--- license: mit tags: - rocm - amd - rdna4 - gfx1201 - ocr - runtime --- # Unlimited-OCR RDNA 4 runtime This is documentation for a lightweight runtime companion to [`baidu/Unlimited-OCR`](https://huggingface.co/baidu/Unlimited-OCR), not a hosted pipeline, new checkpoint, fine-tune, or kernel port. No model weights are duplicated here. The linked runtime verifies direct BF16 Transformers inference on one 16 GB AMD Radeon RX 9070 XT (`gfx1201`) with ROCm 7.2.1 and AMD's PyTorch 2.9.1 wheel set. It provides pinned model preparation, code-integrity checks, offline inference, bounded PDF handling, and tracked reproducible validation fixtures. Source and instructions: Please cite and follow the license terms of the original Baidu model.