Downloads · 30 days
4
13% of all-time downloads
weikaih/3d-vlm-stage1-v2.2
3d-vlm-stage1-v2.2 is a machine learning model from weikaih. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for trellis. The card lists the license as other.
Qwen3.5-2B extended with 24,576 discrete 3D-code tokens, pretrained to be 3D-literate (predict SLAT VQ codes). Stage-1 representation-shaping checkpoint (v2.2, ckpt-1000).
Downloads · 30 days
4
13% of all-time downloads
All-time downloads
32
Public
Parameters
2.3B
5.7 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors5.6 GB · 100%
From the Hugging Face model README
Qwen3.5-2B extended with 24,576 discrete 3D-code tokens, pretrained to be 3D-literate (predict SLAT VQ codes). Stage-1 representation-shaping checkpoint (v2.2, ckpt-1000).
Deploy = understanding VLM + a Stage-2 flow head reading hidden states (NOT a discrete-blueprint emitter). Discrete codes are training-only. Decode the 3 SLAT VQ tokenizers (weikaih/slat-vq-tokenizers-512) for the diagnostic mesh path. Code: https://github.com/weikaih04/3d_vlm