feat(examples): reading-data stage + numbered examples taxonomy (phase 1) - #28
feat(examples): reading-data stage + numbered examples taxonomy (phase 1)#28everettVT wants to merge 2 commits into
Conversation
…nomy Phase 1 of splitting #13 into reviewable pieces: pure examples, zero new package surface. - examples/README.md indexes the researcher workflow as numbered stages (01 reading data ... 08 policy evals); unlanded stages are marked planned with their target scripts named - 01_reading_data: droid_episode_index.py (daft.datasets.droid; column names updated for the 0.7.18 *_cam_video schema), lerobot_episode_index.py (read_episodes/read_tasks/read without decoding video), and the existing egodex_raw_hdf5_video.py relocated into the stage - 02_episode_data: merge_lerobot_datasets.py - two recording sessions into one table by re-indexing episode_index and the global frame index - the generated hand-tracking demo moves to 04_episode_operations/hand_tracking/ (renames; regen default + README links updated) Every landed script runs on a clean env against daft 0.7.18; suite and pre-commit (ruff + ty) green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 085ca7b749
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| daft-physical-ai # interactive | ||
| daft-physical-ai --method wilor --runtime modal --mano-path MANO_RIGHT.pkl --no-input |
There was a problem hiding this comment.
When a reader follows these commands to generate a custom demo, they won't run the generator: the top-level parser creates subparsers and returns after printing help if no command is provided (daft_physical_ai/cli/__init__.py:21-31), while --method/--runtime are registered only on the hands subcommand (daft_physical_ai/cli/hands.py:36-42). Please keep daft-physical-ai hands ... here so both the interactive and noninteractive examples actually scaffold the demo.
Useful? React with 👍 / 👎.
0.7.19 adds native MCAP reading (queued as a stage-01 example), HuggingFace path-handling fixes, and single_file parquet writes. All stage-01/02 examples and the suite revalidated against it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
1 participant
What this is
Phase 1 of splitting #13 into small, atomic PRs. Pure examples - zero new package surface.
examples/README.md: the numbered 01-08 taxonomy (the researcher workflow: read, episode data, transforms, episode ops, inference, write, training handoff, policy evals). Unlanded stages are marked planned with target scripts named - the tree is the roadmap.droid_episode_index.py(nativedaft.datasets.droid; fixed for 0.7.18's*_cam_videocolumns),lerobot_episode_index.py(read_episodes/read_tasks/read, no video decoded), and the existingegodex_raw_hdf5_video.pyrelocated into the stage.merge_lerobot_datasets.py- merge two recording sessions by re-indexingepisode_index+ the global frameindex, then concat.04_episode_operations/hand_tracking/- pure renames (R100), regen-script default and README links updated. Don't review the generateddemo.*contents.Validation
All landed scripts run on a clean env against daft 0.7.18;
pytest(53 passed) and pre-commit (ruff + ty) green.The stack
Next phases, each building on the last: (2) the episode contract, (3) LIBERO demonstrations through it, (4) pose, (5) evals + benchmark data. #13 stays open as the integration preview; it will be closed when the stack lands.
🤖 Generated with Claude Code