CHONK your texts with Chonkie - The no-nonsense RAG Chunking library. It’s ultra light-weight, lightening fast and supports all your CHONKing needs! Now with over 700 ⭐️s on Github

The speed of TikToken with the diverse tokenizer support of 🤗 Tokenizers~ Now use any tokenizer on 🤗 Hub with TikToken, with AutoTikTokenizer. And, get 3-6x tokenizer speed boost out of the box!

Pokemon Card Explorer

Don’t know your Pokemon names? Use this modern search engine to find the Pokemon card you’re looking for from 13000+ cards!

Nadir

A Python library of composable deep-learning optimizers, the latest in research! More than 10+ optimizers are available out of the box with only 2-3 lines of code

Visualize your optimizers working with 80+ general convex and non-convex functions! Make gifs, videos and slide to various points too…

First of it’s kind, novel dataset of multilingual Tabular Natural Language Inference (TNLI) consisting tables and hypothesis in 10 languages including low-resource languages like Afrikaans and Arabic as well

JAX/Flax implimentation of 'Attention Is All You Need' by Vaswani et al. (https://arxiv.org/abs/1706.03762)

Hinglish text is commonly used hindi words written in Roman script and converting them back based on sound is a tough modern problem to solve, which this project makes an attempt at.

Using LSTMs and Transformers to translate low-resource indic language Marathi to English and understanding solutions optimized for low-resourse languages