Downloads · 30 days
84
6% of all-time downloads
Qwen3-VL-2B-Instruct-GazeHeads-AllLayers-ONNX
baulab/Qwen3-VL-2B-Instruct-GazeHeads-AllLayers-ONNX
Qwen3-VL-2B-Instruct-GazeHeads-AllLayers-ONNX is a image-text-to-text model from baulab. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for transformers.js. The card lists the license as apache-2.0.
Same as baulab/Qwen3-VL-2B-Instruct-GazeHeads-ONNX but with every transformer layer decomposed, so the gazeheadmask / gazesign steering inputs are honored for all 448 attention heads (28 layers x 16 heads), not just t…
All-time downloads
1.4K
Public
Repo size
1.2 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx_data1.2 GB · 99%
At a glance
- Task
- Image-Text-to-Text
- Library
- transformers.js
- License
- apache-2.0
- Model type
- qwen3_vl
- Access
- Public
- Created
- Jun 16, 2026
- Updated
- Jun 16, 2026
- SHA
- 587c336c
Try a prompt
Base models
Model card
From the Hugging Face model README
Qwen3-VL-2B-Instruct — Gaze-Head Steering, ALL LAYERS (ONNX, WebGPU)
Same as baulab/Qwen3-VL-2B-Instruct-GazeHeads-ONNX
but with every transformer layer decomposed, so the gaze_head_mask /
gaze_sign steering inputs are honored for all 448 attention heads
(28 layers x 16 heads), not just the gaze-hosting layers. Used for the
interactive demo control conditions (steer all heads / random heads).
Two extra decoder inputs implement the paper intervention (zeros = original):
gaze_head_mask(28, 16) float16 — 1.0 selects a head to steergaze_sign(total_seq_len,) float16 — additive pre-softmax attention bias
Inference is slower than the 7-layer build (all layers use decomposed attention instead of fused GroupQueryAttention). Weight data is bit-identical to the upstream export; only the decoder graph differs.
- Task
- Image-Text-to-Text
- Library
- transformers.js
- Type
- qwen3_vl
- License
- apache-2.0
- Created
- Jun 16, 2026
- Updated
- Jun 16, 2026