Sync DeepSeek OCR HF job code

ds_batch_ocr/stages.py

@@ -7,7 +7,7 @@ from pathlib import Path
 from typing import Any, Dict, Iterable, List, Optional
 
 import shutil
-from datasets import Dataset, Features, Sequence, Value, load_dataset, Image as HfImage
+from datasets import Dataset, Features, Sequence, Value, load_dataset, Image as HfImage, load_
 from PIL import Image, ImageOps
 from torch.utils.data import DataLoader
 
@@ -162,19 +162,11 @@ def _push_dataset_records(
 
         def _record_iterator() -> Iterable[Dict[str, Any]]:
             for record in records:
-                if isinstance(record, dict):
-                    data = dict(record)
-                else:
-                    data = dict(record)
-
-                if "figures" in data and not isinstance(data["figures"], dict):
-                    data["figures"] = _figures_to_columnar(data.get("figures"))
-
+                data = json.loads(record)
                 yield data
 
-        write_jsonl_iter(dataset_path, _record_iterator())
 
-    dataset = Dataset.from_json(str(dataset_path), features=_dataset_features())
+    dataset = Dataset.from_generator(_record_iterator, features=_dataset_features())
 
     token = env_or_none("HF_TOKEN")
     dataset.push_to_hub(
@@ -454,13 +446,7 @@ def run_stage_extract(settings: ExtractSettings) -> None:
             except Exception as exc:  # pragma: no cover - defensive logging
                 LOGGER.warning("Failed to read documents batch %s: %s", file_path, exc)
                 continue
-
-            if not isinstance(batch_data, list):
-                LOGGER.warning("Unexpected batch content in %s; expected list", file_path)
-                continue
-
-            for entry in batch_data:
-                yield entry
+            yield from batch_data
 
     documents_iter_for_push = iter_documents_from_batches(document_batch_files)
     dataset_records_iter = _build_dataset_records_iter(documents_iter_for_push)