基于 PyTorch 的深度学习优化库，通过 ZeRO、模型并行和混合专家等技术，让千亿到万亿参数模型的分布式训练与推理更省显存、更高效。
DeepSpeed 官方示例集，提供各类模型接入 DeepSpeed 分布式训练的现成代码，方便照着改自己的训练脚本。
只需几行代码就能把 PyTorch 模型跑在任意设备或分布式配置上，自动处理混合精度（含 fp8）和 FSDP、DeepSpeed 配置，省去手写分布式训练样板代码的麻烦。
基于 Megatron 和 DeepSpeed 在 GPU 上训练大规模自回归 Transformer 的模型并行框架，用来复现和扩展 GPT-3 级别的语言模型。
LMDeploy 是面向大模型的高效推理与服务工具包，支持模型压缩、部署和在线服务，内置 TurboMind 推理引擎和 CUDA 算子，可加速 Llama、InternLM 等模型的推理。
… 0.19.8 release by @loadams in https://github.com/deepspeedai/DeepSpeed/pull/8871 Update to filter out version.txt from PR gates. by @loadams in https://github.com/deepsp
…sion library by @hwchen2017 in https://github.com/deepspeedai/DeepSpeed/pull/8535 Update version.txt after 0.19.7 release by @loadams in https://github.com/deepspeedai/De
…kpoint export by @gaoxiaomo in https://github.com/deepspeedai/DeepSpeed/pull/8318 Raise the per-tensor norms to norm type when combining them by @vineethsaivs in https://
…lelism by @sfc-gh-thonguyen in https://github.com/deepspeedai/DeepSpeed/pull/8226 Return a copy from OnebitLamb.get lamb coeffs by @vineethsaivs in https://github.com/dee
… 0.19.4 release by @loadams in https://github.com/deepspeedai/DeepSpeed/pull/8219 Restore the custom DCO workflow by @tohtana in https://github.com/deepspeedai/DeepSpeed/
With GPT-NeoX 2.0, we now support upstream DeepSpeed. This enables the use of new DeepSpeed features such as [Curriculum Learning](https://www.deepspeed.ai/tutorials/curr