---
base_model: Roblox/cube3d-v0.5
base_model_relation: quantized
tags:
  - 3d-generation
  - text-to-3d
  - quantized
  - int4
  - torchao
  - rtn
license: apache-2.0
library_name: torchao
pipeline_tag: text-to-3d
---

# 🚀 First INT4 Quantized Cube3D - Run on Half the VRAM

![](https://img.shields.io/badge/VRAM_-14.3_GB_(44%25↓)-brightgreen)
![](https://img.shields.io/badge/Loading-88%25_Faster-blue)
![](https://img.shields.io/badge/Latency-Same-green)

Presenting the **first INT4 quantized version** of [Cube3D v0.5](https://huggingface.co/Roblox/cube3d-v0.5), a text-to-3D mesh generative model. Quantized via **RTN W4A16** (group_size=128) using [torchao](https://github.com/pytorch/ao), it cuts peak VRAM from **25.4 GB → 14.3 GB (44%↓)** while maintaining the same inference speed and comparable shape fidelity - enabling 3D shape generation on much smaller, more accessible GPUs.

| | BF16 + Engine | BF16 + EngineFast | **INT4 + EngineFast** |
|---|:-:|:-:|:-:|
| 🎮 Peak VRAM | 21.7 GB | 25.4 GB | **14.3 GB (44%↓)** ✨ |
| 📦 Setup time | 19.4 s | 206.9 s | **25.1 s (88%↓)** |
| ⏱️ Latency | 90.9 s | 15.0 s | **14.2 s** |

<mark>💡 The 44% VRAM reduction means this model now fits on a single 16 GB GPU (e.g. NVIDIA L4, A10, A2 etc.), bringing high-quality text-to-3D generation to individual researchers and end-user hardware.
</mark>

### Original BF16 vs Quantized INT4 Comparisons:
##### A. Easy Categories (7)
![Easy categories](assets/easy.gif)
##### B. Medium Categories (6)
![Medium categories](assets/medium.gif)
##### C. Complex Categories (2)
![Complex categories](assets/complex.gif)


# Cube3D v0.5 - RTN W4A16 INT4 (torchao)

Post-training quantized version of [Roblox/cube3d-v0.5](https://huggingface.co/Roblox/cube3d-v0.5), a text-to-3D mesh generative model.  
Quantization method: **RTN W4A16**, group_size=128, via [torchao](https://github.com/pytorch/ao) `int4_weight_only`.

## What's in this repo

| File | Size | Description |
|------|------|-------------|
| `shape_gpt_rtn_int4_g128.pt` | 1.26 GB | INT4 quantized GPT weights (torchao pickle) |
| `shape_tokenizer.safetensors` | ~1.10 GB | VQ-VAE decoder — BF16, unchanged from base model |
| `open_model_v0.5.yaml` | tiny | Model architecture config |
| `quant_config.json` | tiny | Quantization metadata |


## New Benchmarking Dataset (15-categories, 170 prompts)

### Shape Quality (Chamfer Distance, 15 categories, 170 prompts):

Median Chamfer Distance: 67.9 × 10⁻³

Best categories: `vehicle_land` (41.4), `geometric_primitive` (46.5), `animal_wild` (53.8).  
Complex categories: `symmetry_topology` (205.8), `abstract_mathematical` (167.9) - high variance.


| Category | Mean | Std | n |
|---|---:|---:|---:|
**Easy** (CD × 10⁻³ < 75)
| vehicle_land | 41.4 | 21.1 | 10 |
| geometric_primitive | 46.5 | 25.8 | 10 |
| animal_wild | 53.8 | 21.2 | 10 |
| animal_domestic | 56.5 | 21.2 | 10 |
| tool_hardware | 66.7 | 44.7 | 10 |
| furniture | 70.4 | 34.2 | 10 |
| musical_instrument | 72.5 | 45.7 | 10 |
**Medium** (CD × 10⁻³ 75–100)
| vehicle_air_water | 75.3 | 36.1 | 10 |
| fine_detail | 79.2 | 54.8 | 10 |
| visualization_stylized | 85.0 | 46.8 | 30 |
| electronics | 92.2 | 50.1 | 10 |
| architecture | 92.8 | 50.0 | 10 |
| nature_plant | 98.2 | 44.0 | 10 |
**Complex** (CD × 10⁻³ > 100)
| abstract_mathematical | 167.9 | 165.1 | 10 |
| symmetry_topology | 205.8 | 242.7 | 10 |

## Requirements

```
torch==2.10.0+cu128
torchvision==0.25.0+cu128
torchaudio==2.10.0
torchao==0.10.0
```

The .pt file is a torchao pickle, torchao enables kernel-supported INT4 inference.

## Usage
Please see the Google Colab tutorial. 


## Quantization details

- **Method**: Round-to-nearest (RTN)
- **Precision**: W4A16 - weights INT4, activations BF16
- **Quantized INT4 layers**: 279 / 282
- **Skipped layers**: `shape_proj` (in_features=16, < group size), `lm_head` (out=4099, output head), `bbox_proj`
- **Torchao Quantization Group size**: 128


## Citation
```bibtex
@article{roblox2025cube,
  title={Cube: A Roblox View of 3D Intelligence},
  author={Roblox},
  journal={arXiv preprint arXiv:2503.15475},
  year={2025}
}
```