提交图

69 次代码提交

作者 SHA1 备注 提交日期
6224f3b993 Update README.md 2026-07-07 23:08:05 +08:00
626f655789 Update README.md 2026-07-07 23:06:37 +08:00
31d63e0884 Rename WZJH_Official.png to WZJHOfficial.png 2026-07-07 23:06:17 +08:00
39ea919076 Rename WZJH Official.png to WZJH_Official.png 2026-07-07 23:06:03 +08:00
aa143a0584 Add files via upload 2026-07-07 23:04:58 +08:00
64862dfc16 Add RW-RL dataset card news (#30)
Co-authored-by: Elvin-yk <180030703+Elvin-yk@users.noreply.github.com>
2026-06-30 12:11:31 +08:00
dc6732617e Add RW-RL dataset link to README (#28)
Co-authored-by: Elvin-yk <180030703+Elvin-yk@users.noreply.github.com>
2026-06-23 19:32:02 +08:00
4425ccd672 Update README.md 2026-04-21 10:19:20 +08:00
082414ca6e Update README.md 2026-04-21 10:18:02 +08:00
5320717cba Update README.md 2026-04-21 10:16:48 +08:00
e573cfd0fe Add files via upload 2026-04-21 10:14:29 +08:00
c1b2795640 feat(logging): add swanlab backend for training runs (#17)
* feat(logging): add swanlab backend + evo1 stage2 optimizations

- Add SwanLab logging support (auto-selects when SWANLAB_API_KEY is set)
- Add stage2 training optimizations: tensor fastpath, gradient checkpointing
- Document stage2 config override behavior and runnable value-model alternative

Cherry-picked from: bfdc28c, bcecabb

* refactor(logging): separate SwanLabLogger from WandBLogger

- Restore WandBLogger to its original clean form with all comments
- Add SwanLabLogger as a separate class with the same interface
- Add make_logger() factory function to select backend based on
  SWANLAB_API_KEY environment variable
- Move evo1 stage2 runtime logging from lerobot_train.py into
  EVO1Policy.__init__ where it belongs
- Add ImportError handling for swanlab dependency

* fix(logging): gracefully fall back to WandB when swanlab is not installed

If SWANLAB_API_KEY is set but the swanlab package is not installed,
make_logger() now logs a warning and falls back to WandBLogger instead
of crashing with an ImportError at training startup.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Elvinyk <223015152@link.cuhk.edu.cn>
Co-authored-by: MrBlue0632 <2028207642@qq.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-14 10:31:40 +08:00
bd3d1fc549 fix: remove evo1 hardcodes and unused resume_pretrain from public code
Public training code should be policy-agnostic. This removes:
- resume_pretrain config field (pretrained_path without --resume already
  provides the same "load weights, fresh optimizer" behavior)
- evo1-specific DDP static graph and stage2 logging in lerobot_train.py
- drop_last sourced from policy config (DataLoader concern, not policy)
- type-inference fallback in PreTrainedConfig that was added to handle
  evo1 checkpoints missing a "type" field

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-12 17:55:27 +08:00
ddb0738a76 fix(pi05): restore tied embed tokens on load (#20)
Co-authored-by: shiki42 <shiki42@users.noreply.github.com>
2026-04-12 16:51:40 +08:00
081d826f26 restore deleted bundled assets and test artifacts (#18) 2026-04-08 17:53:19 +08:00
c65a92c351 support Evo1 policy (#11)
* feat(evo1): integrate evo1 with lerobot training

* chore(evo1): drop v21 export and local tests

* docs(evo1): add setup notes and keep only 4gpu config

* chore(evo1): clean up quality issues

* fix(evo1): use batch_size consistently

* docs(evo1): add handoff notes

* docs(evo1): clarify resume pretrain semantics

* fix(evo1): address review follow-ups

* docs(evo1): refresh inherited issues note

* fix(evo1): update Evo1 model to higher effience

* fix(evo1): GPU version of dynamic process and ratio cache

* fix(evo1): support stage2 checkpoint resume

---------

Co-authored-by: t0mo0n <2423843425@qq.com>
2026-04-03 19:26:55 +08:00
ca80be1e0d fix: change post-recording hint to lerobot-dataset-report
The recording completion hint should point users to the dataset report
tool first, which itself already suggests boundary frame export.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-01 19:13:33 +08:00
80fa8bf47a feat: data quality check via episode boundary frame export (#13)
* feat: add episode boundary frame export script

* feat: add boundary frame export hints after recording and in dataset report

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: print scp hint when running in SSH session

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: address review findings — deduplicate utils, fix CHW handling, fix --json output

- Replace duplicated resolve_dataset_root/load_info/HF_LEROBOT_HOME with imports from existing modules
- Replace custom tensor_to_pil_image with torchvision.transforms.functional.to_pil_image
- Fix select_camera_key to use names field for CHW/HWC shape disambiguation
- Move dataset report tip inside text-only branch to preserve --json output
- Move prepare_output_dir after dataset construction to avoid premature deletion
- Fix resolve_repo_id to always derive from resolved dataset_root
- Remove unnecessary scripts/__init__.py
- Fix episode spec display in README.txt for non-contiguous selections
- Use episodes.column_names for column containment check

