CITATION.cff

cff-version: 1.2.0
message: "If you use this work, please cite both this repository and the HunyuanOCR technical report."
title: "HunyuanOCR-ROCm: an evaluation-backed AMD ROCm port of HunyuanOCR-1.5"
abstract: >-
  Tooling to run Tencent HunyuanOCR-1.5 on AMD gfx1100 (RDNA3) across three
  inference backends (llama.cpp, vLLM, transformers) and score them on
  OmniDocBench v1.6. Evaluation-backed, not precision-aligned.
authors:
  - alias: AIwork4me
type: software
version: "0.1.3"
date-released: "2026-07-18"
license: Apache-2.0
keywords:
  - OCR
  - document-ai
  - AMD-ROCm
  - RDNA3
  - vision-language-model
  - OmniDocBench
repository-code: "https://github.com/AIwork4me/HunyuanOCR-ROCm"
preferred-citation:
  type: article
  title: "HunyuanOCR Technical Report"
  authors:
    - name: "Hunyuan Vision Team"
  year: 2025
  url: "https://arxiv.org/abs/2511.19575"