Upload onnx/model_fp16.onnx with huggingface_hub

Add INT8 quantized version (366MB vs 513MB)

WebGPU v2: no Identity nodes, proper shape inference. Fixes 'unaligned accesses' and 4D/6D shape mismatch.

Fix: strip stale value_info to resolve WebGPU shape inference conflict (4D vs 6D declared shapes)

Upload README.md with huggingface_hub

Delete onnx/model_fp16_cascaded.onnx with huggingface_hub

Upload onnx/model_fp16.onnx with huggingface_hub

Delete onnx/model_uint8.onnx with huggingface_hub

Delete onnx/model_quantized.onnx with huggingface_hub

Delete onnx/model_bnb4.onnx with huggingface_hub

Delete onnx/model_q4.onnx with huggingface_hub

Delete onnx/model_int8.onnx with huggingface_hub

Delete onnx/model_fp16.onnx with huggingface_hub

Remove stale non-cascaded models

Upload README.md with huggingface_hub

Upload onnx/model_fp16_cascaded.onnx with huggingface_hub

Upload onnx/model_fp16.onnx with huggingface_hub

Upload onnx/model_fp16.onnx with huggingface_hub

Upload onnx/model_fp16.onnx with huggingface_hub

Upload onnx/model_bnb4.onnx with huggingface_hub

Upload onnx/model_int8.onnx with huggingface_hub

Upload onnx/model_quantized.onnx with huggingface_hub

Upload onnx/model_uint8.onnx with huggingface_hub

Upload onnx/model_q4.onnx with huggingface_hub

Upload onnx/model.onnx with huggingface_hub

Upload onnx/model_fp16.onnx with huggingface_hub

Upload README.md with huggingface_hub

Upload preprocessor_config.json with huggingface_hub

Upload BiRefNet_config.py with huggingface_hub

Upload config.json with huggingface_hub

initial commit