Qwen3-VL-Embedding-2B — [VLEmb]视觉信息编码v3

English | 简体中文

1. Overview

视觉信息编码模型（~1GB 权重 + 622MB embed 表 + vision）

This repository ships the RK182x deployment artifacts (RKNN graphs + weights) together with the wdnwebui worker manifest (qwen3-vl-embedding-2b.json).

Item	Value
Model ID	wardner/Qwen3-VL-Embedding-2B
Display name	Qwen3-VL-Embedding-2B
Model type	LLM
Compute target	RK182x
Host platform	RK3588
Runtime	RKLLM + RKNN
Capabilities	swcap_vision_embedding
Worker factory	qwen3_vl_embedding
Memory estimate	~2000 MB
Init time	~30 s
Cards required	1
Total weight size	1.60 GB

2. Files

Path	Description	Size
Qwen3-VL-Embedding-2B.embed.bin	fp16 embedding table (vocab × hidden)	593.5 MB
Qwen3-VL-Embedding-2B.tokenizer.gguf	Tokenizer (GGUF, GPT-2 BPE + ByteLevel)	5.7 MB
llm_Qwen3-VL-Embedding-2B.1828.rknn	RKNN graph (RK182x compute card, 1.82 TOPS generation)	15.6 MB
llm_Qwen3-VL-Embedding-2B.weight	Weights paired with the RKNN graph above	787.3 MB
vision_Qwen3-VL-Embedding-2B.1828.rknn	RKNN graph (RK182x compute card, 1.82 TOPS generation)	6.2 MB
vision_Qwen3-VL-Embedding-2B.weight	Weights paired with the RKNN graph above	231.8 MB
qwen3-vl-embedding-2b.json	worker manifest — deliberately plain text	766 B

qwen3-vl-embedding-2b.json is deliberately not in Git LFS. It is the manual of this distribution unit and must stay plain text so it can be diffed, reviewed and previewed on the Hub.

3. Download

hf download wardner/Qwen3-VL-Embedding-2B --local-dir ./Qwen3-VL-Embedding-2B

Or with git (requires Git LFS):

git clone https://huggingface.co/wardner/Qwen3-VL-Embedding-2B

If huggingface.co is unreachable from your network (DNS interception / blocked access), point the client at a mirror instead:

HF_ENDPOINT=https://hf-mirror.com hf download wardner/Qwen3-VL-Embedding-2B \
  --local-dir ./Qwen3-VL-Embedding-2B

Note: with git clone the weights are Git LFS pointers until git lfs pull completes. hf download fetches the real files directly, and does not create a .git directory — neither is required for the model to work; see §4.

4. Installation (wdnwebui)

Clone the repository into the model root so the folder holds both the weights and the worker manifest:

git clone https://huggingface.co/wardner/Qwen3-VL-Embedding-2B /userdata/models/Qwen3-VL-Embedding-2B

Then open the web UI → Compute → the card appears under "local compute" (the backend rescans the model root every ~2 s, no restart needed) → click "enable".

The directory name does not have to match the model name: the SDK follows "the folder the manifest was found in", located through the WDNAI_QWEN3_VL_EMBEDDING_2B_DIR environment variable declared in the manifest.

5. SHA-256 verification

cd ./Qwen3-VL-Embedding-2B
sha256sum -c SHA256SUMS

All entries must report OK before deployment.

6. Caveats and limits

See the notes field of qwen3-vl-embedding-2b.json — it records the per-model specifics (sampling defaults, capability layout, side effects).

7. License

The upstream model is released under the Apache License 2.0. The files here are RKNN-converted derivatives, not unmodified upstream weights. See LICENSE and NOTICE.