Hi, I encountered the following error when I run the guided policy search algorithm:
ValueError: Failed to find PD solution even for very large eta (check that dynamics and cost are reasonably well conditioned)!
Is there any solution?
P.S.
The above error occurred the newest version of GPS.
A few months ago, I run the same algorithm altering iterations 12 to 30 in hyperparams.py:
algorithm = {
・・・
'iterations': 12,
'iterations': 30,
・・・
}
Then, the trajectory of the ILQG (Trajectory Samples) became very different from that of the Neural Networks (Policy Samples). At this time, I encountered the following warning:
"Final KL divergence after DGD convergence is too high."
Hi, I encountered the following error when I run the guided policy search algorithm:
ValueError: Failed to find PD solution even for very large eta (check that dynamics and cost are reasonably well conditioned)!
Is there any solution?
P.S. The above error occurred the newest version of GPS. A few months ago, I run the same algorithm altering iterations 12 to 30 in hyperparams.py:
algorithm = { ・・・
'iterations': 12,
}
Then, the trajectory of the ILQG (Trajectory Samples) became very different from that of the Neural Networks (Policy Samples). At this time, I encountered the following warning: "Final KL divergence after DGD convergence is too high."