so http://github.com/karl3wm/httptransformer now has a resuming class
i can load llama 405b and process 0.00002% of it on my tiny system, then reboot the system and start right off again at 0.00002% and process up to 0.00004% ! :D it doesn't do any paralleliz--
