Surya OCR 2 ONNX FP16 Split Canary
This repository contains a converted/quantized artifact derived from
datalab-to/surya-ocr-2.
What is included
- Source model:
datalab-to/surya-ocr-2
- Runtime/format: ONNX Runtime / WebGPU-oriented experimentation
- Quantization: not quantized; FP16 split vision + language last-logits ONNX canary
- Vision weights included: yes, included as
surya_vision.onnx
- Created for: local OCR/document-understanding experiments and parity testing
Validation status
Export canary diff vs full PyTorch forward: max_abs_diff = 0.0 for the traced sample.
Known caveats
This is a fixed-shape split export canary, not a complete browser OCR app. The full-graph ONNX benchmark runner stalled during ORT session/generation; a split decode harness is still required.
Files
surya_vision.onnx, surya_language_last_logits.onnx, processor assets, and export_config.json.
Usage
Use the vision ONNX and language last-logits ONNX together with a custom decode loop matching export_config.json shapes.
Provenance
This artifact was generated non-destructively from the original Hugging Face checkpoint. It is not a new fine-tune.
If you need production parity, compare against the original model on your own document distribution before deployment.
