|
Download OCR_INDRA/README_BUNDLE.md from prosperious/ocr-projects-archive: direct link, hf CLI and curl.
- Browser
- Download file 905 Bytes
-
https://huggingface.co/prosperious/ocr-projects-archive/resolve/main/OCR_INDRA/README_BUNDLE.md
- Command line
-
hf download hf://prosperious/ocr-projects-archive/OCR_INDRA/README_BUNDLE.md
-
curl -L -o README_BUNDLE.md https://huggingface.co/prosperious/ocr-projects-archive/resolve/main/OCR_INDRA/README_BUNDLE.md
905 Bytes
| # OCR_INDRA experiment bundle | |
| Remote run companion for `docs/REMOTE_RUN_SHEET.md` (start there). | |
| Layout: indra/ = cloned indra-ocr v0.3.2 (commit 3d773ae); models/indra-ocr-v3.1.1/ | |
| = Hub artifacts @ v3.1.1 (OpenRAIL-M — fine-tuning explicitly licensed §II ¶36); | |
| tools/indra_eval = verified inspector/eval harness; tools/exp = experiment code; | |
| INDRA_OCR_NEPALI_COMPLETE_REPORT.docx = full investigation report (read-only). | |
| Setup: `bash tools/exp/remote_setup.sh` (deps + probe + offline verification gate), | |
| then follow the run sheet. `bash tools/exp/verify_bundle.sh` alone re-checks the | |
| toolchain with no data: torch re-implementation → ONNX export → exactness vs the | |
| shipped rec.onnx → rec-lines smoke CER. Offline-verified 2026-10-20 on torch | |
| 2.14.1+cpu / onnxruntime 1.31.0 (export uses the legacy torch exporter on purpose). | |
| Never copy datasets off the remote; only results files leave. | |