cpu encoder service fails partway through inference on intel mac (and m1 macs)

[rbavery]

not sure why since it is containerized for intel cpus. opened an issue on torchserve repo: https://github.com/pytorch/serve/issues/2258