Downloads · 30 days
7
50% of all-time downloads
toposet-rig-vae-z16
udbbdh/toposet-rig-vae-z16
toposet-rig-vae-z16 is a machine learning model from udbbdh. 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 mit.
Weights only. 16-channel per-token latent VAE over [vertices | joints] tokens of a rigged low-poly mesh (vertex coordinates via a Huber regression head, edges/faces/orientation, bone parents, skin weights; no decoder…
All-time downloads
14
Public
Repo size
984 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pth984 MB · 100%
At a glance
- License
- mit
- Access
- Public
- Created
- Sep 16, 2026
- Updated
- Sep 16, 2026
- SHA
- 25ff3b6d
Model card
From the Hugging Face model README
TopoSet rig VAE, z16 / KL 1e-4 (RigSpacetimeBranchVAE) — intermediate checkpoint
Weights only. 16-channel per-token latent VAE over [vertices | joints] tokens of a rigged low-poly mesh (vertex
coordinates via a Huber regression head, edges/faces/orientation, bone parents, skin weights; no decoder RoPE, decodes
with coordinates unknown). Initialised from the z32 model by PCA-projecting the bottleneck onto the top-16 latent
directions (95.3% of variance), then continued with KL 1e-4 (unified normalisation).
rig_vae_ema_step_30000.pth— EMA (0.999) weights at step 30000 of a 300000-step run; still training, the topology heads are still recovering from the bottleneck cut (see below)config.yaml— training config;models.vae.argsare the constructor arguments (z_dim: 16)
Validation at this checkpoint (FLR held-out val, EMA weights)
| step | edge F1 | face P/R | orientation acc | vertex MAE | bone parent acc | skin L1 |
|---|---|---|---|---|---|---|
| 30000 | 0.965 | 1.000 / 0.9999 | 0.985 | 0.00103 (grid 1/1024 = 0.00098) | 1.000 | 0.437 |
For reference the z32 parent reaches edge F1 1.000, vertex MAE 0.00091, skin L1 0.225. Code is not included.
- License
- mit
- Created
- Sep 16, 2026
- Updated
- Sep 16, 2026