wangzixiang
/
lerobot
mirror of https://github.com/huggingface/lerobot.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

added comments from kewang

This commit is contained in:
Michel Aractingi 2024-12-17 18:03:46 +01:00
parent 7e0f20fbf2
commit 7b68bfb73b
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lerobot/common/policies/sac/modeling_sac.py
View File

@ -128,6 +128,7 @@ class SACPolicy(
# perform image augmentation # perform image augmentation
# reward bias # reward bias
# from HIL-SERL code base
# add_or_replace={"rewards": batch["rewards"] + self.config["reward_bias"]} in reward_batch # add_or_replace={"rewards": batch["rewards"] + self.config["reward_bias"]} in reward_batch
@ -207,6 +208,7 @@ class SACPolicy(
def update(self): def update(self):
self.critic_target.lerp_(self.critic_ensemble, self.config.critic_target_update_weight) self.critic_target.lerp_(self.critic_ensemble, self.config.critic_target_update_weight)
# TODO: implement UTD update # TODO: implement UTD update
# First update only critics for utd_ratio-1 times
#for critic_step in range(self.config.utd_ratio - 1): #for critic_step in range(self.config.utd_ratio - 1):
# only update critic and critic target # only update critic and critic target
# Then update critic, critic target, actor and temperature # Then update critic, critic target, actor and temperature