JART Runtime Models
Converted runtime model artifacts for JART.
This repository stores model directories in the layout expected by the JART runtime:
<family>/<variant>/manifest.json
<family>/<variant>/weights.safetensors
The runtime treats manifest.json as the typed model contract and
weights.safetensors as the fp32 runtime weight store. Large weight artifacts
are tracked with Git LFS.
Contents
lama/: LaMa inpainting runtime models.zitspp/: ZITS++ mask inpainting runtime models.EasyOCR/: EasyOCR detection, backbone, and recognition runtime models.paddleocr_v5/: PaddleOCR V5 contracts, prepared fixtures, and runtime safetensors artifacts.paddleocr_vl_1.5/: PaddleOCR-VL 1.5 runtime model, tokenizer files, and manifest.
Usage
Clone with Git LFS enabled:
git lfs install
git clone https://huggingface.co/<namespace>/<repo>
Then point JART commands at one of the contained model directories, for example:
jart mask-inpaint --model JART_models/lama/base --image input.jpg --mask mask.png --output output.png
Notes
These files are converted runtime artifacts for JART, not the original upstream training checkpoints. Upstream model licenses and attribution must be checked against the original source models before publishing or redistributing.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support