policy-value-methods icon indicating copy to clipboard operation
policy-value-methods copied to clipboard

A3C model doesn't converge!

Open QasimWani opened this issue 3 years ago • 0 comments

fluctuating losses. agnostic to number of parallel agents. checked loss function, everything seems fine when referenced across A3C paper and other repos. shared optimizer looks fine. can't figure out the exact issue.

QasimWani avatar Aug 02 '20 12:08 QasimWani