Update README.md
#2
by Ramkasuru - opened
README.md
CHANGED
|
@@ -15,16 +15,28 @@ tags:
|
|
| 15 |
- diffusion
|
| 16 |
---
|
| 17 |
|
| 18 |
-
#
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
## Efficiency
|
| 30 |
|
|
|
|
| 15 |
- diffusion
|
| 16 |
---
|
| 17 |
|
| 18 |
+
# 🚀 First INT4 Quantized Efficient-CubePart3D - 83% Smaller Part Decomposition Model
|
| 19 |
+
|
| 20 |
+
-blueviolet)
|
| 21 |
+
-brightgreen)
|
| 22 |
+
-blue)
|
| 23 |
+
-orange)
|
| 24 |
+
|
| 25 |
+
Presenting the **first INT4 quantized efficient version** of [CubePart](https://huggingface.co/Roblox/cubepart) ([arXiv 2605.28763](https://arxiv.org/abs/2605.28763), SIGGRAPH 2026), an open-vocabulary, part-controllable 3D mesh generative model.
|
| 26 |
+
Quantized via **HQQ (Half-Quadratic Quantization)**, a calibration-free weight-only method, at **W4A16** (4-bit weights, 16-bit activations, group_size=128) using [torchao](https://github.com/pytorch/ao),
|
| 27 |
+
it cuts the diffusion transformer from **8.58 GB → 1.48 GB (82.8%↓)** and peak VRAM from **33.71 GiB → 27.09 GiB (19.6%↓)** while keeping part decomposition near lossless (median chamfer distance **0.0114** per part over 828 parts).
|
| 28 |
+
Companion to [TrNi/efficient-cube3d](https://huggingface.co/TrNi/efficient-cube3d), which does the same for Cube3D.
|
| 29 |
+
|
| 30 |
+
| | BF16 | **INT4 (HQQ W4A16)** |
|
| 31 |
+
|---|:-:|:-:|
|
| 32 |
+
| 💾 Model size (diffusion transformer) | 8.58 GB | **1.48 GB (82.8%↓)** |
|
| 33 |
+
| 🎮 Peak VRAM | 33.71 GiB | **27.09 GiB (19.6%↓)** ✨ |
|
| 34 |
+
| 📦 Setup time | 161.8 s | **166.1 s** (BF16 load + 4.3 s quantize) |
|
| 35 |
+
| ⏱️ Latency | 34.7 s | **66.5 s** (eager mode) |
|
| 36 |
+
| 🎯 Part chamfer distance vs BF16, median | n/a | **0.0114** |
|
| 37 |
+
|
| 38 |
+
<mark>💡 The diffusion transformer is 82.8% smaller on disk and part decomposition stays near lossless on the typical part. Peak VRAM drops less (19.6%) because the Qwen3-VL text encoder and the shape VAE stay in BF16. INT4 is slower here because CubePart needs <code>torch.compile</code> disabled, so dequantization runs in eager mode.
|
| 39 |
+
</mark>
|
| 40 |
|
| 41 |
## Efficiency
|
| 42 |
|