Add torch_cpu/torch_cuda deployments: cumulative checkpoint + cum manifest + updated adapter

Upload folder using huggingface_hub