I found a wave of "alchemy artifact" in the goose factory. The developer should pack it quickly

Fish and sheep Xiao Xiao From the Aofei temple
qubits | official account QbitAI

I have a fat friend , Do sth in school Autopilot The kind of .

I often hear his roast , The open source software that poor students use in their laboratories is maozi ：

It looks a little different from industrial ones . As a tool party , I don't believe it , Fat students who engage in autonomous driving have no benefits ？

Drag this fat friend to toss , I really found some good ones .

Now? , There is one “ Student Edition ” Autopilot simulation system , Simulating physical collisions is like this ：

This is the case in sensor simulation feel：

Although it is “ Student Edition ”, But in fact, such a simulation system ,“ Mother ” It is the industrial simulation system ——

It's just the nearest “TechoDay Tencent technology open day ” On , Tencent autopilot simulation system TAD Sim Official release Education .

In addition to being able to be local like the ontology / The cloud covers the whole process of automatic driving test ,“ Education ” It can also provide customized services for colleges and universities ：

stay Tsinghua University In the first version experienced by the Academy of vehicle transportation , The two sides cooperated to develop a learning driving strategy training system , It can realize the simulation scene of random and complex traffic flow , There is also a dedicated interface for reinforcement learning .

It is not only for those who engage in autonomous driving research . We experienced a circle , Found in this developer activity , Tencent has opened up a lot of good things for developers .

What are the details , Let's move on .

“ Alchemy artifact ” And so on ： Lightweight + high efficiency

First look AI aspect .

Now? , Tencent's “ Alchemy artifact ”—— "One-stop" work style AI Develop service platform TI The platform has released Public cloud version .

in other words , This time, the dependence of tools on the environment is further weakened , more Lightweight 、 Open the box 了 .

To be specific , Ease of use aspect ,TI The public cloud version of the platform provides data acquisition during training 、 mark , To various preset algorithm models and frameworks when building , And management MLOps Tools etc. .

performance aspect , Official data show , stay CV、NLP And so on ,TI It can provide algorithm or operation and maintenance engineers with the acceleration capability of various frameworks , The acceleration ratio can reach 200%+.

The key technology behind it , This is the new upgrade TI-ACC Acceleration function .

in addition , Based on Cloud Computing The elastic expansion Characteristics , The platform can adjust the scale according to the required computing resources .

for example , I believe many friends have tried some online Demo when , The model takes several minutes to reason 、 Sometimes the operation fails .

Sometimes this happens for reasons other than network speed , It is also possible that the server will “ It didn't slow down ”, Or the resources cannot be scheduled .

But if it's based on TI Platform construction is similar AI Model , And open the corresponding interface on the Internet , The server will schedule flexibly according to the demand , There will be no waste or waste “ Burst ” The situation of .

Said so much TI Platform capabilities , Is it easy to use ？

We take helping an enterprise develop a commodity identification application as an example .

This AI There are only two requirements for the model ： In addition to accurate product identification , The reasoning speed should not be too slow , Only in this way can it have the value of practical application .

choose Detectron2 frame 、 be based on Python After building the code , Originally, a series of operations such as finding data were required , But in TI It can be solved directly on the platform , Only before and after Half day , You can try it directly .

Compared with developers who build and reason by themselves , This product identification model is in TI The acceleration on the platform is nearly twice as fast .

It is worth mentioning that , This time, TI The public cloud version of the platform is not only for enterprise users , It is also open to individual developers for the first time .

Good things developed on the cloud , It is not limited to AI.

From cloud native to big data , From audio and video to the Internet of things , Today's developers are using the power of tools , Let the overtime -2h（ Manual formation ）.

such as CODING DevOps platform , It can make R & D and operation and maintenance more automated , Make the product release time from “ quarter ”、“ year ” In units of , Shorten to “ God ” perhaps “ Zhou ” In units of .

Orbit Cloud native application management , It can also provide enterprises with centralized cloud native application governance capabilities , Lower the threshold for cloud native application developers , Cloud native application development 、 deliver 、 Operation and maintenance life cycle management .

Tencent cloud also released a newly upgraded fully managed data warehouse product this time ——CDW ClickHouse, Filled in ClickHouse There is no technology gap for subsequent capacity expansion , Can offer low cost 、 High throughput write 、 10 billion millisecond level big data analysis services .

For example Audio and video cloud creation platform , Provide audio and video editing directly in the cloud , By the way, we can provide AI Smart subtitles 、 Intelligent strip removal and other functions , The repetitive work in the clip is further compressed , Natural efficiency up.

△ You can also collaborate directly in the cloud , Cutting with colleagues

There are also some, such as audio and video players sdk（ The same as Tencent video ）、 And Tencent's first IOT device insight analysis product IoT Insight, The digital twin modeling of IOT equipment assets can be realized in a simple process 、 Equipment asset data analysis and other one-stop services .

If you look carefully , You will find that these tools seem to have one thing in common ——

So that you can “ cloud ” Engage in development .

you 're right , In this age of telecommuting , One is connected WiFi Notebook , Developers can work anywhere , Even building AI Model .

How cool it is to simplify and develop ？

actually , In this AI Technology sinks 、 At a point in time when technology iterations are fast , For frontline developers , The first obstacle to development , It boils down to “ resources ” Two words .

In terms of hardware, there is no need to say , Especially for AI In terms of development , Top research institutions / Enterprises can still rely on resources “ Do wonders vigorously ”, Small and medium-sized laboratories 、 The developers of the enterprise have to be careful , Spend every cent of your money on the cutting edge .

Even if the hardware is fixed ,“ Software ” In fact, there are a lot of troubles in .

from 0 To 1 Build new applications , Breaking through the technical threshold in the early stage naturally costs a lot of money , At the later maintenance stage , The cost of time and energy invested should not be underestimated .

That's why , Nowadays, cloud development based on cloud platform will attract more and more attention .

Only with AI Take application development as an example ,IDC Newly released 《2021H2 China AI Cloud Service Market Research Report 》 Show ,2021 In the second half of , China AI The scale of public cloud service market has reached 26.02 RMB 100 million , Occupy AI Of the overall software market 13.4%.

One side , The cloud platform can provide cost-effective hardware resources .

On the other hand , Cloud platforms often release the technical capabilities and experience accumulated by large technology companies for a long time .

With this TechoDay Take the development tools shown on , Its ability to cover cloud primordial 、 big data 、 The Internet of things 、AI、 Audio and video 、 Virtual simulation and other fields , And it is assembled to the cloud in a modular way , Allow developers to “ Onekey ” Low threshold call .

On such a standardized infrastructure , It is possible to unify the development tool chain and develop engineering .

Compared with the traditional development mode , Naturally, it saves developers a lot of tools “ Counterpulsation ” Trouble , Improved development efficiency .

Further more , With the blessing of cloud automation tools , In terms of operation and maintenance at the later stage of the project , Developers can also save a lot of energy .

After all , Reducing costs and increasing efficiency , It's still very fragrant .（ Manual formation ）

that , These new tools on the cloud , You can experience it ？

PS： Click on 【 Read the original 】 Can get 《 Tencent cloud tools Guide ： Lightweight cloud development and cloud applications 》 Oh ~

—  End  —

Focus on me here , Remember to mark the star ～