LibreQwen3VLEmbedding2b

Qwen3-VL-Embedding-2B weights mirrored for LibreYOLO's embed task (LibreVLM tier).

Source

Mirrored from Qwen/Qwen3-VL-Embedding-2B at commit 9f2f7e710d6d81056aa5c0a4f04764fec6bb7bda. Copyright (c) Tongyi Lab, Alibaba Group. Licensed under the Apache License 2.0.

Modifications

No learned parameters were changed. This repository preserves the Hugging Face snapshot files needed by LibreYOLO's VLM wrapper and replaces the model card with LibreYOLO-specific loading notes.

Usage

from libreyolo import LibreVLM

model = LibreVLM("qwen3-vl-embedding-2b")
pages = model.embed("pages/")  # (N, 2048), one unit row per image
query = model.embed_text(
    ["total due on the invoice"],
    prompt="Retrieve images or text relevant to the user's query.",
)
scores = query @ pages.T

License

Apache License 2.0. See LICENSE and NOTICE.