According to the official introduction ,Awkward Array For nesting 、 Data of different sizes , Include Any length A list of 、 Record 、 Mixed types and missing data , It's similar to NumPy.
It looks like an upgraded version NumPy ah .
Sure enough , Arrays of different lengths can be directly put together for operation .
also , Official expression Awkward Array Not only is it easier to use , stay Speed and Memory It also has the advantage of magnitude .
See if it can be arranged ～
https://pypi.org/project/awkward/
I'm sure you're right Jupyter Notebook No stranger .
When you have Jupytext This little plug-in can put Jupyter Notebook and IDE Perfect combination , Does that sound great ！
From now on Jupyter Notebook Can be stored as Markdown File or multilingual Script file .
Jupytext The main things that can be done are ：
Jupyter Notebook Version control
Edit in your favorite text editor 、 Merge or refactor Notebook
stay Notebook Upper use Q&A Check
stay Python Used in ：
This project is in Github Previous 5k+star.
https://github.com/mwouts/jupytext
Than Streamlit also Light weight Of UI Design Library ,Gradio It makes it easy for you to “ Get along well with ” Your model , You can drag and drop pictures directly in the browser , Paste text , Recording sound , wait .
As long as launch() The parameter in the function is set to share=True, And get one Shareable Of website , The friend who gets the link can open it on both the computer and the mobile phone , Living is a Applet .
Always need to do Demo My little friend looks like , This project is in Github Previous 4.5k+star.
https://github.com/gradio-app/gradio
This Hub He is good at data management and data preprocessing .
It can handle Any kind of , Any size The data of , And because the data is stored in the cloud , So it can be seamless in Any machine Visit .
Data compressed into binary bytes can be stored anywhere , And will only be obtained when needed , So there was no TB Class a hard disk can also handle TB Level data .
Hub Provides an important API, Support data in common tools （PyTorch etc. ） Use on , Data versioning , Data conversion and other functions .
This project is in github Previous 4.1k+star.
https://github.com/activeloopai/Hub
AugLy yes facebook The latest data enhancement library , Support at the same time voice , Text , Images and video Data of type , Contains 100 Varied Enhancements .
Data is very important for model training , It is very difficult to label large-scale data . Due to human resources , And model characteristics , Data enhancement is more and more widely used .
AugLy Of advantage ：
The processing type is more comprehensive . Other data enhancement Libraries , for example Albumentations and NVIDIA DALI, Mainly responsible for image related data processing , Text data does not support .
The treatment is very Hommization .AugLy A picture can be made into a memo , In the picture / Superimpose text on video /Emojis, Forward screenshots on social media , It can also help you deal with things like copy detection 、 Hate speech detection or copyright infringement .
This project is in Github Previous 4.1k+star.
https://github.com/facebookresearch/AugLy
Evidently It is a tool used to monitor the effect of the model , Can be obtained from Pandas DataFrame or csv File to generate interactive Visual reporting and JSON Format Of Effect introduction . stay Jupyter Notebook Can be used in .
At present... Is available 6 Kind of The report ： Data drift 、 Numerical target drift 、 Classification target drift 、 Regression model performance 、 Classification model performance and probabilistic classification model performance .
This project is in Github Previous 1.8k+star.
https://github.com/evidentlyai/evidently
If you are familiar with YOLO Words , Then you might be right Open vision Launched this year YOLOX Interested in .
YOLO That's the target detection algorithm , Can be used in cars Autopilot And other cutting-edge technologies .
and YOLOX yes YOLO Anchor free version of , The design is simpler , But the performance is better ！ Its goal is to build a bridge between research and industry , And bridging the gap between the two sides .
This Github The open source project on has been won in just half a year 5.2k+star.
https://github.com/Megvii-BaseDetection/YOLOX
Just like its name ,LightSeq It is a support developed by ByteDance BERT、GPT、Transformer And so on ultrafast Inference engine .
You can see its performance , Than FasterTransformer still more Fast.
LightSeq The supported models are also very comprehensive .
In a word, it's just two words “ To use ”. This project is in Github Previous 1.9k+star.
https://github.com/bytedance/lightseq
Want to predict COVID-19 The speed of recovery ？ Let's have a look LinkedIn It is developed for forecasting demand of its own time series Greykite Well .
Comprehensive function （ Multiple time trends ）, The interface is intuitive , Fast prediction speed and strong scalability are its biggest highlights .
The three algorithms applied to the above ：
Silverkite (Greykite’s flagship algorithm)
Facebook Prophet
Auto Arima
If you are interested, go and study it , This project is in Github Previous 1.4k+star.
https://github.com/linkedin/greykite
Now , In search engines and other applications , Semantic recognition I'm getting higher and higher , Because it can effectively avoid the limitation of word matching .
However, the neural network involved in semantic recognition may make many people feel that their heads are big ,Jina and Finetuner Can help you solve these problems .
Jina Is a neural search framework , So that anyone can A few minutes Build an extensible deep learning search application in .
Finetuner coordination Jina Help you with neural networks Adjustable parameter , To get the best results of the neural search task .
Jina and Finetuner Suitable for people with little experience , Friends who want to try again .
https://github.com/jina-ai/finetuner
Reference link ：
https://tryolabs.com/blog/2021/12/21/top-python-libraries-2021
- END -
contrast Excel The cumulative sales of the series of books reached 15w book , Make it easy for you to master data analysis skills , You can click on the link below to learn about purchasing ：
