Upload README.md with huggingface_hub

Delete model.onnx with huggingface_hub

Delete int8/model.onnx.data with huggingface_hub

Delete int8/model.onnx with huggingface_hub

Delete fp32/model.onnx.data with huggingface_hub

Delete fp32/model.onnx with huggingface_hub

Delete fp16/model.onnx.data with huggingface_hub

Delete fp16/model.onnx with huggingface_hub

feat: add iOS-optimized LAM weights (fp16, 192MB external data)

feat: add iOS-optimized LAM variant (opset 18, native LayerNorm, 608 nodes)

fix: replace fp16 iOS weights with fp32 (383MB, WASM compatible)

fix: replace fp16 iOS variant with fp32 (WASM EP lacks fp16 kernels)

feat: add iOS-optimized LAM weights (fp16, 192MB external data)

feat: add iOS-optimized LAM variant (opset 18, native LayerNorm, 608 nodes)

docs: update config.json with surgical fp16 as recommended, int8 warning

docs: comprehensive model card with surgical fp16 details, int8 findings, platform guide

feat: surgical fp16 (model_fp16.onnx.data) - external data format, decomposed LayerNorm preserved

feat: surgical fp16 (model_fp16.onnx) - external data format, decomposed LayerNorm preserved

Upload model_fp16.onnx with huggingface_hub

Upload fp16/model.onnx.data with huggingface_hub

Upload fp16/model.onnx with huggingface_hub

Upload int8/model.onnx.data with huggingface_hub

Upload int8/model.onnx with huggingface_hub

Upload fp32/model.onnx.data with huggingface_hub

Upload fp32/model.onnx with huggingface_hub

Upload model.onnx with huggingface_hub

Upload config.json with huggingface_hub

Upload README.md with huggingface_hub

initial commit