Ludic – an LLM-RL library for the era of experience

Official Implementation of Dynamic erf (Derf).

Serving multiple LoRA finetuned LLM as one

An Efficient "Factory" to Build Multiple LoRA Adapters

Using GRPO and a modified compositional reward function to train an opensource model on the 1890 Dakota Dictionary

Streamline on-policy/off-policy distillation workflows in a few lines of code

An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models

Throughput-oriented multi-turn inference engine for KernelBench [ICML '25]

Official implementation of "Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding"

Automating analysis from trace files

Advanced quantization toolkit for LLMs and VLMs. Support for WOQ, MXFP4, NVFP4, GGUF, Adaptive Schemes and seamless integration with Transformers, vLLM, SGLang, and llm-compressor

Triton-based Symmetric Memory operators and examples

Provide with pre-build flash-attention package wheels on Linux and Windows platforms using GitHub Actions

Deep learning at the speed of light.

SimKO: Simple Pass@K Policy Optimization

The official implementation of MARS: Unleashing the Power of Variance Reduction for Training Large Models

QuTLASS: CUTLASS-Powered Quantized BLAS for Deep Learning

Supporting code for the blog post on modular manifolds.

Tongyi Deep Research, the Leading Open-source Deep Research Agent

Simplifying reinforcement learning for complex game environments

Efficient non-uniform quantization with GPTQ for GGUF

Fast low-bit matmul kernels in Triton