[RyanPCo]

_video_fps divided 30 fps by world_size to undo a DistributedSampler stride
that never applies: the val heads are CombinedLoaders Lightning does not
shard (every rank runs every val batch; 1 rank = 223 val steps, 8 ranks =
8 x 223 in the 2026-09-16 DDP check), so rank 0 renders every frame and the
8-GPU videos came out as 8x slow motion (416 frames at 4 fps).

🤖 Generated with Claude Code

https://claude.ai/code/session_01HjCQBqQCr2igFEMR7yUovw

[RyanPCo]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• pi: initialize from PaliGemma with a scratch action expert #677 : 2 dependent PRs (#678 , #680 )
• perf: vectorized image augs and a torch.compile knob for HPT and Pi; Pi clips gradients like openpi #676
• loader: one zarr read per key, not one per keymap entry #674
• eval: build the pinned-episode video batches only where they are used #673
• checkpoints: last.ckpt as a symlink, not a second full copy #672
• trainHydra: metric_frames_per_episode: auto, and warn on empty prompts #671
• transforms: per-step wrist framing for proprio history #670
• viz: one keypoint renderer, the horizon trace #669
• configs: history and backbone-LR knobs on the keypoint wrist HPT model #668
• trainer: ddp_bf16_compress, DDP with a bf16 gradient all-reduce #667
• eval: overlay videos play at the source rate under DDP #666 👈 (View in Graphite)
• hpt: build the policy on CPU and let Lightning place it #664 : 1 other dependent PR (#665 )
• hpt: never re-initialise pretrained stems; hash-check them; mask text padding #645 : 1 other dependent PR (#646 )
• obs: proprio history (K proprio steps, one token each, learned time embedding) #650
• loss: mask the repeat-last action padding out of the HPT loss #649
• resnet: keep a frozen backbone in eval mode; give backbones their own LR group #652
• augs: per-sample image augmentation for HPT train-mode augs #651
• eval: deterministic validation (seeded prompts + flow sampling); log-only grad-norm MAD #644
• data: mecka_fold_flagship_top3_hpt_6d, the 3-operator flagship train split #643
• val: subsampled metric loaders + pinned contiguous video loaders #642
• infra: mecka 6D / keypoint recipes, 300M models, defaults, and the rollout revert #640
• eval: layout-aware frame-invariant metrics, and a train-split viz head #659
• data: stop letting one bad episode or a rotation channel take a run down #658
• norm: guard degenerate channels, horizon pooling, and no rot6d normalization #657
• keypoints: the 144-D wrist-first hand-keypoint action becomes the human default #656
• mecka: fix the double-mirrored LEFT wrist frame, and correct existing zarrs #655
• 6d: PI goes 6D-only; Eva and Human gain the continuous-rotation modes #654
• 6d: continuous rotation representation, bimanual layouts, shared norm helpers #653
• train: give each Slurm slurm job gets its own cache dir to avoid conflict #638 : 1 other dependent PR (#639 )
• write git version into the run dir for reproducibility #637
• train: don't load PI base weights on resume and verify that checkpoint path exists #636
• configs: declare proprio and actions dims once per HPT model config #634 : 1 other dependent PR (#635 )
• cache norm stats for data configs to minimize recomputation #633
• import cloud packages lazily to support local training #631 : 1 other dependent PR (#632 )
• cache dataset resolver for faster training initialization #630
• validate episode hashes in data config before training #628
• modify pi camera map to match data #629
• add tests to ensure training works between every model and data vendor #626
• run CI tests on every PR #625
• add env variables for user-specific paths #624
• build: set pyproject and uv as the dependency managers #623
• fix(configs): fix paths for data hydra yaml configs #622
• refactor(algo): delete ACT #627 : 1 other dependent PR (#621 )
• tests: create dedicated test folder with unit and integration tests #620
• add pyright type checker #619
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[RyanPCo]

Automated adversarial review (read-only; findings were checked by a second skeptical pass before posting): 1 medium, 1 nit.

[RyanPCo]

medium (correctness): This is only right when val_dataloader returns a list. With a single head (train_viz=false, no unseen-op split, no video episodes), val_dataloader returns a bare CombinedLoader. Lightning then injects a DistributedSampler into its inner DataLoaders, so rank 0 renders every W-th frame and a fixed 30 fps gives a W-times fast-forward video. Either always return a list from val_dataloader so sharding does not depend on the number of heads, or keep the world_size division for the single-loader case, and add a test for that path.

[RyanPCo]

nit (simplification): _video_fps now just returns int(source_fps), and its only caller uses the default. Consider a VIDEO_FPS = 30 constant with a one-line why ('val loaders are unsharded, so rank 0 renders every frame') in place of the 11-line docstring with the 2026-09-16 measurement and the account of the old bug. Later PRs in the stack (eval_video._write_video, pl_data_utils._video_gate) point to 'see _video_fps', so those references would need to move too. The world_size loop in the test then has nothing to test.

[RyanPCo]

Review findings: status after the fixup pass. The fixes are squashed into this PR's commits, and the whole stack was force-pushed.

Fixed

• Removing the world_size fps division is wrong when val_dataloader returns a single CombinedLoader: val_dataloader now always returns a list, so Lightning shards no val loader. The fixup targets eval: build the pinned-episode video batches only where they are used #673's commit because a fixup at eval: overlay videos play at the source rate under DDP #666 conflicted on autosquash. New 2-rank DDP test checks both ranks see every frame; with the old bare return they saw 0,2,4,6 and 1,3,5,7. Also updated test_infra_port_wiring (outside the PR).
• _video_fps is now an identity function with a history-lesson docstring: Replaced it with a VIDEO_FPS = 30 constant and a one-line reason. Dropped its world_size test, and updated the 'see _video_fps' references added by eval: build the pinned-episode video batches only where they are used #673 and eval: log validation videos to wandb #684.