VECTOR FEED
Installation
Requirements
- Python 3.10+
- PyTorch 2.x (CUDA or CPU)
- NVIDIA GPU recommended for VLM and Pipeline backends
(Volta/Ampere/Hopper)
- 16 GB+ RAM for standard Pipeline usage
- Additional VRAM for VLM backend (8 GB+ for 7B models, 24 GB+ for 72B
models)
pip Install
| Backend |
Install command |
| Pipeline (full) |
pip install "vector-feed[pipeline]" |
| VLM (vllm) |
pip install "vector-feed[vlm-vllm]" |
| VLM (lmdeploy) |
pip install "vector-feed[vlm-lmdeploy]" |
| VLM (transformers) |
pip install "vector-feed[vlm-transformers]" |
| All |
pip install "vector-feed[all]" |
Docker
Global (HuggingFace models)
git clone https://github.com/opendatalab/VECTOR-FEED.git
cd VECTOR-FEED
docker compose --profile global up
China (ModelScope mirrors)
docker compose --profile china up
First Run
CLI
vector-feed -p document.pdf
Python
from vector_feed.cli.common import do_parse
result = do_parse(
data=b"..." # PDF bytes
)
Verify Installation
