normal expects all elements of std >= 0.0

[mengxi1030]

I’m very glad to see such an excellent open-source project. While reproducing the Go2 locomotion on rough terrain, I encountered the following bug:

Total timesteps: 685768704
Iteration time: 6.69s
Time elapsed: 13:05:38
ETA: 00:26:53

Error executing job with overrides: []
Traceback (most recent call last):
File "/data/fangyanping/conda/envs/env_isaaclab/lib/python3.11/site-packages/isaaclab/source/isaaclab_tasks/isaaclab_tasks/utils/hydra.py", line 101, in hydra_main
func(env_cfg, agent_cfg, *args, **kwargs)
File "/data/fangyanping/workspace/robot/legged_rl_lab/scripts/rsl_rl/train.py", line 242, in main
runner.learn(num_learning_iterations=agent_cfg.max_iterations, init_at_random_ep_len=True)
File "/data/fangyanping/conda/envs/env_isaaclab/lib/python3.11/site-packages/rsl_rl/runners/on_policy_runner.py", line 149, in learn
loss_dict = self.alg.update()
File "/data/fangyanping/conda/envs/env_isaaclab/lib/python3.11/site-packages/rsl_rl/algorithms/ppo.py", line 249, in update
self.policy.act(obs_batch, masks=masks_batch, hidden_states=hid_states_batch[0])
File "/data/fangyanping/conda/envs/env_isaaclab/lib/python3.11/site-packages/rsl_rl/modules/actor_critic.py", line 119, in act
return self.distribution.sample()
File "/data/fangyanping/conda/envs/env_isaaclab/lib/python3.11/site-packages/torch/distributions/normal.py", line 74, in sample
return torch.normal(self.loc.expand(shape), self.scale.expand(shape))
RuntimeError: normal expects all elements of std >= 0.0

The training command I used is:

python scripts/rsl_rl/train.py
--task=LeggedRLLab-Isaac-Velocity-Rough-Unitree-Go2-v0
--num_envs 4096

May I ask if there is a known solution for this issue?