Search Results for "transformer"

ts-blank-space is a fast type-stripping compiler that converts TypeScript to JavaScript. It supports a modern subset of TypeScript by erasing the types and replacing them with whitespace. That's it. It is not a type checker and does not perform any other code transformations. The underlying technique can improve build performance and simplify development. The implementation is pure TypeScript. It is simple enough to read and understand in a few minutes because it is only 700 lines of code...

Transformers4Rec is an advanced recommendation system library that leverages Transformer models for sequential and session-based recommendations. The library works as a bridge between natural language processing (NLP) and recommender systems (RecSys) by integrating with one of the most popular NLP frameworks, Hugging Face Transformers (HF). Transformers4Rec makes state-of-the-art transformer architectures available for RecSys researchers and industry practitioners.

...This objective sidesteps generative pixel losses and avoids heavy negative sampling, producing features that transfer strongly with linear probes and minimal fine-tuning. The design scales naturally with Vision Transformer backbones and flexible masking strategies, and it trains stably at large batch sizes. i-JEPA’s predictions are made in embedding space, which is computationally efficient and better aligned with downstream discrimination tasks. The repository provides training recipes, data pipelines, and evaluation code that clarify which masking patterns and architectural choices matter most.

...Included notebooks examine topics such as MicroGrad-style autodiff, mixture density networks, evolution strategies, floating-point behavior, and KNN versus SVMs. Several notebooks explore minimal character-level recurrent neural networks and transformer-related ideas. The repository also contains Rust experiments and lecture material alongside the notebooks. Most of the code is written as self-contained experiments that can be read and modified independently. Its value is primarily educational, showing small implementations of technical concepts without a large framework around them.

Polysemy is a high-performance, zero-boilerplate effect system for Haskell, designed to simplify the handling of side effects in functional programs. Unlike traditional monad transformer stacks, Polysemy uses a modern approach based on freer monads and interpreters, allowing developers to define, compose, and interpret effects in a more modular and testable way. It aims to offer both flexibility and performance without sacrificing type safety or expressiveness.

Transformer is a TensorFlow implementation of the architecture introduced in the Attention Is All You Need paper. It was created as a readable and relatively modular reference for understanding and experimenting with Transformer-based machine translation. The updated implementation corrects issues involving masking, positional encoding, and other parts of the original code.

...Users needing updated implementations are directed toward related layers in the author's bert4keras project. Its main value today is as a concise historical example of early Transformer-style attention code.