Use DETERMINISTIC sampling in PPO #3676

Job	Run time
pytorch_tutorial_build_worker (1, 15, linux.g5.12xlarge.nvidia.gpu)	15m 5s
pytorch_tutorial_build_worker (2, 15, linux.g5.4xlarge.nvidia.gpu)	53m 55s
pytorch_tutorial_build_worker (3, 15, linux.g5.4xlarge.nvidia.gpu)	22m 12s
pytorch_tutorial_build_worker (4, 15, linux.g5.4xlarge.nvidia.gpu)	32m 52s
pytorch_tutorial_build_worker (5, 15, linux.g5.4xlarge.nvidia.gpu)	23m 9s
pytorch_tutorial_build_worker (6, 15, linux.g5.4xlarge.nvidia.gpu)	20m 38s
pytorch_tutorial_build_worker (7, 15, linux.4xlarge.nvidia.gpu)	22m 32s
pytorch_tutorial_build_worker (8, 15, linux.4xlarge.nvidia.gpu)	16m 57s
pytorch_tutorial_build_worker (9, 15, linux.4xlarge.nvidia.gpu)	22m 50s
pytorch_tutorial_build_worker (10, 15, linux.4xlarge.nvidia.gpu)	18m 9s
pytorch_tutorial_build_worker (11, 15, linux.4xlarge.nvidia.gpu)	19m 9s
pytorch_tutorial_build_worker (12, 15, linux.4xlarge.nvidia.gpu)	33m 52s
pytorch_tutorial_build_worker (13, 15, linux.4xlarge.nvidia.gpu)	44m 55s
pytorch_tutorial_build_worker (14, 15, linux.4xlarge.nvidia.gpu)	38m 27s
pytorch_tutorial_build_worker (15, 15, linux.4xlarge.nvidia.gpu)	30m 5s
pytorch_tutorial_build_manager	22m 22s
	7h 17m 9s

Provide feedback