Spaces:
Running
on
Zero
Running
on
Zero
metadata
title: WD EVA02 LoRA ONNX Tagger
emoji: 🖼️
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 4.28.3
app_file: app.py
license: apache-2.0
WD EVA02 LoRA ONNX Tagger
This Space demonstrates image tagging using a fine-tuned WD EVA02 model (converted to ONNX format).
Model Repository: celstk/wd-eva02-lora-onnx
How to Use:
- Upload an image using the upload button.
- Alternatively, paste an image URL into the browser (experimental paste handling).
- Adjust the tag thresholds if needed.
- Choose the output mode (Tags only or include visualization).
- Click the "Predict" button.
Note:
- This Space uses a model from a private repository (
celstk/wd-eva02-lora-onnx
). You might need to duplicate this space and add your Hugging Face token (HF_TOKEN
) to the Space secrets to allow downloading the model files. - Image pasting behavior might vary across browsers.
- If you require GPU acceleration, uncomment the
hardware: cuda-t4-small
line above and ensure the environment has the necessary CUDA libraries compatible withonnxruntime-gpu
. The current setup defaults to CPU due to potential CUDA library mismatches in the standard Spaces environment.