🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.

Kernel sources for https://huggingface.co/kernels-community

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.

Go ahead and axolotl questions

DFlash: Block Diffusion for Flash Speculative Decoding

Train transformer language models with reinforcement learning.

SGLang is a high-performance serving framework for large language models and multimodal models.

The agent that grows with you

Distil SN97 — Competitive Model Distillation on Bittensor

PyTorch native quantization and sparsity for training and inference

Train the smallest LM you can that fits in 16MB. Best model wins!

NanoGPT (124M) in 2 minutes

Load compute kernels from the Hub

Accessible large language models via k-bit quantization for PyTorch.

An interface library for RL post training with environments.

Fast and memory-efficient exact attention

🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support

Memory optimized Mixture of Experts

Ring attention implementation with flash attention

Efficient Triton Kernels for LLM Training

Lighteval is your all-in-one toolkit for evaluating LLMs across multiple backends

Mamba SSM architecture

A simple way of measuring triton kernels

Atropos is a Language Model Reinforcement Learning Environments framework for collecting and evaluating LLM trajectories through diverse environments

KernelBench: Can LLMs Write GPU Kernels? - Benchmark with Torch -> CUDA problems