Use DETERMINISTIC sampling in PPO #3698

Job	Run time
pytorch_tutorial_build_worker (1, 15, linux.g5.12xlarge.nvidia.gpu)	14m 20s
pytorch_tutorial_build_worker (2, 15, linux.g5.4xlarge.nvidia.gpu)	52m 27s
pytorch_tutorial_build_worker (3, 15, linux.g5.4xlarge.nvidia.gpu)	21m 5s
pytorch_tutorial_build_worker (4, 15, linux.g5.4xlarge.nvidia.gpu)	29m 46s
pytorch_tutorial_build_worker (5, 15, linux.g5.4xlarge.nvidia.gpu)	21m 27s
pytorch_tutorial_build_worker (6, 15, linux.g5.4xlarge.nvidia.gpu)	19m 26s
pytorch_tutorial_build_worker (7, 15, linux.4xlarge.nvidia.gpu)	21m 18s
pytorch_tutorial_build_worker (8, 15, linux.4xlarge.nvidia.gpu)	16m 6s
pytorch_tutorial_build_worker (9, 15, linux.4xlarge.nvidia.gpu)	22m 27s
pytorch_tutorial_build_worker (10, 15, linux.4xlarge.nvidia.gpu)	17m 34s
pytorch_tutorial_build_worker (11, 15, linux.4xlarge.nvidia.gpu)	17m 44s
pytorch_tutorial_build_worker (12, 15, linux.4xlarge.nvidia.gpu)	33m 0s
pytorch_tutorial_build_worker (13, 15, linux.4xlarge.nvidia.gpu)	46m 6s
pytorch_tutorial_build_worker (14, 15, linux.4xlarge.nvidia.gpu)	41m 7s
pytorch_tutorial_build_worker (15, 15, linux.4xlarge.nvidia.gpu)	28m 40s
pytorch_tutorial_build_manager	21m 14s
	7h 3m 47s

Provide feedback