[AnikethCheluva]

Stack context

Follow-up to #269, above #268 → #267 → #263. This is the requested real/sim
menu organization and retirement of the remaining E1 surfaces. No merge or
training launch is included.

What and why

• Put organize stationery, fold/stack towels, wireless earbuds and multitask
  under experiment/real/, each with complete baseline, arc_vel, arc_dur
  entry points. Defaults bind model, data/splits, optimizer, evaluator, trainer,
  checkpoints and logging; machine/credential bindings remain external.
• Put LIBERO, its historical profiles, OAT/LIBERO, PushT and synthetic
  experiments under experiment/sim/. Update Hydra defaults, CLI selectors,
  launcher allowlists, profile registries, tests and documentation.
• Remove the E1 experiment/model/data/evaluator configs, Ravens config menu,
  and stationary_recipe. Delete E1 embodiment/image compatibility shims,
  campaign scripts, obsolete E1 launcher and their config-only tests.
• Retain only the shared infrastructure required by current real recipes under
  neutral names: frozen YAM HPT configs, embodiment resolvers, progress sampler
  and the existing ResNetMLP image stem. No tokenizer/detokenizer arithmetic
  changes in this layer.
• Preserve dated calibration evidence; map its old selector through the
  migration table when replaying a current recipe.

Compatibility

This intentionally removes legacy import/config entry points. Older
checkpoints and running jobs must keep their recorded source and normalizer.
The original Lambda training checkout is not switched. No datasets, weights,
normalizers or checkpoint files are deleted; removed source is in Git history.

Verification

• All 343 shipped YAMLs compose and resolve in explicit offline audit contexts.
• All 12 real recipes match their pre-removal resolved model, data, evaluator,
  trainer, callback and logger settings, allowing only exact module alias
  replacements and freshly generated W&B IDs.
• Complete-entry-point, clock/shape/rollout alignment, shared resolver,
  simulator selector and deployment tests are included.
• Combined affected-test run: 1,070 passed, 4 skipped, 1 deselected. Its one
  remaining failure was a config-only test of the intentionally deleted Ravens
  menu; that obsolete test was subsequently removed and the retained Ravens
  runtime tests plus lightweight recursive-audit tests were rerun: 10 passed.
• Ruff checks pass for all modified Python files.
• All 34 retained PushT recipes are byte-equivalent as unresolved composed
  configurations (excluding Hydra selector metadata) to parent commit
  b6686ec9ad4fc8bbc89c97263fb63ec1891e1ae2. The older pre-parent-cleanup
  fixture is preserved; a separately pinned fixture guards this move.
• Offline composition does not certify dataset accessibility, credentials,
  pretrained weights or GPU training readiness.
• Lambda's persistent next checkout was synced to this commit. CPU-only Slurm
  job 411859 passed new-source imports and all 12 real recipe/codec/evaluator
  constructions (COMPLETED, exit 0:0, 2 CPUs, 8 GiB, zero GPUs).
  Existing training source checkouts and their shared environment were preserved.

[AnikethCheluva]

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

• arc: one clock per arm, absolute-duration clocks with speed-bounded isotonic decode #272 : 2 dependent PRs (#273 , #274 )
• Organize real and sim experiments and remove retired E1 surfaces #270 👈 (View in Graphite)
• configs: organize Cartesian experiments by task #269
• arc: consolidate velocity/duration clocks and M×18/M×20 streams #268
• arc: encode first-row start delays in non-E1 clocks #267
• Bring Aniketh's ARC work onto main with the YAM task recipes and the station rollout runtime #263 : 1 other dependent PR (#266 )
• main

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