https://vision-array.github.io/A minimal but rich note site on CV and AI. 2026-02-17T23:30:44-08:00 vision-array https://vision-array.github.io/ Jekyll © 2026 vision-array /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png 2026-01-22T04:00:00-08:00 2026-02-17T23:29:31-08:00 https://vision-array.github.io/posts/ray-data-shuffling/ vision-array

Ray data shuffling Shuffling or randomizing the order of the data becomes important for ML workloads specifically for Training. I have tried to visualize some of the data shuffling approaches that Ray offers through these diagrams. This blog and diagrams assume some familiarity with Ray Data. file shuffle shuffles the order of input files loaded to the worker before reading less compute...

2025-09-01T05:00:00-07:00 2025-09-06T16:18:05-07:00 https://vision-array.github.io/posts/object-storage-s3/ vision-array

Object Storage S3

2025-02-22T04:00:00-08:00 2025-09-06T16:18:05-07:00 https://vision-array.github.io/posts/sys-design-intro/ vision-array

system design basics in a hurry basicworking of mobile or web application service any basic mobile or web application needs three things at minimum: server user with the application dns(domain naming service) then for further scaling it to 50 to 100 users and growing, we add a database (one good usecase for database is to store user data) web tier webtier servers are used for se...

2024-12-29T03:00:00-08:00 2024-12-29T03:00:00-08:00 https://vision-array.github.io/posts/yolov3-v5/ vision-array

YOLO v3 and v5 yolov3 Single shot detector i.e the object detection (classificatio + localization/regression) happens within a single network fully convolutional layers - no fully connected layer no max pooling - convolution with stride 2 and residual connections to prevent from gradient problems uses FPN - feature pyramid network by upsampling the feature map(s) in the later part of...

2024-05-11T04:00:00-07:00 2024-05-11T04:00:00-07:00 https://vision-array.github.io/posts/cuda-basics/ vision-array

CUDA basics CUDA Basic Syntax for defining a kernel with a grid size and thread blocks Host Code: Code that runs on CPU Device Code: Code that runs on GPU Grid: collection of all threads launched for a kernel - distributed into thread blocks Thread blocks: a set of threads in a grid grouped together Syntax given for a 32 thread : dim3 block(4,1,1); or dim3 block(4); ( it will be implicit...