Fix vllm token

README.md

@@ -73,6 +73,8 @@ See full scores [below](#benchmark-table).
 pip install chandra-ocr
 ```
 
+If you're going to use the huggingface method, we also recommend installing [flash attention](https://github.com/Dao-AILab/flash-attention).
+
 ### From Source
 
 ```bash

chandra/model/hf.py

@@ -74,7 +74,7 @@ def load_model():
         "device_map": device_map,
     }
     if settings.TORCH_ATTN:
-        kwargs["attn_implementation"] = settings.TORCH_ATTN_IMPLEMENTATION
+        kwargs["attn_implementation"] = settings.TORCH_ATTN
 
     model = Qwen3VLForConditionalGeneration.from_pretrained(
         settings.MODEL_CHECKPOINT, **kwargs

chandra/scripts/vllm.py

@@ -17,8 +17,6 @@ def main():
         "-v",
         f"{os.path.expanduser('~')}/.cache/huggingface:/root/.cache/huggingface",
         "--env",
-        f"HUGGING_FACE_HUB_TOKEN={os.getenv('HF_TOKEN')}",
-        "--env",
         "VLLM_ATTENTION_BACKEND=TORCH_SDPA",
         "-p",
         "8000:8000",

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "chandra-ocr"
-version = "0.1.2"
+version = "0.1.4"
 description = "OCR model that converts documents to markdown, HTML, or JSON."
 readme = "README.md"
 requires-python = ">=3.10"