pepijn223 71aacda05e feat(lingbot_va): implement training / fine-tuning (flow-matching loss) ...

- Implement LingBotVAPolicy.forward(): dual-stream flow-matching training loss
  (latent + action, timestep-weighted, action-masked) ported from upstream train.py;
  VAE-encodes camera clips, UMT5-encodes the task, noises both streams, runs the
  block-causal flex-attention training pass (forward_train).
- training_loss_from_streams() core + _build_training_streams() data prep (action
  scatter into the 30-d space, multi-frame VAE encode incl. robotwin_tshape).
- get_optim_params returns only trainable transformer params (LoRA/PEFT friendly);
  VAE/UMT5 stay frozen. Training needs attn_mode='flex'.
- Add a tiny-config single-training-step test (forward->loss->backward->AdamW) and a
  Training/fine-tuning section in the docs.

Co-authored-by: Cursor <cursoragent@cursor.com>

2026-06-06 15:38:41 +02:00

..

artifacts

feat(encoding parameters): adding support for user provided video encoding parameters (#3455)

2026-05-14 23:46:42 +02:00

async_inference

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

cameras

test(cameras): skip flaky async_read test (#3106)

2026-03-08 14:02:33 +01:00

configs

Add extensive language support (#3467)

2026-05-19 14:46:11 +02:00

datasets

fix(datasets): bound VideoDecoderCache to prevent OOM on large datasets (#3614)

2026-05-19 16:54:25 +02:00

envs

feat(envs): add RoboTwin 2.0 benchmark (#3315)

2026-04-20 17:46:39 +02:00

fixtures

feat(encoding parameters): adding support for user provided video encoding parameters (#3455)

2026-05-14 23:46:42 +02:00

mocks

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

motors

fix(feetech): motor position readings overflow (#3373)

2026-04-13 22:39:58 +02:00

optim

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

policies

feat(lingbot_va): implement training / fine-tuning (flow-matching loss)

2026-06-06 15:38:41 +02:00

processor

fix(pi0, pi05): stabilize torch.compile and expand test coverage (#3610)

2026-05-22 10:29:34 +02:00

rewards

feat(rewards): add ROBOMETER reward model (#3627)

2026-05-29 21:45:39 +02:00

rl

RL stack refactoring (#3075)

2026-05-12 15:49:54 +02:00

robots

feat(robots): natively integrate Seeed Studio reBot B601-DM arm (#3624)

2026-05-18 19:49:21 +02:00

scripts

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

teleoperators

feat(robots): natively integrate Seeed Studio reBot B601-DM arm (#3624)

2026-05-18 19:49:21 +02:00

training

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

transport

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

utils

Add extensive language support (#3467)

2026-05-19 14:46:11 +02:00

__init__.py

chore(doc): add license header to all files (#818)

2025-03-05 17:56:51 +01:00

conftest.py

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

test_available.py

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00

test_cli_peft.py

feat(rollout): decouple policy deployment from data recording with new lerobot-rollout CLI (#3413)

2026-04-28 00:57:35 +02:00

test_control_robot.py

feat(rollout): decouple policy deployment from data recording with new lerobot-rollout CLI (#3413)

2026-04-28 00:57:35 +02:00

test_robomme_env.py

feat(envs): add RoboMME benchmark (#3311)

2026-04-20 20:21:27 +02:00

test_rollout.py

feat(rollout): decouple policy deployment from data recording with new lerobot-rollout CLI (#3413)

2026-04-28 00:57:35 +02:00

test_yaml_policy_path.py

Fix policy.path in YAML configs (PR #3145 followup) (#3597)

2026-05-26 14:01:19 +02:00

utils.py

feat(dependencies): minimal default tag install (#3362)

2026-04-12 20:03:04 +02:00