TheSequence   @TheSequenceAI

Top AI newsletter with over 105k readers from the main AI Labs and enterprises. Founded by @JRdothoughts & @Kseniase_



















 

  Tweets by TheSequence  

TheSequence    @TheSequenceAI    11/30/2021      

Mixture of experts is a deep learning model architecture that makes scaling easier. Recently, @MSFTResearch open-sourced Tutel. It's is a high-performance MoE library to facilitate the development of large-scale DNN models. Find it here: https://t.co/dnmIj9VDfU
  
          6



TheSequence    @TheSequenceAI    11/30/2021      

Congratulations, @huggingface! Could you tell us more about your preparation/the process? https://t.co/sOIw7KlfVK
  
          3



TheSequence    @TheSequenceAI    11/30/2021      

How to get started as an open-source contributor? Join the talk to learn from open-source stars with @hackerrank! https://t.co/ctVbGbe9g5
  
          7



TheSequence    @TheSequenceAI    11/30/2021      

Robust ML observability includes: 1) Model lineage, validation, comparison 2) Data quality monitoring & troubleshooting 3) & more! Take a deeper look at the ML observability: https://t.co/a6IUBnc1Jv
  
          3



TheSequence    @TheSequenceAI    11/30/2021      

Many ML teams will use the terms model monitoring and model observability interchangeably. In reality, there are relevant differences between the two. ML observability features are more interested to understand why something is happening. 1/2
  
          3



TheSequence    @TheSequenceAI    11/30/2021      

Two major schools: 1. Data Parallelism 2. Model Parallelism Both techniques have benefits and limitations. Model parallelism methods are only applicable to certain types of neural nets. Data parallelism has limitations in terms of achieving generalization. 2/2
  
    1         2



TheSequence    @TheSequenceAI    11/30/2021      

Modern deep neural networks are large and require incredibly large training datasets. The traditional sequential approach is simply impractical. But we can use parallel training. The idea of parallelizable training is intuitive but incredibly hard to achieve. 1/2
  
    1         7



TheSequence    @TheSequenceAI    11/29/2021      

Horovod has become one of the go-to frameworks for distributed training. Horovod is open source and hosted by @LFAIDataFdn. The project can be downloaded at https://t.co/3EwbmVYQyX 3/3
  
    1         1



TheSequence    @TheSequenceAI    11/29/2021      

ML companies are hiring: 1. @JinaAI_ Open-source neural search 2. @Verbit_ AI-powered transcription 3. @RescaleInc Intelligent computing platform for digital R&D 4. @DeliverrInc AI-powered e-commerce fulfillment startup 5. @lifevoxel Medical platform for AI and Visualization
  
          4



TheSequence    @TheSequenceAI    11/28/2021      

ML Research of this week: 1. @Walmarttech Deep Learning Demystified 2. @GoogleAI Predictive Text Selection and Federated Learning 3. @mldcmu Safety Envelopes in Robotic Interactions
  
    3         6



TheSequence    @TheSequenceAI    11/28/2021      

Linear Algebra course materials by @zicokolter are open! It would be a great option if you: 1) have some experience with matrices 2) want to know notation & matrix manipulations frequently used in ML/optimization @SCSatCMU @CSDatCMU https://t.co/LWBCeP3GM3
  
    1         3



TheSequence    @TheSequenceAI    11/27/2021      

Linear Algebra course by @MIT is free and open for everyone! Find: 1. Videos and lecture notes 2. Assignments: problem sets with solutions 3. Exams and solutions Learn the basics: https://t.co/WFxMIiBUId
  
    3         9



TheSequence    @TheSequenceAI    11/27/2021      

Features: 1) 200+ custom-built high-quality scenes 2) 2600+ custom-designed household objects across 100+ object types 3) Agent Types 4) 200+ Actions 5) Images 6) Metadata You can use AI2-Thor here: https://t.co/UzssajybZU 3/3
  
    1      



TheSequence    @TheSequenceAI    11/27/2021      

Components of CoreNLP: 1. Pipelines: take a raw text as input and processes it through a series of annotators 2. CoreDocuments: encapsulate the set of annotations produced by Pipelines 3. Annotations: abstracted under a consistent programming model 2/2
  
          3



TheSequence    @TheSequenceAI    11/27/2021      

AI Infrastructure Alliance aligns startups and community members to: 1. Define key components of the AI & ML canonical stack 2. Equalize the competitive landscape 3. Develop ideal best practices and architectures 4. & more! Visit @AiInfra to read about the community!
  
    1         2



TheSequence    @TheSequenceAI    11/27/2021      

That's truly the "app store" for ML apps. There's so much interesting there. You should check it out: https://t.co/QroNMAvmri
  
    1         2