2051dd38fc
[HIL-SERL] Review feedback modifications ( #1112 )
2025-05-15 15:24:41 +02:00
c7a3973653
[PORT HIL-SERL] Better unit tests coverage for SAC policy ( #1074 )
2025-05-14 16:41:36 +02:00
bbc6b7d841
Added comment on SE(3) in kinematics and nits in lerobot/envs/utils.py
2025-05-12 18:05:22 +02:00
b104f8b012
Added number of steps after success as parameter in config
2025-05-09 18:09:10 +02:00
633edcb3af
added names in record_dataset function of gym_manipulator
2025-05-07 13:58:24 +02:00
adbf8bb85e
Cleaning configs
2025-05-07 10:26:32 +02:00
ad132c9c39
[HIL SERL] Env management and add gym-hil ( #1077 )
...
Co-authored-by: Michel Aractingi <michel.aractingi@gmail.com >
2025-05-07 09:39:21 +02:00
70d55c77e9
Merge branch 'main' into user/adil-zouitine/2025-1-7-port-hil-serl-new
2025-05-06 16:43:44 +02:00
5998203a33
[Port HIL-SERL] Final fixes for reward classifier ( #1067 )
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2025-05-05 11:33:09 +02:00
8cfab38824
Fix typos ( #1070 )
2025-05-05 10:35:32 +02:00
6fa7df35df
[PORT HIL-SERL] Add unit tests for SAC modeling ( #999 )
...
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2025-05-05 09:27:42 +02:00
6d723c45a9
feat(encoding): switching to PyAV for ffmpeg related tasks ( #983 )
2025-04-29 17:39:35 +02:00
42bf1e8b9d
Update tutorial ( #1021 )
...
Co-authored-by: Simon Alibert <75076266+aliberts@users.noreply.github.com >
2025-04-28 09:00:32 +02:00
4257fe5045
rename reward classifier
2025-04-25 18:38:52 +02:00
ea89b29fe5
checkout normalize.py to prev commit
2025-04-25 18:10:59 +02:00
50e9a8ed6a
cleaning
2025-04-25 17:22:02 +02:00
bd4db8d747
[Port HIl-Serl] Refactor gym-manipulator ( #1034 )
2025-04-25 16:34:54 +02:00
a8da4a347e
Clean the code
2025-04-24 17:22:54 +02:00
b8c2b0bb93
Clean the code and remove todo
2025-04-24 16:10:56 +02:00
c58b504a9e
[HIL-SERL]Remove overstrict pre-commit modifications ( #1028 )
2025-04-24 13:48:52 +02:00
299effe0f1
[HIL-SERL] Update CI to allow installation of prerelease versions for lerobot ( #1018 )
...
Co-authored-by: imstevenpmwork <steven.palma@huggingface.co >
2025-04-24 10:18:03 +02:00
c5845ee203
Fix linter issue
2025-04-22 10:37:08 +02:00
a7a51cfc9c
Refactor SACPolicy and configuration to replace 'grasp_critic' terminology with 'discrete_critic'. Update related methods and comments for clarity and consistency in handling discrete actions.
2025-04-18 14:57:03 +00:00
c1ee25d9f7
nits in configuration classifier and control_robot
2025-04-18 16:18:13 +02:00
9886520d33
Added option to add current readings to the state of the policy
2025-04-18 16:18:13 +02:00
3b24ad3c84
Fixes for the reward classifier
2025-04-18 16:18:13 +02:00
54c3c6d684
Enhance MLP class in modeling_sac.py with detailed docstring and refactor layer construction for improved readability. Simplify layer addition logic by removing unnecessary conditions and ensuring consistent handling of activations and dropout.
2025-04-18 14:15:06 +00:00
fb92935601
[pre-commit.ci] auto fixes from pre-commit.com hooks
...
for more information, see https://pre-commit.ci
2025-04-18 13:33:37 +00:00
dcd850feab
Refactor SACObservationEncoder to improve modularity and readability. Split initialization into dedicated methods for image and state layers, and enhance caching logic for image features. Update forward method to streamline feature encoding and ensure proper normalization handling.
2025-04-18 15:10:22 +02:00
1ce368503d
Refactor SACPolicy initialization by breaking down the constructor into smaller methods for normalization, encoders, critics, actor, and temperature setup. This enhances readability and maintainability.
2025-04-18 15:10:22 +02:00
fb075a709d
Refactor input and output normalization handling in SACPolicy for improved clarity and efficiency. Consolidate encoder initialization logic and remove redundant else statements.
2025-04-18 15:10:22 +02:00
3424644ecd
Fix init temp
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
2025-04-18 15:10:22 +02:00
c37936f2c9
Update log_std_min type to float in PolicyConfig for consistency
2025-04-18 15:10:22 +02:00
c5382a450c
fix caching
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
2025-04-18 15:10:22 +02:00
9e5f254db0
change the tanh distribution to match hil serl
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
2025-04-18 15:10:22 +02:00
8122721f6d
match target entropy hil serl
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
2025-04-18 15:10:22 +02:00
5c352ae558
stick to hil serl nn architecture
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
2025-04-18 15:10:22 +02:00
9386892f8e
Refactor modeling_sac and parameter handling for clarity and reusability.
...
Co-authored-by: s1lent4gnt <kmeftah.khalil@gmail.com >
2025-04-18 15:10:22 +02:00
267a837a2c
fix encoder training
2025-04-18 15:10:22 +02:00
28b595c651
[pre-commit.ci] auto fixes from pre-commit.com hooks
...
for more information, see https://pre-commit.ci
2025-04-18 15:10:22 +02:00
9fd4c21d4d
General fixes in code, removed delta action, fixed grasp penalty, added logic to put gripper reward in info
2025-04-18 15:10:22 +02:00
02e1ed0bfb
[pre-commit.ci] auto fixes from pre-commit.com hooks
...
for more information, see https://pre-commit.ci
2025-04-18 15:10:22 +02:00
e18274bc9a
fix caching and dataset stats is optional
2025-04-18 15:10:22 +02:00
68c271ad25
Add rounding for safety
2025-04-18 15:10:22 +02:00
203315d378
fix sign issue
2025-04-18 15:10:22 +02:00
d5a87f67cf
Handle gripper penalty
2025-04-18 15:10:22 +02:00
8bcf41761d
fix caching
2025-04-18 15:10:22 +02:00
7c2c67fc3c
Enhance SAC configuration and replay buffer with asynchronous prefetching support
...
- Added async_prefetch parameter to SACConfig for improved buffer management.
- Implemented get_iterator method in ReplayBuffer to support asynchronous prefetching of batches.
- Updated learner_server to utilize the new iterator for online and offline sampling, enhancing training efficiency.
2025-04-18 15:10:22 +02:00
70130b9841
Enhance SACPolicy to support shared encoder and optimize action selection
...
- Cached encoder output in select_action method to reduce redundant computations.
- Updated action selection and grasp critic calls to utilize cached encoder features when available.
2025-04-18 15:10:22 +02:00
6167886472
Enhance SACPolicy and learner server for improved grasp critic integration
...
- Updated SACPolicy to conditionally compute grasp critic losses based on the presence of discrete actions.
- Refactored the forward method to handle grasp critic model selection and loss computation more clearly.
- Adjusted learner server to utilize optimized parameters for grasp critic during training.
- Improved action handling in the ManiskillMockGripperWrapper to accommodate both tuple and single action inputs.
2025-04-18 15:10:22 +02:00
Adil Zouitine