- Who
- Hugging Face
- What
- publishes a fine-tuning guide scripts, containers, and a walkthrough for fine-tuning TRELLIS.2 with LoRA
- Quote
In this blog, we provide fine-tuning scripts and containers to fine-tune TRELLIS.2, a popular 3D model on your 3D assets, that can run both locally and on Hugging Face Jobs.
What happened
Hugging Face provides scripts and Docker containers for fine-tuning TRELLIS.2 on custom 3D assets locally or with Hugging Face Jobs. In an experiment using 188 footwear objects from Google Scanned Objects, the authors added LoRA adapters to the sparse-structure and shape DiTs while freezing the texture DiT. On nine held-out objects, the adapters improved the mean F-score from 0.276 to 0.803 and reduced Chamfer distance from 0.174 to 0.033.
Why it matters
The guide shows how to target geometry stages while freezing the texture stage, tying those choices to the dataset and reporting results on nine held-out objects.
Read next
See the full storyHugging Face Publishes a TRELLIS.2 Fine-Tuning Guide for 3D FootwearThe story so far
Hugging Face published scripts, Docker containers, and a walkthrough for fine-tuning TRELLIS.2 on custom 3D assets, either locally or with Hugging Face Jobs. In an experiment on 188 footwear objects from Google Scanned Objects, the authors added LoRA adapters to the sparse-structure and shape diffusion transformers while keeping the texture transformer frozen. On nine held-out objects, the adapters raised mean F-score from 0.276 to 0.803 and lowered Chamfer distance from 0.174 to 0.033. The guide requires Hugging Face account access to DINOv3, a gated model. Users supplying their own dataset must upload its meshes to the Hub before starting.