Advanced Quantization Algorithm for LLMs and VLMs, with support for CPU, Intel GPU, CUDA and HPU.

Causal depthwise conv1d in CUDA, with a PyTorch interface

SYCL* Templates for Linear Algebra (SYCL*TLA) - SYCL based CUTLASS implementation for Intel GPUs

Puzzles for learning Triton

A Python-embedded DSL that makes it easy to write fast, scalable ML kernels with minimal boilerplate.

[ICLR 2025] Palu: Compressing KV-Cache with Low-Rank Projection

Curated collection of papers in MoE model inference

Perplexity GPU Kernels

A subset of PyTorch's neural network modules, written in Python using OpenAI's Triton.

Reproducing R1 for Code with Reliable Rewards

Tile primitives for speedy kernels

s1: Simple test-time scaling

CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image

Meta Lingua: a lean, efficient, and easy-to-hack codebase to research LLMs.

oneAPI - Data Parallel C++ course for students

Fast Matrix Multiplications for Lookup Table-Quantized LLMs

A batched offline inference oriented version of segment-anything

Applied AI experiments and examples for PyTorch

Helpful tools and examples for working with flex-attention

Run PyTorch LLMs locally on servers, desktop and mobile

PyTorch media decoding and encoding

A PyTorch native platform for training generative AI models

A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc.

TorchMultimodal is a PyTorch library for training state-of-the-art multimodal multi-task models at scale.

Train transformer language models with reinforcement learning.

20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.

Go ahead and axolotl questions

PyTorch native post-training library

Implement a ChatGPT-like LLM in PyTorch from scratch, step by step