neural-horner v8 — attributed baseline fork
Unmodified open-source v8 checkpoint by Robby Sneiderman (https://github.com/Robby955/neural-horner, MIT). Submitted to the SAIR Modular Arithmetic Challenge as an attributed baseline; no additional training by this team and no changes to the weights.
The shipped checkpoint is L=2048; the upstream manifest text predates it and still describes the earlier L=1024 artifact. The manifest here has been corrected on that point and on the tier range (1-10), and its training_description records the provenance above. Nothing else differs.
Our own radix-4 descendant of this model is at: https://huggingface.co/Lei285714/mod-arith-k2
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support