The annual salary of machine learning positions is 500,000+. Now there is an introductory platform that provides free learning and practical operation. Don't you know?

On the AI circle living Bodhisattva, it is none other than Teacher Li Mu.

Previously, I wrote "Hands-on Deep Learning", which was an introductory classic in the circle. Later, I taught Stanford AI courses for free at Station B. A video of a difficult and hard-core dissertation was played 360,000. Many research groups went from tutors to undergraduate apprentices. All my life is chasing fan.

With such great sharing, it’s no wonder many people proudly claim to "have been worshipped by the teacher"——

Recently, I discovered that some platforms have done something more exciting.

Not to mention that this platform provides free computing power, it also puts the introductory classic "hands-on deep learning" into the platform, free for everyone to learn and practice, called: Amazon SageMaker Studio Lab.

According to the introduction, the platform is based on JupyterLab and provides free GPU and CPU computing power +15G permanent storage. It is also linked to GitHub and supports the use of mainstream machine learning tool components and open source resource packs. Developers can combine the "Hands-on Deep Learning" textbook Train the model yourself and see the results.

Moreover, they claim that they only need an email address, no official account, and no credit card.

is there such a good thing?

will take everyone to test it together today.

Can you really practice the "hands-on deep learning" case?

By linking studiolab.sagemaker.aws, we can log in to the web platform to explore the reality.

• can I find "hands-on deep learning"?

As you can see from the interface, the platform provides GPU/CPU computing power options, and can be used directly without payment.

In the lower right corner, we can see the Dive into Deep Learning (abbreviated as D2L).

Click Open D2L notebooks directly to open:

After starting the project, the system automatically loads D2L resources and stores them in our cloud folder.

The README file has also been opened. Here, the system environment configuration, the summary of the book, the audience-oriented, and the catalog framework are all available. There are links to each chapter at the end of the article, from which you can directly enter.

At this point, you can learn deep learning through the platform combined with course content and practical operations——

So what is the actual effect?

• can run anywhere and everywhere code to show you

Take the classic AlexNet part of the classics as an example to get a feel for it.

On the platform, AlexNet's inheritance and development, basic principle explanations are presented, and the model definition and construction code can be run.

In order to ensure a better understanding for beginners, we can also build a single-channel data example to observe the output of the 8 layers in AlexNet. The purpose is to help us intuitively understand the role of different layers:

The most important thing is that the training model link is also open for practical operation, just select the code part to run.

However, the process allows a long time. Let's train for about 7 minutes under the GPU option. Wait slowly!

Seeing that the curve is drawn slowly

It is worth mentioning that, because the entire interface can be used as a notebook to add code to record learning and thinking——

Therefore, even if our course is completed, we can add a code column at the end of the text, and program the homework based on the exercises at the end of the chapter.

• From the basics of mathematics and physics to the actual operation environment configuration, all arrangements are clearly arranged. White

The above display is just a subsection. In fact, from the introduction of concepts such as fully connected layer, convolution, and pooling, to the explanation of ResNet and DenseNet...they are presented and operated in the free environment of Amazon SageMaker Studio Lab, and they are all arranged clearly.

The platform also carefully considers that our high-level linear algebra has different foundations. Not everyone has passed 90 points. It also gives an introduction and code implementation of single-variable calculus, maximum likelihood and other mathematical foundations, and also attaches it. The environment configuration method is a very reliable posture.

After the above verification, this Amazon SageMaker Studio Lab can indeed implement the "hands-on deep learning" of the great god for free and complete——

For those who want to get started or even have a deep grasp of the AI/ML technology, this theory + practice study method is naturally more efficient, and the transition will be more seamless when switching to actual work or scientific research or even entrepreneurship in the future.

In fact, its ability is not only at this level.

A free platform for developers

From the name, you also discovered that the company behind the launch of Amazon SageMaker Studio Lab is Amazon Cloud Technology.

This cutting-edge technology manufacturer launched a free platform this time, not only turning "hands-on deep learning" into a theory + practical exercise field, but also wants to face data scientists, enterprise developers, university teachers and students——

provides a free low-threshold entry machine learning opportunity inclusive of benefits.

In fact, before Amazon Cloud Technology, there were many open machine learning platforms in the industry——

So, what are the new highlights of Amazon SageMaker Studio Lab this time?

