-
Notifications
You must be signed in to change notification settings - Fork 459
Closed
Description
a quick performance monitoring indicated that the bottleneck was concentrated on 1 of my CPU
by profiling the go2_train.py example i get that the 2 slowest functions are torch.as_tensor and _validate_2D_io_variables()
which seem to be called at every step, is it intended ? thanks
the log :
2131668/1608293 2.661 0.000 3.953 0.000 /home/me/Documents/myenv2/lib/python3.12/site-packages/taichi/lang/kernel_impl.py:1214(_getattr)
2400 0.024 0.000 5.266 0.002 /home/me/Documents/rsl_rl/rsl_rl/algorithms/ppo.py:90(act)
4400 10.881 0.002 13.065 0.003 /home/me/Documents/rsl_rl/rsl_rl/modules/actor_critic.py:119(update_distribution)
4400 0.010 0.000 13.644 0.003 /home/me/Documents/rsl_rl/rsl_rl/modules/actor_critic.py:134(act)
83974 22.945 0.000 22.945 0.000 {method 'item' of 'torch._C.TensorBase' objects}
67959 31.006 0.000 31.006 0.000 {built-in method torch.as_tensor}
13558 0.111 0.000 31.402 0.002 /home/me/Documents/Genesis/genesis/engine/solvers/rigid/rigid_solver_decomp.py:3205(_validate_2D_io_variables)
Metadata
Metadata
Assignees
Labels
No labels