Feature Extraction
Transformers
Safetensors
Chinese
English
qwen3_5
image-text-to-text
multimodal-embedding
text-embedding
image-embedding
video-embedding
mrl
custom_code
Instructions to use tencent/WeMM-Embedding-9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use tencent/WeMM-Embedding-9B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="tencent/WeMM-Embedding-9B", trust_remote_code=True)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("tencent/WeMM-Embedding-9B", trust_remote_code=True) model = AutoModelForMultimodalLM.from_pretrained("tencent/WeMM-Embedding-9B", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
| #!/usr/bin/env python3 | |
| """Align SGLang 0.5.9 video preprocessing with WeMM-Embedding.""" | |
| import importlib.metadata | |
| import py_compile | |
| import shutil | |
| from pathlib import Path | |
| import sglang | |
| EXPECTED_VERSION = "0.5.9" | |
| REPLACEMENTS = ( | |
| ( | |
| "IMAGE_FACTOR = 28", | |
| "IMAGE_FACTOR = 32 # patch_size=16 * spatial_merge_size=2", | |
| ), | |
| ( | |
| " idx = np.linspace(0, total_frames - 1, num=nframes, dtype=np.int64)", | |
| " idx = torch.linspace(0, total_frames - 1, nframes).round().long().cpu().numpy()", | |
| ), | |
| ( | |
| """ video = torchvision.transforms.functional.resize( | |
| video, | |
| [resized_height, resized_width], | |
| interpolation=InterpolationMode.BILINEAR, | |
| ) | |
| """, | |
| """ video = torchvision.transforms.functional.resize( | |
| video, | |
| [resized_height, resized_width], | |
| interpolation=InterpolationMode.BICUBIC, | |
| antialias=True, | |
| ).float() | |
| """, | |
| ), | |
| ) | |
| def main() -> None: | |
| version = importlib.metadata.version("sglang") | |
| if version != EXPECTED_VERSION: | |
| raise RuntimeError(f"Expected sglang=={EXPECTED_VERSION}, found {version}") | |
| path = ( | |
| Path(sglang.__file__).resolve().parent | |
| / "srt" | |
| / "multimodal" | |
| / "processors" | |
| / "qwen_vl.py" | |
| ) | |
| text = path.read_text(encoding="utf-8") | |
| if all(new in text for _, new in REPLACEMENTS): | |
| print("SGLang video preprocessing is ready.") | |
| return | |
| updated = text | |
| for old, new in REPLACEMENTS: | |
| if updated.count(old) != 1: | |
| raise RuntimeError(f"Unexpected SGLang source: {old.splitlines()[0]}") | |
| updated = updated.replace(old, new, 1) | |
| backup = path.with_suffix(path.suffix + ".original") | |
| if not backup.exists(): | |
| shutil.copy2(path, backup) | |
| path.write_text(updated, encoding="utf-8") | |
| py_compile.compile(str(path), doraise=True) | |
| print("SGLang video preprocessing is ready.") | |
| if __name__ == "__main__": | |
| main() | |