Downloads · 30 days
0
Qwen2.5-VL-7B-Instruct-W4A8-ConvRot
rekuen/Qwen2.5-VL-7B-Instruct-W4A8-ConvRot
Qwen2.5-VL-7B-Instruct-W4A8-ConvRot is a machine learning model from rekuen. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
A W4A8 quantization of Qwen/Qwen2.5-VL-7B-Instruct for use as a text encoder.
Access
Public
Updated Sep 24, 2026
Repo size
7.2 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors7.2 GB · 100%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- Sep 24, 2026
- Updated
- Sep 24, 2026
- SHA
- 173b6ef7
Base models
Model card
From the Hugging Face model README
Qwen2.5-VL-7B-Instruct — W4A8 ConvRot
A W4A8 quantization of Qwen/Qwen2.5-VL-7B-Instruct for use as a text encoder.
This quantization is unofficial and is not affiliated with the Qwen team or Alibaba Cloud.
Quantization
- Format:
asym_w4a8_int8(int4 weights, int8 activations) with ConvRot rotation. Group size is 16 and rotation group size is 256. - Quantized: the 196 language-model linear projections across all 28 decoder layers.
- Kept in bf16: the full vision tower, token embeddings,
lm_headand all norms.
License
Apache 2.0, the same license as the base model (see LICENSE). The original model is Copyright Alibaba Cloud. The only change is the weight quantization described above.
- License
- apache-2.0
- Created
- Sep 24, 2026
- Updated
- Sep 24, 2026