Let's start with the configuration first.

The platform provides 15G or more permanent storage, 16G memory, 4 CPUs, and the GPU is NVIDIA Tesla T4, which is slightly higher than other mainstream platforms.

Due to the use of the newer architecture of Nvidia Tesla T4, its mixed-precision computing speed indicators are correspondingly higher. In addition, the free version uses the same architecture as SageMaker Studio, which is equivalent to a layer of enterprise-level Buff, and the stability is more guaranteed.

It is worth noting that the platform advertises 4 hours GPU + 12 hours CPU, but in fact, we can still open Runtime again after the time is up, and the original files still exist.

But if you want to mine coins, forget it...

The platform expressly prohibits the use of SageMaker Studio Lab for production activities, and the direct title is found by mining encrypted currency.

configuration, let's look at the actual operation.

In terms of operability, Amazon SageMaker Studio Lab is more concise and intuitive than other platforms.

The interface can not only create Jupyter Notebook files, but also supports us to directly create Terminal tabs and Markdown format files.

In addition, this platform is equipped with Conda and Pip resource package managers to avoid repeated installation of open source software packages, saving trouble and worry.

You don't even need to use the command line to pull GitHub projects, just click the button on the left.

If there is a yml environment configuration file in the cloned project, when the project is created, the Conda environment will also be established synchronously.

The platform is also linked to Github.

Add the following content to our own Github project README document:

[](https://studiolab.sagemaker.aws/import/github/org/repo/blob/master/path/to/notebook.ipynb)

You can add the "Open in Studio Lab" button——

Others can access the Amazon SageMaker Studio Lab platform with just one click.

Of course, we can also create a new ipynb by uploading or copying manually.

Take the classic image classification algorithm as an example.

After copying an ipynb file from another platform, it can be used without modification. At most, it means manually installing dependencies.

Actual training is basically the same as on other platforms, and sometimes, even slightly faster.

Finally, we have to look at data security issues that many people care about.

We do machine learning and often hold a large amount of private facial information, even hospital patient information that has not been desensitized. In order to protect the privacy and data security of others, we have to look at this aspect.

Amazon SageMaker Studio Lab is born out of an enterprise-level application, and promises that everyone's data will be encrypted and protected, and if the account is deleted, all related data will be deleted accordingly, and the platform party promises not to keep it.

At present, many famous schools and enterprises have used Amazon SageMaker Studio Lab and endorsed their sites.

Among them, there is the School of Engineering of the University of Pennsylvania, where ENIAC was born, the Department of Finance, Santa Clara University, California, and Hugging Face.

There are also many domestic followers.

A doctoral student in the field of machine learning from a 985 Polytechnic University in the south said that although the direction of their research group is traditional machine learning, it still needs deep learning to assist in verification.

Since the laboratory’s computing equipment was purchased several years ago, with the increase in personnel and research directions, especially on the eve of submission, competition for computing resources is common. Amazon SageMaker Studio Lab is indeed attractive to them.

After talking about the advantages of this free platform, the next question is: how to apply?

Let's talk about it here.

No need for an Amazon Cloud Technology account, just log in to the official link studiolab.sagemaker.aws/requestAccount and fill in the email and related information.

However, in order to ensure that everyone can use it through the application as soon as possible, there are some Tips, I hope you will pay attention to:

It is recommended that the language be used in English, and the name of the relevant organization should be clearly filled in, and the suffix of the mailbox should match the English name of the organization. This kind of application is more credible and reliable.

If you meet the above conditions, you can get an invitation within 24 hours of the pro-test. Pay attention to check your email.

Advanced version of seamless migration

As mentioned earlier, Amazon SageMaker Studio Lab and the professional version of Amazon SageMaker Studio have the same architecture, so if you want to migrate from the beginner free version to the professional version, it is certainly not a big deal.

For professional developers, this certainly provides more advanced research and entrepreneurial possibilities.

More specifically, the professional version of Amazon SageMaker Studio has provided developers with a comprehensive set of features from start to finish:

For example, it provides large-scale distributed training for our productive large-scale model training needs. Use partitioning algorithms to automatically split large models and data sets in GPU instances to improve parallelism and speed up training.

