Repository navigation
Plan Skillspace coverage and record simulator episode evidence - #3
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Skillspace collection needs a bounded plan for useful evidence, and simulation research needs state/command recordings before we can learn task outcomes. This adds a local coverage planner and a separate versioned episode-tape recorder, with an execution map connecting collection, review, learning, benchmarks and installable candidates.
/coverageprovides editable task goals/conditions, prioritized capture briefs, strict JSON import/export, declared source-group counts and a mandatory planned/blocked benchmark description. Revisions preserve review history; unchanged saves preserve the revision. Exact duplicates and transitively related recordings/sessions/objects cannot inflate independent-group quotas or cross confirmation/train/development boundaries. Reviews, rights and lineage remain declarations, not authenticated approval. The default records are prominently fabricated examples.dvidia-episode-taperecords the existing authored MuJoCo controller at control boundaries, keeping measured state separate from policy requests, issued commands and final applied controls. It retains contact forces, sensor availability/age, authored phases, outcomes, source IDs and consistency hashes. Validation/inspection are standard-library-only; native recording uses the existing optional arm extra.dvidia-skillspace-coverage, collection/incentive design and the execution map are documented. Proposed recognition/funded campaigns issue no points, rewards, payments or public credit. Collection targets, model training, simulation qualification and physical readiness remain separate states.Validation: all 176 tests passed with the existing MuJoCo 3.15.0 environment; the CPU/base suite passed 119 tests with 9 optional-native skips. The tape suite includes fresh nominal, open-jaw and release/recovery recording cases and CLI round trips. The HTTP suite checks raw duplicate-key rejection, the dedicated bounded 1 MiB planner import, unchanged 32 KiB training limits, local Origin/Host guards and preservation of an existing training result. Installed entry points work. Desktop/mobile browser checks covered plan creation/editing, revision behavior, provenance warnings, case addition/removal, actual exports and training navigation; no horizontal overflow at 390 px.
Limits: no new policy, learned supervisor, human-video action bridge, task-general simulator or physical-robot qualification. Tapes sample control boundaries rather than intermediate native ticks. Existing runtime and frozen benchmark artifacts are unchanged; older release/Hugging Face pilot artifacts are not replaced.