Machine Learning World

Архітектура Twitter Timeline | Fan Out | System Design

Архітектура великих систем завжди була чимось таким величним про що всі мріяли але мало хто бачив. Тож я хочу змінити цем момент і зробити знання більш відкритими, розповідаючи про різні цікаві рішення та підходи. Цього разу ми розберемо архіхтектуру Twitter Timeline (Feed), та спробуємо її трошки зоптимізувати додаючи прості але еффективні алгоритми. ❤️ Допомогти в порятунку та лікуванні безхатніх тварин https://uah.fund/donate ⚓ Підписатися на мій LinkedIn https://www.linkedin.com/in/creotiv/

Інвестиції в контент: Як змінити своє життя та кар'єру

У сучасному світі контент є ключовим елементом комунікації. Відсутність його в вашій стратегії може стати причиною того, що вас просто не помітять. Відтак, не дивно, що багатьом молодим спеціалістам складно знайти роботу. ❤️ Допомогти в порятунку та лікуванні безхатніх тварин https://uah.fund/donate ⚓ Підписатися на мій LinkedIn https://www.linkedin.com/in/creotiv/

GitHub - cvg/LightGlue: LightGlue: Local Feature Matching at Light Speed

LightGlue: Local Feature Matching at Light Speed. Contribute to cvg/LightGlue development by creating an account on GitHub.

What Is ChatGPT Doing … and Why Does It Work?

Stephen Wolfram explores the broader picture of what's going on inside ChatGPT and why it produces meaningful text. Discusses models, training neural nets, embeddings, tokens, transformers, language syntax.

GitHub - facebookresearch/AnimatedDrawings: Code to accompany "A Method for Animating Children's Drawings of the Human Figure"

Code to accompany "A Method for Animating Children's Drawings of the Human Figure" - GitHub - facebookresearch/AnimatedDrawings: Code to accompany "A Method for A...

YOLOv8 native tracking | Step-by-step tutorial | Tracking with Live Webcam Stream

The YOLOv8 team just released (March 9, 2023) native support for object tracking algorithms (ByteTrack and BoT-SORT): https://docs.ultralytics.com/tasks/tracking/ The video will show you how to use the new YOLOv8 object tracking functionality using a live webcam stream. Whether you're an experienced developer or a beginner in the field of computer vision, this video covers everything you need to know to get started with object tracking using YOLOv8. With step-by-step instructions, you'll learn how to use state-of-the-art deep-learning models to track objects in videos and live streams. Chapters: 00:00 Introduction 00:44 Setup Python environment 01:28 Real-time object tracking in the terminal with YOLOv8 CLI 02:17 Bulk video tracking with YOLOv8 SDK 04:56 Setting up inference loop with YOLOv8 SDK 06:11 Bounding box annotation with Supervision 08:18 YOLOv8 tracking ⭐ 10:26 Real-time object counting with Supervision 12:57 Conclusion Resources: ⭐ GitHub repository with demo project: https://github.com/SkalskiP/yolov8-native-tracking 🌏 Roboflow: https://roboflow.com 🌌 Roboflow Universe: https://universe.roboflow.com ⭐ Supervision repository: https://github.com/roboflow/supervision 🎬 Track & Count Objects using YOLOv8 ByteTrack & Supervision: https://youtu.be/OS5qI9YBkfk 🎬 Count People in Zone | 3 Models: YOLOv5, YOLOv8 and Detectron2: https://youtu.be/l_kf9CfZ_8M 🎬 YOLOv8 Object Counting in Real-time with Webcam, OpenCV, and Supervision: https://youtu.be/QV85eYOb7gk 🎬 YOLOv8: How to Train for Object Detection on a Custom Dataset: https://youtu.be/wuZtUMEiKWY 📓 Learn more about YOLOv8 and other Computer Vision models with Roboflow Notebooks: https://github.com/roboflow/notebooks Stay updated with the projects I'm working on at https://github.com/roboflow and https://github.com/SkalskiP! ⭐

GPT in 60 Lines of NumPy | Jay Mody

Implementing a GPT model from scratch in NumPy.

Disney's FRAN neural network can change your age - Candid.Technology

Disney researchers have come up with a new neural network that can de-age actors in-footage while maintain high-quality output.

Golang Developer Tech Lead | Newxel

We are looking for a mature Senior Golang Developer to take part in feature development on the server side.