For example, the data labeling function Ground Truth Plus has brought in human experts, combined with machine learning to assist pre-labeling, greatly reducing labeling errors and increasing the labeling rate.

Another example is Amazon SageMaker Data Wrangler. This function is oriented to the data preparation stage in machine learning, and data selection, cleaning, and exploration can be performed through a visual interface. You only need one-click import, and you can quickly standardize and convert large quantities of data with a variety of structures without code. Deloitte, one of the "Big Four", adopted this function. Data preparation that was originally completed in a few months is now compressed to a few days.

In addition, Amazon SageMaker Studio also includes access control management, model monitoring, serverless reasoning functions, reasoning configuration recommendations... until the full cycle of industrialized AI/ML services are all rounded up.

Many of the above are newly launched functions of Amazon Cloud Technology re:Invent 2021 this year, which to a large extent demonstrate the company's understanding of needs and the forward-looking technology of this company——

For professional developers and data scientists, whether it is research or entrepreneurship, these features provide more possibilities.

What's more interesting is that Amazon Cloud Technology seems to focus on more than just operating businesses. There are also many inclusive activities that "do not make money" for us.

The machine learning marathon project is a manifestation.

This activity is several times a year, and the platform will present test questions in the application of AI-related fields, covering AI automation programming, disaster prevention or damage assessment, etc.

The activity will test the challenger's skills in CV, NLP, etc. During this period, the relevant platform and resources will also be provided by the platform. The winner will receive a prize of up to $50,000.

There are a lot of activities like this that are geared to actual needs and benefit developers from technology, and each has its own fun or social value.

Amazon DeepRacer, a self-driving racing game with a zero threshold to get started with machine learning, has millions of followers and 140,000 participating developers;

There are also activities to cooperate with the non-profit organization Girls in Tech to help more women understand and learn on mobile devices, and eliminate the gender gap in the technology circle;

There is also the latest release of Amazon SageMaker Canvans, which has attracted high attention in the circle, and is aimed at internal analysts and operators with 0 code experience to help them apply the technology of machine learning in their actual business.

behind the scenes 161c28dbdf1119

Finally, how to evaluate Amazon Cloud Technology's SageMaker Studio Lab?

From a business perspective, these practices are certainly beneficial to the ecological construction of the home, and they are necessary actions for leading companies to maintain their position. In addition, a large number of companies often recruit 500,000 related practitioners with an annual salary of 500,000, which is also a good thing for many developers.

After all, the shortage of artificial intelligence is visible to the naked eye. Most developers have rich programming experience and mathematical foundations. The only barrier is that they are not familiar with machine learning. Make up this piece and drink the soup, why not do it?

However, from the perspective of the industry, the above actions are indeed promoting the landing of cutting-edge technologies——

But the push here is not to be faster, but to be wider.

You know, the automobile was just invented a hundred years ago, and only mechanical experts can become car owners, so that in society at that time, drivers were a profession with cutting-edge technology in their hands.

You must also know that 30 years ago, the PC and the Internet were only toys for a small number of developers, so that the development of a website can quickly make people rich and accumulate assets as high as tall buildings.

Therefore, a hundred years ago, people could not understand a society where everyone could drive. Therefore, 30 years ago, it was also difficult for people to imagine that they could move their fingers and have their own Internet platform.

The same goes for machine learning today. Even the most cutting-edge technology masters can only glimpse a small part of the AI floor plan. Only by repeated dimensionality reduction, this technology can enter the scene of thousands of industries, and it will produce different frequencies of reverberation in the hands of people with different backgrounds and different experiences——

This is not only a manifestation of the value of Amazon's cloud technology inclusive layout, but also the common heart of teachers such as Li Mu.

So, how much energy can AI technology release in the future?

answer has to be found in every individual developer and every scene rift.

For more technical wind directions, long press the picture below to understand:

Unconsciously, by the end of the year, for the technology circle, the development of various technologies and industries during the year has not only stood at the peak, but also experienced ups and downs.

At the last moment of 2021, Amazon Cloud Technology wants to listen to the voices of developers in the cloud computing field. For this reason, the award-winning survey of cloud computing developers has officially opened. We sincerely invite all partners to participate, and multiple gifts are waiting for you!