DeepSpeed-MII GitHub repository (org moved from microsoft/)

Fields

id

sources/niche-deepspeed-mii-repo

URL

https://github.com/deepspeedai/DeepSpeed-MII

Publisher

DeepSpeed (formerly Microsoft)

Kind

repo

Accessed

2026-10-03

notes

MII README, served from deepspeedai/DeepSpeed-MII: github.com/microsoft/DeepSpeed-MII returns HTTP 200 after redirect to the deepspeedai org, so old links still resolve. Under the hood MII is DeepSpeed-Inference, which picks an inference path from model architecture, model size, batch size and available hardware, with pre-compiled wheels to avoid long kernel builds. Newest news item in the README is 2024/01 (DeepSpeed-FastGen: Mixtral, Phi-2, Falcon). API metadata 2026-10-03: created 2022-03-23, last push 2025-06-30, v0.3.3 (2025-03-25), Apache-2.0, 2114 stars.

Notes

MII README, served from deepspeedai/DeepSpeed-MII: github.com/microsoft/DeepSpeed-MII returns HTTP 200 after redirect to the deepspeedai org, so old links still resolve. Under the hood MII is DeepSpeed-Inference, which picks an inference path from model architecture, model size, batch size and available hardware, with pre-compiled wheels to avoid long kernel builds. Newest news item in the README is 2024/01 (DeepSpeed-FastGen: Mixtral, Phi-2, Falcon). API metadata 2026-10-03: created 2022-03-23, last push 2025-06-30, v0.3.3 (2025-03-25), Apache-2.0, 2114 stars.

Cross-references

This record references no other records.