tangger/lerobot
1
0
Fork 1
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

- Fixed big issue in the loading of the policy parameters sent by the learner to the actor -- pass only the actor to the update_policy_parameters and remove strict=False

Browse Source 
- Fixed big issue in the normalization of the actions in the `forward` function of the critic -- remove the `torch.no_grad` decorator in `normalize.py` in the normalization function
- Fixed performance issue to boost the optimization frequency by setting the storage device to be the same as the device of learning.

Co-authored-by: Adil Zouitine <adilzouitinegm@gmail.com>
This commit is contained in:
Michel Aractingi
2025-02-19 16:22:51 +00:00
committed by AdilZouitine
parent d9cd85d976
commit 795063aa1b
7 changed files with 68 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
lerobot/configs/policy/sac_maniskill.yaml
View File

@@ -64,13 +64,29 @@ policy:
action: [7]
# Normalization / Unnormalization
input_normalization_modes: null
input_normalization_modes:
observation.state: min_max
input_normalization_params:
observation.state:
min: [-1.9361e+00, -7.7640e-01, -7.7094e-01, -2.9709e+00, -8.5656e-01,
1.0764e+00, -1.2680e+00, 0.0000e+00, 0.0000e+00, -9.3448e+00,
-3.3828e+00, -3.8420e+00, -5.2553e+00, -3.4154e+00, -6.5082e+00,
-6.0500e+00, -8.7193e+00, -8.2337e+00, -3.4650e-01, -4.9441e-01,
8.3516e-03, -3.1114e-01, -9.9700e-01, -2.3471e-01, -2.7137e-01]
max: [ 0.8644, 1.4306, 1.8520, -0.7578, 0.9508, 3.4901, 1.9381, 0.0400,
0.0400, 5.0885, 4.7156, 7.9393, 7.9100, 2.9796, 5.7720, 4.7163,
7.8145, 9.7415, 0.2422, 0.4505, 0.6306, 0.2622, 1.0000, 0.5135,
0.4001]
output_normalization_modes:
action: min_max
output_normalization_params:
action:
min: [-10.0, -10.0, -10.0, -10.0, -10.0, -10.0, -10.0]
max: [10.0, 10.0, 10.0, 10.0, 10.0, 10.0, 10.0]
output_normalization_shapes:
action: [7]
# Architecture / modeling.
# Neural networks.