Skip to main content

cleanba impala Asteroids-v5

Created on March 26|Last edited on March 29

10M20M30M40Mglobal_step050001000015000200002500030000
exp_name: cleanba_impala_envpool_machado_atari_wrapper Run set
exp_name: cleanba_impala_test2 Run set 3
Run set
10
Run set 3
10
Run set 3



cleanba
3
moolib
3
moolib w/o stuff
cleanba new
6
cleanba w/ last action reward
7
Run set 6
6
original moolib



cleanba
3
moolib
3
moolib w/o stuff
cleanba new
2



cleanba
3
moolib
3
cleanba-new
3
Run set 4
3
Run set 5



cleanba impala (not working)
1
cleanba_impala_envpool_machado_atari_wrapper_a0_l0_d1_n10
3
cleanba_impala_envpool_machado_atari_wrapper_a0_l0_d1_n10_moolib_optimizer
1
moolib
3
smooth_clip
1
no ppo layer init
1
Run set 7
1
moolib no reward normalization
moolib no reward normalization no
--max-grad-norm 40 --ent-coef 0.0006
4
--ent-coef 0.0006
1
--max-grad-norm 40
1
Run set 13
3



cleanba impala (not working)
1
cleanba_impala_envpool_machado_atari_wrapper_a0_l0_d1_n10
3
cleanba_impala_envpool_machado_atari_wrapper_a0_l0_d1_n10_moolib_optimizer
1
moolib
3
smooth_clip
1
no ppo layer init
1
Run set 7
1
moolib no reward normalization
moolib no reward normalization no
--max-grad-norm 40 --ent-coef 0.0006
4
--ent-coef 0.0006
1
--max-grad-norm 40
1
Run set 13
3
--max-grad-norm 40 --ent-coef 0.0006 --local-num-envs 60
10