Repository navigation
docs/MUSEBENCH.md summarizes the MuseBench paper and dataset card (composition, card-vs-paper discrepancies, protocol, o - #8
Merged
Conversation
Walks each of the 75 MuseBench tar shards header by header over HTTP range requests (GNU long names, POSIX prefix, pax paths), resumable per shard, writing index/<shard>.jsonl and .done under MUSEBENCH_ROOT. Assumption: shards are standalone tars starting at offset 0 (verified on part1_002, part2_001, part3_008, part5_015).
verify.py inserts the repo root into sys.path so it runs under python -I. Later stages (acquire, leak, run) report 'no verifier yet' until their milestones land.
Summarizes arXiv:2610.20310v1 (Tables 1, 4, 10 checked against the arXiv HTML), the HF dataset card, their discrepancies, pinned shard revisions, the ArtifactBench zero-shot protocol, acquisition rules and licenses. No ArtifactNet scores are included (repo is public).
Per-header range requests cost ~0.7 s each; shards with ~18k small members would take ~5.7 h. A single forward GET per shard discards member data locally and resumes with Range: bytes=<offset>- after a dropped connection. Assumption: single-stream throughput (~12 MB/s measured) makes this faster than range mode for member-dense shards.
All 75 shards indexed (351,188 files, 798.0 GB, 351,183 audio = paper total minus the 9,983 unshipped Genius-Lyrics tracks). Corrects the generator-count rows of the card-vs-paper table. Assumption: magenta_generated_music is the paper's Lyria RealTime (identical 19,903 count and hours); marked as inferred.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Milestone of an auto-claude run.
Check:
test -f docs/MUSEBENCH.md && systemd-run --user --scope -q -p MemoryMax=6G ~/RTX_ENV/bin/python -m pytest -m "not slow" -q -p no:cacheprovider && ~/RTX_ENV/bin/python scripts/check_docs.py && ~/RTX_ENV/bin/python -I scripts/musebench/verify.py indexPlan: .claude/auto-plan.md (local to the run).