File “/home/ubuntu/isaac_gym/isaacgym/python/MC_modify_v0.0/rsl_rl/rsl_rl/modules/actor_critic.py”, line 113, in act_student
self.distribution.update(action_mean)
File “/home/ubuntu/isaac_gym/isaacgym/python/MC_modify_v0.0/rsl_rl/rsl_rl/utils/utils.py”, line 206, in update
self.distribution = Normal(logits, self.std.reshape(self.dim))
File “/home/ubuntu/anaconda3/envs/isaac_gym/lib/python3.8/site-packages/torch/distributions/normal.py”, line 56, in init
super().init(batch_shape, validate_args=validate_args)
File “/home/ubuntu/anaconda3/envs/isaac_gym/lib/python3.8/site-packages/torch/distributions/distribution.py”, line 68, in init
raise ValueError(
ValueError: Expected parameter loc (Tensor of shape (24576, 12)) of distribution Normal(loc: torch.Size([24576, 12]), scale: torch.Size([24576, 12])) to satisfy the constraint Real(), but found invalid values:
tensor([[ 0.9591, -1.2083, 1.5889, …, -1.3291, 0.6237, 1.4871],
[-0.4616, -1.0655, 1.6378, …, 1.3159, 2.3434, -0.1360],
[-0.0212, -0.0849, 0.2686, …, 1.4187, 0.6896, 2.6110],
…,
[-0.5143, -2.1863, 2.7050, …, 0.9772, 0.2155, 0.3762],
[-0.7246, 0.5205, 1.0125, …, -0.0216, 1.4684, 2.9036],
[-0.7701, -0.6081, 1.2474, …, -0.1054, 2.1328, 1.6689]],
device=‘cuda:0’, grad_fn=)
I am trying to train quadruped robot.
It seems this situation occurs randomly,and it usually occurs when the number of iterations is around 2000.My obesrvations space includes additional value such as friction,restitution,load and center of mass.
How can I solve this problem？