hangX/lerobot
1
0
Fork 0
forked from tangger/lerobot
Code Pull Requests Activity

- Added additional logging information in wandb around the timings of the policy loop and optimization loop.

Browse Source 
- Optimized critic design that improves the performance of the learner loop by a factor of 2
- Cleaned the code and fixed style issues

- Completed the config with actor_learner_config field that contains host-ip and port elemnts that are necessary for the actor-learner servers.

Co-authored-by: Adil Zouitine <adilzouitinegm@gmail.com>
This commit is contained in:
Michel Aractingi
2025-01-29 15:50:46 +00:00
parent 2ae657f568
commit 8cd44ae163
6 changed files with 461 additions and 313 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
lerobot/configs/policy/sac_manyskill.yaml
View File

@@ -8,8 +8,7 @@
# env.gym.obs_type=environment_state_agent_pos \
seed: 1
dataset_repo_id: null
dataset_repo_id: null
training:
# Offline training dataloader
@@ -75,15 +74,18 @@ policy:
# discount: 0.99
discount: 0.80
temperature_init: 1.0
num_critics: 2
num_critics: 2 #10
num_subsample_critics: null
critic_lr: 3e-4
actor_lr: 3e-4
temperature_lr: 3e-4
# critic_target_update_weight: 0.005
critic_target_update_weight: 0.01
utd_ratio: 2
utd_ratio: 2 # 10
actor_learner_config:
actor_ip: "127.0.0.1"
port: 50051
# # Loss coefficients.
# reward_coeff: 0.5