Sync from GitHub via hub-sync

deepseek-ocr-vllm.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",

deepseek-ocr2-vllm.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",

hunyuan-ocr-1.5.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",

lfm2-extract.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "transformers",
 #     "tqdm",

lighton-ocr2.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",
@@ -63,7 +63,7 @@ from datetime import datetime
 
 import torch
 from datasets import load_dataset
-from huggingface_hub import DatasetCard, login
+from huggingface_hub import DatasetCard
 from PIL import Image
 from toolz import partition_all
 from tqdm.auto import tqdm
@@ -335,10 +335,11 @@ def main(
     # Track processing start time
     start_time = datetime.now()
 
-    # Login to HF if token provided
+    # Set the env var instead of login(): the old huggingface_hub in this image
+    # raises KeyError on a Jobs token in login(). Hub calls read HF_TOKEN.
     HF_TOKEN = hf_token or os.environ.get("HF_TOKEN")
     if HF_TOKEN:
-        login(token=HF_TOKEN)
+        os.environ["HF_TOKEN"] = HF_TOKEN
 
     logger.info(f"Using model: {MODEL}")
 

nuextract3.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=3.1.0",
+#     "datasets>=3.1.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "toolz",

ocr-vllm-judge.py

@@ -2,7 +2,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "rich",

olmocr2-vllm.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",

ovis-ocr2.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "toolz",

paddleocr-vl-1.6.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",

qianfan-ocr.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",

surya-ocr-bucket.py

@@ -2,7 +2,7 @@
 # requires-python = ">=3.10"
 # dependencies = [
 #     "surya-ocr==0.20.0",
-#     "datasets>=3.1.0",
+#     "datasets>=3.1.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "imagecodecs",
@@ -1061,10 +1061,10 @@ def main(args: argparse.Namespace) -> None:
     start_time = datetime.now(timezone.utc)
 
     hf_token = args.hf_token or os.environ.get("HF_TOKEN")
+    # Set the env var instead of login(): the old huggingface_hub in this image
+    # raises KeyError on a Jobs token in login(). Hub calls read HF_TOKEN.
     if hf_token:
-        from huggingface_hub import login
-
-        login(token=hf_token)
+        os.environ["HF_TOKEN"] = hf_token
 
     exts = {e.strip().lower() for e in args.extensions.split(",") if e.strip()}
     io_mode = resolve_io_mode(args.io_mode, args.input_source)

surya-ocr.py

@@ -2,7 +2,7 @@
 # requires-python = ">=3.10"
 # dependencies = [
 #     "surya-ocr",
-#     "datasets>=3.1.0",
+#     "datasets>=3.1.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "toolz",
@@ -83,7 +83,7 @@ from typing import Any, Dict, List, Optional, Tuple
 from urllib.request import urlopen
 
 from datasets import load_dataset
-from huggingface_hub import DatasetCard, login
+from huggingface_hub import DatasetCard
 from PIL import Image
 from toolz import partition_all
 from tqdm import tqdm
@@ -529,9 +529,11 @@ def main(
     check_vllm_available()
     start_time = datetime.now(timezone.utc)
 
+    # Set the env var instead of login(): the old huggingface_hub in this image
+    # raises KeyError on a Jobs token in login(). Hub calls read HF_TOKEN.
     HF_TOKEN = hf_token or os.environ.get("HF_TOKEN")
     if HF_TOKEN:
-        login(token=HF_TOKEN)
+        os.environ["HF_TOKEN"] = HF_TOKEN
 
     # Import Surya only after env is set.
     from surya.input.load import load_pdf

unlimited-ocr-vllm.py

@@ -1,7 +1,7 @@
 # /// script
 # requires-python = ">=3.11"
 # dependencies = [
-#     "datasets>=4.0.0",
+#     "datasets>=4.0.0,<5.1",  # 5.1 needs huggingface_hub>=1.31; the image ships an older 1.x on PYTHONPATH, which wins
 #     "huggingface-hub",
 #     "pillow",
 #     "tqdm",
@@ -80,6 +80,10 @@ logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 MODEL = "baidu/Unlimited-OCR"
+# Last revision tested with this recipe and image. The 2026-10-14 upstream update
+# ("compatibility with transformers") makes the remote code import addict and
+# matplotlib at config load, which this image lacks. Re-test before moving it.
+MODEL_REVISION = "07dea832e22aefee32ad281d4b80551282e1c168"
 
 # Prompt and no-repeat-ngram knobs straight from the model card / vLLM recipe (single image).
 PROMPT = "<image>document parsing."
@@ -271,8 +275,14 @@ def main(
     logger.info(f"Initializing vLLM with model: {model}")
     logger.info("This may take a few minutes on first run...")
 
+    # The pin applies to the default model only; another --model loads its main branch.
+    revision = MODEL_REVISION if model == MODEL else None
+    logger.info(f"Model revision: {revision or 'main'}")
     llm = LLM(
         model=model,
+        revision=revision,
+        code_revision=revision,
+        tokenizer_revision=revision,
         trust_remote_code=True,
         max_model_len=max_model_len,
         gpu_memory_utilization=gpu_memory_utilization,