Commit 3fd8a42e by yzh

second_amp

parent c3b1f4f9
...@@ -333,7 +333,8 @@ class Elf12DHStandCfg(LeggedRobotCfg): ...@@ -333,7 +333,8 @@ class Elf12DHStandCfg(LeggedRobotCfg):
max_contact_force = 700 # forces above this value are penalized max_contact_force = 700 # forces above this value are penalized
class scales: class scales:
ref_joint_pos = 2.2 # ref_joint_pos = 0
# ref_joint_pos = 2.2
feet_clearance = 1. feet_clearance = 1.
feet_contact_number = 2.0 feet_contact_number = 2.0
# gait # gait
...@@ -401,7 +402,7 @@ class Elf12DHStandCfgPPO(LeggedRobotCfgPPO): ...@@ -401,7 +402,7 @@ class Elf12DHStandCfgPPO(LeggedRobotCfgPPO):
class algorithm(LeggedRobotCfgPPO.algorithm): class algorithm(LeggedRobotCfgPPO.algorithm):
entropy_coef = 0.001 entropy_coef = 0.001
learning_rate = 1e-4 learning_rate = 1e-4
num_learning_epochs = 4 num_learning_epochs = 3
gamma = 0.994 gamma = 0.994
lam = 0.9 lam = 0.9
num_mini_batches = 4 num_mini_batches = 4
...@@ -425,11 +426,11 @@ class Elf12DHStandCfgPPO(LeggedRobotCfgPPO): ...@@ -425,11 +426,11 @@ class Elf12DHStandCfgPPO(LeggedRobotCfgPPO):
load_run = -1 # -1 = last run load_run = -1 # -1 = last run
checkpoint = -1 # -1 = last saved model checkpoint = -1 # -1 = last saved model
resume_path = None # updated from load_run and chkpt resume_path = None # updated from load_run and chkpt
amp_reward_coef = 1.0 amp_reward_coef = 0.2
amp_motion_files = MOTION_FILES amp_motion_files = MOTION_FILES
amp_num_preload_transitions = 2000000 amp_num_preload_transitions = 2000000
amp_task_reward_lerp = 0.9 amp_task_reward_lerp = 0.9
amp_discr_hidden_dims = [1024, 512] amp_discr_hidden_dims = [1024, 512]
amp_replay_buffer_size = 1000000 amp_replay_buffer_size = 1000000
min_normalized_std = [0.01, 0.01, 0.01, 0.01] * 3 min_normalized_std = [0.001, 0.001, 0.001, 0.001] * 3
\ No newline at end of file \ No newline at end of file
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册登录 后发表评论