Sync DeepSeek OCR HF job code

ds_batch_ocr/config.py

@@ -105,6 +105,7 @@ class InferenceSettings:
 class ArtifactLocator:
     strategy: str = "local"
     repo_id: Optional[str] = None
+    repo_type: Optional[str] = None
     job_id: Optional[str] = None
     job_owner: Optional[str] = None
     uri: Optional[str] = None
@@ -117,6 +118,7 @@ class ArtifactLocator:
         env = os.environ
 
         repo_id = (env.get(f"{stage}_JOB_REPO") or "").strip() or (env.get(f"{stage}_REPO_ID") or "").strip() or None
+        repo_type = (env.get(f"{stage}_REPO_TYPE") or "").strip() or None
         job_id = (env.get(f"{stage}_JOB_ID") or "").strip() or None
         job_owner = (env.get(f"{stage}_JOB_OWNER") or "").strip() or None
         uri = (env.get(f"{stage}_ARTIFACT_URI") or "").strip() or None
@@ -142,6 +144,7 @@ class ArtifactLocator:
         locator = cls(
             strategy=strategy,
             repo_id=repo_id,
+            repo_type=repo_type,
             job_id=job_id,
             job_owner=job_owner,
             uri=uri,
@@ -154,6 +157,7 @@ class ArtifactLocator:
             {
                 "strategy": locator.strategy,
                 "repo_id": locator.repo_id,
+                "repo_type": locator.repo_type,
                 "job_id": locator.job_id,
                 "job_owner": locator.job_owner,
                 "uri": locator.uri,

ds_batch_ocr/hf_io.py

@@ -25,28 +25,36 @@ def unpack_archives(target_dir: Path) -> None:
         archive.unlink()
 
 
-def download_job_artifact(repo_id: str, target_dir: Path) -> None:
+def download_job_artifact(repo_id: str, target_dir: Path, repo_type: Optional[str] = None) -> None:
     LOGGER.info("Downloading job artifact %s -> %s", repo_id, target_dir)
     actual_repo_id = repo_id
-    repo_type = None
+    normalized_repo_type = (repo_type or "").strip() or None
 
     if repo_id.startswith("jobs/"):
         parts = repo_id.split("/", 2)
         if len(parts) == 3:
-            repo_type = "job"
+            normalized_repo_type = normalized_repo_type or "job"
             actual_repo_id = f"{parts[1]}/{parts[2]}"
         else:
             LOGGER.warning("Unexpected jobs repo format: %s", repo_id)
     elif repo_id.startswith("datasets/"):
-        repo_type = "dataset"
+        normalized_repo_type = normalized_repo_type or "dataset"
         actual_repo_id = repo_id.split("/", 1)[1]
     elif repo_id.startswith("models/"):
-        repo_type = "model"
+        normalized_repo_type = normalized_repo_type or "model"
         actual_repo_id = repo_id.split("/", 1)[1]
 
+    if normalized_repo_type:
+        normalized_repo_type = normalized_repo_type.lower()
+        if normalized_repo_type not in {"dataset", "model", "space", "job"}:
+            LOGGER.warning("Unknown repo_type '%s'; defaulting to dataset", normalized_repo_type)
+            normalized_repo_type = "dataset"
+    else:
+        normalized_repo_type = "dataset"
+
     snapshot_download(
         repo_id=actual_repo_id,
-        repo_type=repo_type,
+        repo_type=normalized_repo_type,
         local_dir=target_dir,
         local_dir_use_symlinks=False,
         ignore_patterns=("logs/**",),
@@ -91,10 +99,14 @@ def resolve_stage_dir(base_dir: Path, locator: ArtifactLocator) -> Path:
 def _handle_hf_hub(locator: ArtifactLocator, base_dir: Path) -> None:
     repo_id = locator.repo_id or locator.uri
     if repo_id:
-        download_job_artifact(repo_id, base_dir)
+        download_job_artifact(repo_id, base_dir, repo_type=locator.repo_type)
         return
     if locator.job_id and locator.job_owner:
-        download_job_artifact(f"jobs/{locator.job_owner}/{locator.job_id}", base_dir)
+        download_job_artifact(
+            f"jobs/{locator.job_owner}/{locator.job_id}",
+            base_dir,
+            repo_type=locator.repo_type or "job",
+        )
         return
     LOGGER.debug("HF locator missing repo/job information; treating as local artifacts.")