Skip to content

ATTRIBUTION.md

Latest commit

 

History

History
64 lines (47 loc) · 2.89 KB

ATTRIBUTION.md

File metadata and controls

64 lines (47 loc) · 2.89 KB
Raw

Attribution — NInfer RTX 5090 binaries

The ninfer and ninfer-serve binaries distributed by this project's GitHub Releases are not original work of this project. They are compiled, unmodified, from:
Project Neroued/ninfer
Branch master
Revision feaf4dd0983fdaeb2ba4c06eec6da350e644fb3a
Version string none — upstream publishes no tags or releases
License Apache License 2.0 — see LICENSE

What was and wasn't changed

No source changes were made. The binaries are a stock build of the revision above, configured exactly as upstream's own Dockerfile configures them.

The only difference from what upstream would produce is the build environment, recorded here so the result is reproducible:
Target architecture sm_120a (RTX 5090)
Build image nvidia/cuda:13.1.2-devel-ubuntu24.04
Build host Vast.ai RTX 5090, Ubuntu 24.04, driver 610.43.03
Build time 302 s at -j128
Built 2026-08-20 16:39:34 UTC

Upstream's CMakeLists.txt rejects any value of CMAKE_CUDA_ARCHITECTURES other than 120a and any CUDA compiler older than 13.1, so neither is a choice this project made.

Every tarball also carries a BUILDINFO.txt recording the exact revision, CUDA version, driver version, GPU, job count and wall-clock build time of that specific build.

Why this exists

Upstream ships no binaries for any platform. There are no releases and no tags on the repository, and their README states:

There is no install target or packaged binary distribution; NInfer is run from its source build tree.

Their only packaging path is a Dockerfile you build yourself, which requires a machine that already has an RTX 5090 and a CUDA 13.1 toolchain. This project publishes the output of that build so that renting a 5090 for an hour does not mean spending the first half of it compiling.

Apache-2.0 obligations

Per §4 of the Apache License 2.0, redistribution of these binaries carries the full text of the license (included as LICENSE in this directory and as an asset on every release that contains binaries), retains all attribution notices, and states the origin above. Upstream ships no NOTICE file, so there is none to propagate.

The Apache-2.0 license covers the binaries only. The ninfer orchestration script in this repository is separate work under the MIT License — see the repository root LICENSE.

Model artifacts

The .ninfer checkpoints this CLI downloads are not redistributed by this project; they are pulled directly from their Hugging Face repositories at provision time. They are derived from Qwen/Qwen3.6-27B, Qwen/Qwen3.8-27B and Qwen/Qwen3.6-35B-A3B, all Apache-2.0.