fix: regenerate FA FP16 ONNX with fp32-fp16 Cast nodes

Add CK Flash Attention FP16 ONNX model (model_fa_fp16.onnx)

Upload onnx/model_sdpa_fp16.onnx with huggingface_hub

Upload onnx/special_tokens_map.json with huggingface_hub

Upload onnx/tokenizer_config.json with huggingface_hub

Upload onnx/tokenizer.json with huggingface_hub

Add ONNX model: label_mapping.json

Add ONNX model: category_mapping.json

Add ONNX model: config.json

Add ONNX model: model.onnx

Update to rank 32 merged model with supplement data (80% accuracy)

Upload config.json with huggingface_hub

Upload label_mapping.json with huggingface_hub

Upload folder using huggingface_hub

initial commit