Word alignments based on character-level attention maps in Whisper with unsupervised head selection.

The Official PyTorch Implementation of "NVAE: A Deep Hierarchical Variational Autoencoder" (NeurIPS 2020 spotlight paper)

An interactive tool for analyzing, executing, and improving dynamic programming algorithms.

Productive, portable, and performant GPU programming in Python.

Optax is a gradient processing and optimization library for JAX.

A curated list for awesome discrete diffusion models resources.

Embedded segmental K-means (ES-KMeans) in Python.

Integrate Git version control with automatic commit-and-sync and other advanced features in Obsidian.md

MiniLLM is a minimal system for running modern LLMs on consumer-grade GPUs

Optimizing inference proxy for LLMs

Google Research

Python implementation of soft-DTW.

a Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization in pure C.

Apache Spark - A unified analytics engine for large-scale data processing

Python code of the paper Model order reduction of deep structured state-space models: A system-theoretic approach

An official implementation of "BPE-Dropout: Simple and Effective Subword Regularization" algorithm.

Unsupervised text tokenizer for Neural Network-based text generation.

Make huge neural nets fit in memory

A DDSP-based neural voice synthesiser.

Development repository for the Triton language and compiler

SimulEval: A General Evaluation Toolkit for Simultaneous Translation

A library for efficient similarity search and clustering of dense vectors.

Voice Conversion With Just Nearest Neighbors

Paper list of simultaneous translation / streaming translation, including text-to-text machine translation and speech-to-text translation.

Fast and differentiable time domain all-pole filter in PyTorch.

You like pytorch? You like micrograd? You love tinygrad! ❤️

Solve puzzles. Learn CUDA.