I found the main issue. This is because my gpu’s 6gb ram is not enough for most of the large gpu memory allocation.
from HumanoidAMP.yaml decreased
#numEnvs: ${resolve_default:4096,${…num_envs}}
numEnvs: ${resolve_default:200,${…num_envs}}
from HumanoidAMPPPO.yaml decreased
#minibatch_size: 32768
minibatch_size: 32
#amp_minibatch_size: 4096
amp_minibatch_size: 16
now they are working without any crash and error
releted topics :
1)Gym cuda error: running out of memory - #3 by jimingre
2)AssertionError