Instructions to use chibifire/anny-camera-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use chibifire/anny-camera-lora with PEFT:
from peft import PeftModel from transformers import AutoModel base_model = AutoModel.from_pretrained("OmniGen2/OmniGen2") model = PeftModel.from_pretrained(base_model, "chibifire/anny-camera-lora") - Notebooks
- Google Colab
- Kaggle
anny-camera-lora
A camera-control LoRA for OmniGen2, trained on renders of the ANNY body to set the camera azimuth from the prompt.
What it is for
This repo holds a PEFT LoRA adapter for OmniGen2 and its adapter config. The adapter holds only the trained attention tensors, not the base model. It was trained on chibifire/anny-render-corpus to make the camera azimuth in the prompt control the view. The evaluation fits the ANNY body to detected keypoints and compares the recovered azimuth to the requested azimuth. This adapter is undertrained. It learned the three cardinal directions and overfit at other angles. It does not give general camera control.
Provenance
- Base model: OmniGen2/OmniGen2 at revision
df5dca8a981d74e6c3af214c145f5c735fe72367. - Mirror of that base revision: chibifire/omnigen2-base-df5dca8a. Use upstream when you can. The mirror is the fallback.
- Training data: chibifire/anny-render-corpus.
- Training code: weftspun/anny-render-corpus. Citation metadata is in
CITATION.cff.
Use
The adapter does not work without its base. Get both, then load the adapter onto the OmniGen2 transformer with PEFT.
hf download OmniGen2/OmniGen2 --revision df5dca8a981d74e6c3af214c145f5c735fe72367 --local-dir OmniGen2
hf download chibifire/anny-camera-lora --local-dir anny-camera-lora
Licence
Apache 2.0. Terms: https://www.apache.org/licenses/LICENSE-2.0.
- Downloads last month
- 13
Model tree for chibifire/anny-camera-lora
Base model
OmniGen2/OmniGen2