have collaborated with the TensorFlow community to optimize models for their respective processors through TensorRT and OpenVINO Toolkit. 

TensorFlow Serving addresses the deployment and hosting of models. With tight integration with Kubeflow, the Kubernetes ecosystem is taking advantage of the scale of containers for training and inference of TensorFlow models. 

5. Backed by Google’s research and development

Finally, TensorFlow is a critical project for Google. It has invested millions of dollars in research and development to advance machine learning and bring those capabilities to TensorFlow. Google is leveraging TensorFlow for many of its products and services, including Google Assistant, Nest, Android, etc. TensorFlow is one of the key differentiating factors of the Google Cloud AI platform that offers end-to-end capabilities in the form of cognitive services and managed ML PaaS.

TensorFlow played a crucial role in the growth of machine learning and artificial intelligence. Thank you TensorFlow for enabling and empowering developers, and wish you a happy anniversary!

Source link

Be the first to comment

Leave a Reply

Your email address will not be published.