Utils for Unsloth

2-5X faster 80% less memory LLM finetuning

An official repository for GPTailor

Seemless interface of using PyTOrch distributed with Jupyter notebooks

Fork for Token Order Prediction

Flash Dynamic Mask Attention

C++ extensions in PyTorch

AlphaGo Moment for Model Architecture Discovery.

Long-RL: Scaling RL to Long Sequences

The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.

Nano vLLM

CPM.cu is a lightweight, high-performance CUDA implementation for LLMs, optimized for end-device inference and featuring cutting-edge techniques in sparse architecture, speculative sampling and qua…

A PyTorch native platform for training generative AI models

FLA repo fork for Semi-Compressed Attention (SCAN)

🔥 A minimal training framework for scaling FLA models

Training and serving large-scale neural networks with auto parallelization.

A framework for few-shot evaluation of language models.

PCCL (Prime Collective Communications Library) implements fault tolerant collective communications over IP

Tensors and Dynamic neural networks in Python with strong GPU acceleration

Experimental repository for research implementation of NoLoCo.

Bringing BERT into modernity via both architecture changes and scaling