Pipeline Parallelism Python Packages

deepspeed

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

colossalai

Making large AI models cheaper, faster and more accessible

colossalai-nightly

Making large AI models cheaper, faster and more accessible

omniback

Serving Inside Pytorch

petals

🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading

torchgpipe

A GPipe implementation in PyTorch

hippo-llm

Local LLM inference + embedding & search in one package. Run 30B on consumer hardware, RAG without ChromaDB.

hami-core

Serving Inside Pytorch

fleet-x

飞桨大模型开发套件，提供大语言模型、跨模态大模型、生物计算大模型等领域的全流程开发工具链。

custom-colossalai

Making large AI models cheaper, faster and more accessible

fleet-lightning

飞桨大模型开发套件，提供大语言模型、跨模态大模型、生物计算大模型等领域的全流程开发工具链。

adeepspeed

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

elf-pipeline

Automated Memory-Aware Pipeline Parallel Training of Neural Networks

test-petals

🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading

torchslicer

Split learning framework for PyTorch — partition a model across devices and train end-to-end without any single node holding the full model.

paddle-fleet

飞桨大模型开发套件，提供大语言模型、跨模态大模型、生物计算大模型等领域的全流程开发工具链。