段错误 (核心已转储)

################################################################################
Learning iteration 110/150000

Computation: 4464 steps/s (collection: 10.770s, learning 0.238s)
Value function loss: 0.0094
Surrogate loss: -0.0121
Vel predict loss: 0.0184
AMP loss: 0.1644
AMP grad pen loss: 0.1755
AMP mean policy pred: -0.6572
AMP mean expert pred: 0.6566
Mean action noise std: 0.56
Mean reward: 12.91
Mean episode length: 786.38
Mean episode rew_action_rate: -0.2167
Mean episode rew_cheat: -0.0300
Mean episode rew_collision: -0.0047
Mean episode rew_dof_acc: -0.1419
Mean episode rew_dof_error: -0.0158
Mean episode rew_feet_air_time: -0.0352
Mean episode rew_feet_edge: 0.0000
Mean episode rew_feet_stumble: 0.0000
Mean episode rew_lin_vel_z: -0.0318
Mean episode rew_stuck: -0.1442
Mean episode rew_torques: -0.0687
Mean episode rew_tracking_ang_vel: 0.2061
Mean episode rew_tracking_lin_vel: 0.9077
Mean episode terrain_level: 0.0815
--------------------------------------------------------------------------------
Total timesteps: 5455872
Iteration time: 11.01s
Total time: 1192.41s
ETA: 1610188.2s

段错误 (核心已转储)
Unable to locate the error location, the software version we are using is as follows
torch 2.3.1
torchaudio 2.3.1
torchvision 0.18.1
isaacgym 1.0rc4
legged_gym 1.0.2

Start by reproducing the training run that reaches iteration 110 with the listed torch, torchaudio, torchvision, Isaac Gym, and legged_gym versions. Investigate the segmentation fault using the reported log, since the issue provides no error location or source file. Done means identifying the failure location and documenting or verifying a reliable fix.

Written by the indexing model from the issue text.