hugpadhugpad

IFM / K2-Horizon-7B-FP8

Text Generation9B81919Updated Oct 15, 2026
Hugging Face

Downloads

819

Likes

19

Last updated

Oct 15, 2026

License

apache-2.0

[!NOTE] This repository contains an FP8-quantized version of IFM/K2-Horizon-7B. All linear layers except lmhead are quantized to FP8: - Weights: static FP8, one scale per 128128 block. - Activations: dynamic FP8, one scale per 1128 group along the input-channel dim. The FP8 m…

Tasks & library

  • text generation
  • transformers

Tags

  • transformers
  • k2_horizon
  • text-generation
  • fp8
  • k2-horizon
  • 7b

Technical specifications

Parameters9B
Librarytransformers
Languagesen
Base modelIFM/K2-Horizon-7B
Licenseapache-2.0