tiny res96 recipe, 200-epoch trunk, stage-2 e99

README.md

@@ -0,0 +1,108 @@
+---
+license: mit
+library_name: cod-vae
+pipeline_tag: feature-extraction
+tags:
+- 3d
+- shape-reconstruction
+- autoencoder
+- vae
+- occupancy
+---
+
+# COD-VAE 16 x 4 (tiny)
+
+A tiny, decode-first [COD-VAE](https://arxiv.org/abs/2503.08737) that compresses a
+3D shape into **16 latent vectors of 4 dimensions = 64 numbers** and
+decodes them back into an occupancy field. Same latent shape as
+[cod-vae-16x4](https://huggingface.co/TimSchneider42/cod-vae-16x4) and
+[cod-vae-16x4-small](https://huggingface.co/TimSchneider42/cod-vae-16x4-small),
+but built for pipelines whose wall clock is the decode forward+backward through a
+frozen decoder (e.g. reconstruction-reward RL): ~6.5M parameters, roughly
+**4x faster than `-small` and 33x faster than the full-size model**.
+
+> **Note:** the latent shape matches its siblings, but every model defines its own
+> latent space — latents from one cannot be decoded with another.
+
+Trained with [`cod-vae`](https://github.com/TimSchneider42/cod-vae), a PyTorch/JAX
+reimplementation of COD-VAE (Cho et al., ICCV 2025). The weights are a
+self-contained npz and load with either backend. The JAX decode numbers below
+include the channel-last plane layout (cod-vae >= 56b2c82).
+
+## Architecture vs the -small recipe
+
+| | cod-vae-16x4-small | this model |
+|---|---|---|
+| embed dim / heads | 256 / 4 | 128 / 4 |
+| encoder | 3 blocks x 3 layers, 512 patches, mlp 4 | 2 blocks x 2 layers, 256 patches, mlp 2 |
+| refinement decoder | 6 layers, 16-px patches | 4 layers, 32-px patches |
+| query planes (`query_dim`) | 16 channels at 128² | 8 channels at 96² |
+| latent decoder layers | 12 | 6 |
+| total parameters | ~35M | ~6.5M |
+
+The shipped config pins `attention_implementation="default"` (the XLA path), which
+is measurably faster than cuDNN's fused kernel on these short sequences.
+
+## Decode speed (H100, JAX float16, batch 1024 x 2048 queries, fwd+bwd through the full latent)
+
+`num_latents` and `latent_dim` barely move the decode cost, so these numbers
+(measured on the 16x8 variant) hold for the whole `-tiny` family.
+
+| model | step | throughput |
+|---|---|---|
+| cod-vae-16x8 (full) | ~350 ms | 2.9k shapes/s |
+| cod-vae-16x8-small | 43.5 ms | 23.6k shapes/s |
+| **cod-vae-16x8-tiny** | **8.0 ms** | **127k shapes/s** |
+
+## Held-out reconstruction quality
+
+| source | held-out shapes | volume IoU | near-surface accuracy |
+|---|---|---|---|
+| ABC (CAD parts) | 128 | 0.6854 | 0.7074 |
+
+For reference, cod-vae-16x4-small reaches 0.762 / 0.746 on the same protocol — the extra
+~4x decode speedup costs additional quality. The 16x8 configuration was qualified
+against a hard floor of 0.75 ABC volume IoU before the grid was trained.
+
+## Usage
+
+```python
+import trimesh
+from cod_vae import CODVAE
+
+vae = CODVAE.from_pretrained("TimSchneider42/cod-vae-16x4-tiny")
+
+mesh = trimesh.load("bunny.obj", force="mesh")
+latent, transform = vae.encode_mesh(mesh, return_transform=True)   # (16, 4)
+reconstruction = vae.decode_mesh(latent, transform=transform)      # trimesh.Trimesh
+```
+
+Latents can also be computed from raw surface point clouds and decoded at arbitrary
+query points:
+
+```python
+latents = vae.encode(points)                          # (N, 3) in [-1, 1]^3
+logits = vae.decode(latents, queries)                 # occupancy logits, positive inside
+volume = vae.decode_volume(latents, resolution=128)   # dense logit grid
+```
+
+Install with `pip install cod-vae[torch,hub]` (or `cod-vae[jax,hub]`).
+
+## Training data and recipe
+
+Same 110,077-shape merged dataset and two-stage recipe as the `-small` grid: a
+200-epoch stage-1 trunk per `num_latents` (shared by its row) and a fresh 100-epoch
+stage 2 per cell with 6 latent-decoder layers; see the
+[training guide](https://github.com/TimSchneider42/cod-vae/blob/main/TRAINING.md)
+for the exact commands.
+
+## Citation
+
+```bibtex
+@inproceedings{cho2025cod,
+  author={Cho, In and Yoo, Youngbeom and Jeon, Subin and Kim, Seon Joo},
+  title={Representing 3D Shapes with 64 Latent Vectors for 3D Diffusion Models},
+  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
+  year={2025}
+}
+```