Add torch_cuda deployment: checkpoints + updated manifest (ascend_310p preserved)

Update model card: multi-deployment (torch/onnx) support

Add model card with source model attribution

Upload folder using huggingface_hub

initial commit