NVIDIA Broadcast 1.4 Adds Eye Contact and Vignette Effects With Virtual Background Enhancements

Plus new options to mirror your camera and take a selfie.

Let's build GPT: from scratch, in code, spelled out.

We build a Generatively Pretrained Transformer (GPT), following the paper "Attention is All You Need" and OpenAI's GPT-2 / GPT-3. We talk about connections to ChatGPT, which has taken the world by storm. We watch GitHub Copilot, itself a GPT, help us write a GPT (meta :D!) . I recommend people watch the earlier makemore videos to get comfortable with the autoregressive language modeling framework and basics of tensors and PyTorch nn, which we take for granted in this video. Links: - Google colab for the video: https://colab.research.google.com/drive/1JMLa53HDuA-i7ZBmqV7ZnA3c_fvtXnx-?usp=sharing - GitHub repo for the video: https://github.com/karpathy/ng-video-lecture - nanoGPT repo: https://github.com/karpathy/nanoGPT - my website: https://karpathy.ai - my twitter: https://twitter.com/karpathy - our Discord channel: https://discord.gg/3zy8kqD9Cp Supplementary links: - Attention is All You Need paper: https://arxiv.org/abs/1706.03762 - OpenAI GPT-3 paper: https://arxiv.org/abs/2005.14165 - OpenAI ChatGPT blog post: https://openai.com/blog/chatgpt/ Chapters: 00:00:00 intro: ChatGPT, Transformers, nanoGPT, Shakespeare baseline language modeling, code setup 00:07:52 reading and exploring the data 00:09:28 tokenization, train/val split 00:14:27 data loader: batches of chunks of data 00:22:11 simplest baseline: bigram language model, loss, generation 00:34:53 training the bigram model 00:38:00 port our code to a script Building the "self-attention" 00:42:13 version 1: averaging past context with for loops, the weakest form of aggregation 00:47:11 the trick in self-attention: matrix multiply as weighted aggregation 00:51:54 version 2: using matrix multiply 00:54:42 version 3: adding softmax 00:58:26 minor code cleanup 01:00:18 positional encoding 01:02:00 THE CRUX OF THE VIDEO: version 4: self-attention 01:11:38 note 1: attention as communication 01:12:46 note 2: attention has no notion of space, operates over sets 01:13:40 note 3: there is no communication across batch dimension 01:14:14 note 4: encoder blocks vs. decoder blocks 01:15:39 note 5: attention vs. self-attention vs. cross-attention 01:16:56 note 6: "scaled" self-attention. why divide by sqrt(head_size) Building the Transformer 01:19:11 inserting a single self-attention block to our network 01:21:59 multi-headed self-attention 01:24:25 feedforward layers of transformer block 01:26:48 residual connections 01:32:51 layernorm (and its relationship to our previous batchnorm) 01:37:49 scaling up the model! creating a few variables. adding dropout Notes on Transformer 01:42:39 encoder vs. decoder vs. both (?) Transformers 01:46:22 super quick walkthrough of nanoGPT, batched multi-headed self-attention 01:48:53 back to ChatGPT, GPT-3, pretraining vs. finetuning, RLHF 01:54:32 conclusions Corrections: 00:57:00 Oops "tokens from the _future_ cannot communicate", not "past". Sorry! :)

arXiv Xplorer

Semantic search for arXiv.

Check Out TorchOpt: An Efficient Library For Differentiable Optimization Built Upon PyTorch

Differentiable optimization-based algorithms, such as MAML, OptNet, and MGRL, have flourished recently. Meta-gradient, or the gradient term of outer-loop variables obtained by differentiating through the inner-loop optimization process, is one of the crucial components of differentiable optimization. Machine learning models can improve the sampling efficiency and the ultimate performance by utilizing meta-gradients. There are various difficulties in creating differentiable optimization algorithms. Before implementing algorithms with gradient flows on complex computational graphs, developers must realize different inner-loop optimization techniques. Examples include Gumbel-Softmax to differentiate through discrete distribution, function interpolation for differentiable combinatorial solvers, explicit gradient computation of unrolled optimization, implicit gradient

Voxel51 // FiftyOne

We build software that enables ML engineers to build better datasets and models, more quickly. Try FiftyOne, our powerful platform for dataset curation, analysis, and model development.

Zero-shot Object Detection

Zero-shot object detection (ZSD) is the task of object detection where no labeled training data is available for some target object classes.