Amend some codes to init training for get up better
This commit is contained in:
@@ -35,7 +35,7 @@ params:
|
||||
normalize_input: True
|
||||
normalize_value: True
|
||||
value_bootstrap: True
|
||||
num_actors: 16384 # 同时训练的机器人数量
|
||||
num_actors: 32768 # 同时训练的机器人数量
|
||||
reward_shaper:
|
||||
scale_value: 1.0
|
||||
normalize_advantage: True
|
||||
@@ -45,7 +45,7 @@ params:
|
||||
lr_schedule: adaptive
|
||||
kl_threshold: 0.013
|
||||
score_to_win: 20000
|
||||
max_epochs: 5000
|
||||
max_epochs: 500000
|
||||
save_best_after: 50
|
||||
save_frequency: 100
|
||||
grad_norm: 1.0
|
||||
|
||||
Reference in New Issue
Block a user