Feature Extraction
Transformers
Safetensors
English
Korean
multilingual
qwen3_5
vision-language
embedding
multimodal-embedding
mmeb
custom_code
Instructions to use Urock-AI/eddy-vl_embedding_v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Urock-AI/eddy-vl_embedding_v2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="Urock-AI/eddy-vl_embedding_v2", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("Urock-AI/eddy-vl_embedding_v2", trust_remote_code=True) model = AutoModel.from_pretrained("Urock-AI/eddy-vl_embedding_v2", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Commit History
Yesterday
-
State the compute reduction directly
Verified signature7d69a16 -
Clarify model card wording and keep the embedding token visible
Verified signature45f0ed5
5 days ago
-
Describe parameter bars as weights used for embedding
Verified signature6e189b7 -
Note that lm_head is unused for embedding vectors
Verified signature9429e90 -
Describe the pruned stack as restored by distillation
Verified signaturee1be45d -
Clarify that lm_head stays for serving
Verified signatureb341d45 -
Explain why lm_head stays in the checkpoint
Verified signaturea302945
6 days ago
-
Remove the ambiguous WeMM identity sentence
Verified signature5288ffc -
Set WeMM-Embedding-2B as the only base model
Verified signature045bcb5 -
Point SGLang 0.5.9 users at the WeMM serving steps
Verified signaturea7c8b94 -
Remove SGLang 0.5.9 video patch
Verified signature247e464 -
Show a complete embedding example
Verified signature296bf06 -
Record transformers 5.2.0, the version that loads this model
Verified signatureec69ded -
Drop the WeMM arXiv link from the model card
Verified signature75bf488 -
Use the model name as the citation title
Verified signature44449f9 -
Name WeMM-Embedding-2B instead of teacher
Verified signature77fa5ab -
Add a short v2 fit summary above the scores
Verified signaturefeffa72 -
Add a v2 lead paragraph without the v1 technical report
Verified signature55b8b2f -
Add in-house Qwen and Eddy task scores
Verified signaturec9b3127 -
Compare parameter counts across four embedding models
Verified signature7b6d2c3 -
Drop the exact parameter and FLOP tables
Verified signature4399b65 -
Show parameter bars without lm_head
Verified signaturee959750 -
Drop the kept-versus-teacher bar chart
Verified signature16837a4 -
Add selected in-house MMEB-v2 task scores
Verified signature4878340 -
Keep score comparison on the in-house run only
Verified signaturee6e30a4 -
Show kept-score bars on a 0 to 100 scale
Verified signature2ecf896 -
Add eddy-vl embedding v2
Verified signature2586ee6 -
initial commit
Verified signature0a2501d