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: shuyuan <shuyuan@shuyuandeMac-mini.local>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-01 19:12:21 +08:00
7e4e6701d7 fix: ensure human-inloop schema fields are added for all teleop recordings
The previous condition gated on enable_collector_policy_id (defaults False),
so pure teleop recordings still missed the three complementary_info fields.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-27 14:35:52 +08:00
88f2e56b08 chore: remove AGENTS.md
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-27 14:12:38 +08:00
cad806fa60 feat: add cooldown for intervention toggle key to prevent rapid re-triggers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-27 14:10:10 +08:00
57af3361ce fix: stabilize human-inloop dataset schema 2026-03-19 10:35:31 +08:00
0b720b1e0b Merge pull request #10 from JAVAdcc/javadcc/headless-tty-record-controls
add TTY controls for headless human-inloop recording
2026-03-16 14:16:57 +08:00
eefe5c70be Add TTY controls for headless human-inloop recording 2026-03-15 17:45:31 +08:00
6eb44f69bc Merge pull request #7 from MINT-SJTU/develop
Develop
2026-03-11 21:16:18 +08:00
791122f975 Merge pull request #6 from post-train/codex/piper-teleop-lightweight-send-only
Codex/piper teleop lightweight send only
2026-03-11 21:15:33 +08:00
fa20ecaf8d feat(piper-teleop): optimize lightweight teleoperation
- add send-only follower support and skip unnecessary observation fetches

- add bi-piper process isolation and fresh feedback handling for manual control

- split piper and piperx gravity compensation defaults and expand teleop tests
2026-03-11 10:03:51 +08:00
ccfa70e801 Update README.md 2026-03-10 18:42:56 +08:00
512ce187b3 Merge pull request #3 from MINT-SJTU/codex/success-reset-pose
Reset human-in-loop robot to stored pose after success
2026-03-09 17:49:29 +08:00
ea32523d4e fix(human-inloop): reset to stored pose after success
- return the robot to the stored reset pose for both success and failure outcomes
- clarify reset logging so it matches the generalized behavior
- add a controller test covering success-triggered resets
2026-03-09 17:35:02 +08:00
dfd78e12ae Merge develop into main (Piper integration + CI/test updates)
Merge develop into main (Piper integration + CI/test updates)
2026-03-07 22:28:41 +08:00
cb29370c78 feat(piper): add bimanual PiPER/PiPER-X support and refresh docs
- add bi_piper/bi_piperx follower and leader implementations with configs

- wire bimanual Piper types into factories, scripts, and async inference mappings

- add pairing validation and bimanual parsing/teleop test coverage

- update README news and AgileX teleop/data-collection command templates
2026-03-07 22:27:09 +08:00
0d2ed91280 Revert "feat(policy): add minimal swing policy for hil record smoke runs"
This reverts commit 0714316bab.
2026-03-07 16:36:06 +08:00
0714316bab feat(policy): add minimal swing policy for hil record smoke runs 2026-03-07 12:45:43 +08:00
70c4b2d665 fix(piper): normalize int-like ctrl_mode enum to int 2026-03-07 12:07:54 +08:00
5e1d1cc999 Revert "fix(piper): robust ctrl_mode parsing for enum/string values"
This reverts commit 2190f5c237.
2026-03-07 11:44:14 +08:00
2190f5c237 fix(piper): robust ctrl_mode parsing for enum/string values 2026-03-07 11:40:42 +08:00
1be38954e6 fix(ci): make pre-commit quality checks pass 2026-03-07 02:21:09 +08:00
b69ffe3c02 fix(can): honor effective fd mode in speed_test 2026-03-07 02:09:58 +08:00
20ce8ca651 Revert "test(value): skip viz tests when scipy is unavailable"
This reverts commit fe9869c2bd.
2026-03-07 01:56:21 +08:00
7049938b87 feat(piper): add dedicated piperx types and auto URDF selection
- add dedicated piperx leader/follower types with shared piper logic
- auto-select bundled gravity-comp URDF by teleop type
- simplify piper teleop tests and tighten gravity-comp rollback/init safety
- bundle piper/piperx assets and refresh README/LFS guidance
2026-03-07 01:51:08 +08:00
f68f8eb543 refactor(piper): simplify teleop tests and gravity-comp init
- parametrize piper/piperx roundtrip teleop test to remove duplication

- stop gravity-comp loop before disconnect on connect rollback

- avoid repeating URDF/LFS checks after loop initialization
2026-03-07 01:50:26 +08:00
8934fdbabf fix(piper): guard LFS pointer URDF and normalize asset file modes
- detect Git LFS pointer URDF early and raise actionable runtime error

- add regression test for LFS-pointer URDF guard

- normalize PiPER/PiPER-X asset modes from 100755 to 100644
2026-03-07 01:34:17 +08:00
92d9d829b3 feat(piper): add dedicated piperx types and auto URDF selection
- add piperx_leader/piperx_follower config and runtime classes reusing piper logic

- remove manual gravity_comp_urdf_path and select bundled URDF by teleop type

- add bundled piper/piperx description assets tracked via git lfs

- update README piper setup and refresh piper teleop tests
2026-03-07 00:48:17 +08:00
fe9869c2bd test(value): skip viz tests when scipy is unavailable
- add pytest.importorskip('scipy') at module import time to avoid collection failures in minimal test environments
2026-03-06 17:24:35 +08:00
300f5a386d fix(piper): bool calibration flag and robust can setup/test fallback 2026-03-06 16:56:25 +08:00
5975798aef test(fast): align fast-test behavior with current runtime constraints
- gate PI0/PI0Fast tests on HF token availability for gated models

- update value infer viz tests to current API ownership and arguments

- align piper calibration assertion with runtime error message

- set HF_HUB_DISABLE_XET in fast_tests workflow
2026-03-06 16:31:02 +08:00
fa435f8312 fix(ci): resolve quality workflow failures 2026-03-06 09:48:31 +08:00
ae8efef693 fix(piper): decouple connect enable from calibrate flag 2026-03-06 01:35:05 +08:00
271518d52f fix(piper): enable follower after connect-time calibration 2026-03-06 00:52:37 +08:00