{
  "repo_id": "Reza2kn/surya-ocr-2-coreml-runtime",
  "base_model": "datalab-to/surya-ocr-2",
  "base_model_commit": "3b3d4cdf88d6928b0acdc75181b13206ea67c4a3",
  "format": "CoreML MLProgram packages",
  "packages": {
    "vision_fp16": "surya_vision_fp16.mlpackage",
    "vision_int8": "surya_vision_int8.mlpackage",
    "language_prefill_fp16": "surya_prefill_fp16_seq300_cache512.mlpackage",
    "language_decode_step_fp16": "surya_decode_step_fp16_cache512.mlpackage"
  },
  "fixed_shapes": {
    "vision_pixel_values": [
      1024,
      1536
    ],
    "vision_image_embeds": [
      256,
      1024
    ],
    "prefill_sequence_length": 300,
    "decode_step_sequence_length": 1,
    "max_full_attention_cache_length": 512
  },
  "validation_summary": {
    "prefill_coreml_first_token_matches": true,
    "decode_iterative_tokens_match": "9/9",
    "combined_prefill_decode_tokens_match": "9/9",
    "vision_fp16_prefill_decode_tokens_match": "9/9",
    "vision_int8_prefill_decode_tokens_match": "9/9",
    "native_swift_vision_int8_prefill_decode_tokens_match": "9/9",
    "canary_text": "<p>Invoice "
  },
  "host_responsibilities": [
    "provide a CGImage or preprocessed pixel_values",
    "choose stopping criteria",
    "post-process generated OCR HTML"
  ]
}