Apollo3D/Gym_GPU
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

change parameter

Browse Source
This commit is contained in:
ChenXi
2026-03-20 10:51:07 -04:00
parent 1fbc9dccac
commit 6d2ad9846a
2 changed files with 6 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
rl_game/get_up/config/ppo_cfg.yaml
View File

@@ -41,19 +41,19 @@ params:
normalize_advantage: True normalize_advantage: True
gamma: 0.99 gamma: 0.99
tau: 0.95 tau: 0.95
learning_rate: 3e-4 learning_rate: 2e-4
lr_schedule: adaptive lr_schedule: adaptive
kl_threshold: 0.008 kl_threshold: 0.008
score_to_win: 20000 score_to_win: 20000
max_epochs: 500000 max_epochs: 500000
save_best_after: 50 save_best_after: 50
save_frequency: 100 save_frequency: 100
grad_norm: 1.0 grad_norm: 0.5
entropy_coef: 0.05 entropy_coef: 0.005
truncate_grads: True truncate_grads: True
bounds_loss_coef: 0.001 bounds_loss_coef: 0.001
e_clip: 0.2 e_clip: 0.2
horizon_length: 64 horizon_length: 32
minibatch_size: 4096 minibatch_size: 4096
mini_epochs: 5 mini_epochs: 5
critic_coef: 1 critic_coef: 1

5
rl_game/get_up/config/t1_env_cfg.py
View File

@@ -1,5 +1,4 @@
import random import random
import numpy import numpy
import numpy as np import numpy as np
import torch import torch
@@ -343,7 +342,7 @@ class T1GetUpRewardCfg:
# 7. 成功终极大奖 # 7. 成功终极大奖
is_success = RewTerm( is_success = RewTerm(
func=lambda env, keys: env.termination_manager.get_term(keys), func=lambda env, keys: env.termination_manager.get_term(keys),
weight=1000.0, weight=300.0,
params={"keys": "standing_success"} params={"keys": "standing_success"}
) )
@@ -383,4 +382,4 @@ class T1EnvCfg(ManagerBasedRLEnvCfg):
actions = T1ActionCfg() actions = T1ActionCfg()
episode_length_s = 6.0 episode_length_s = 6.0
decimation = 4 decimation = 2