This repository was archived by the owner on Aug 31, 2026. It is now read-only.
Correct performance claims, remove fabricated persona and self-audit docs - #9
Merged
Conversation
…docs Cleanup pass ahead of archiving this repository. Claims (P0-1): - Rewrite README performance sections to the committed 250-measurement CSV (robocache/bench/results/benchmark_h100_20251106_172811.csv: 0.184 / 2.605 / 20.05 ms P50, ~3.8-110x kernel-level vs PyTorch CPU), the measured 1.30x end-to-end training result (14.04 vs 18.28 ms/step), and the documented 0.74x regression at 64x4096->1024x32. - Remove every "10-20x faster training" claim; replace with measured numbers. - State that benchmark inputs were synthetic dataset-shaped tensors (torch.randn), not real Isaac Gym / TartanAir / nuScenes / KITTI data; retitle and correct docs/validation/REAL_WORLD_VALIDATION.md. - Drop "Battle-tested", "Compute Sanitizer verified", and CI claims that never ran; add a Known Issues section (CPU voxel fallback incorrect, stale multimodal tests). - Mark PRODUCTION_STATUS.md and GPU_CI_STATUS.md as historical; fix the nonexistent .ncu-rep artifact path; correct KNOWN_LIMITATIONS.md CPU latencies to the CSV values. - Add archive banner: last GPU validation 2025-11-08 on H100 PCIe at 0db3726. Persona / self-audit (P0-2): - Strip "Expert CUDA/NVIDIA Engineer (15+ years)" author-persona lines from 53 files (kernels, benchmarks, profiling and validation docs), keeping the code and measurements. - Delete self-issued audit/marketing docs: artifacts/PROOF_OF_EXCELLENCE.md, CODEX_RESPONSE_FINAL_STATUS.md, codex_audit_completion_report.md and seven sibling audit-response docs, docs/internal/ (17 status files), and docs/validation/EXCELLENCE_CONFIRMED.md. Raw measurement artifacts (artifacts/h100/, artifacts/a100/, bug_fixes/, benchmark CSVs/JSON) are kept. Metadata (P1-2, P1-3, P1-7): - robocache/pyproject.toml: license MIT -> Apache-2.0 (matches LICENSE), Apache classifier, requires-python >=3.10. - Replace all "pip install robocache" instructions with the from-source install (package was never published to PyPI); correct SECURITY.md pinning/ Dependabot claims and CHANGELOG distribution claims. - Align version strings on 1.0.0 (_version.py, tests/__init__.py). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01142EwVwkXFRGnYu7YgHSYL
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01142EwVwkXFRGnYu7YgHSYL
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
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.
Pre-archive cleanup, part 1 of 2 (claims and provenance). Addresses audit findings P0-1, P0-2, P1-2, P1-3, P1-7.
Performance claims (P0-1)
robocache/bench/results/benchmark_h100_20251106_172811.csv(5 seeds x 50 repeats = 250 measurements per CUDA config): 0.184 / 2.605 / 20.05 ms P50, ~110x / ~15x / ~3.8x vs PyTorch CPU, labelled as kernel-vs-CPU microbenchmarks.PRODUCTION_STATUS.md/robocache/profiling/NCU_H100_TRAJECTORY_RESAMPLE.md).robocache/benchmarks/results/h100_validated_20251105.json).grep -rn "10-20"now matches only honest ms/percent/slower statements.torch.randn);docs/validation/REAL_WORLD_VALIDATION.mdretitled with a correction note pointing atrobocache/benchmarks/real_world_datasets.py.Fabricated persona and self-audits (P0-2)
artifacts/PROOF_OF_EXCELLENCE.md,artifacts/CODEX_RESPONSE_FINAL_STATUS.md,artifacts/codex_audit_completion_report.md, plus sibling audit-response docs inartifacts/, all ofdocs/internal/, anddocs/validation/EXCELLENCE_CONFIRMED.md.artifacts/h100/,artifacts/a100/,artifacts/bug_fixes/, benchmark CSV/JSON) - those are real captures with GPU/driver stamps.Metadata (P1-2, P1-3, P1-7)
robocache/pyproject.tomlMIT -> Apache-2.0 + Apache classifier (matchesLICENSE).pip install robocacheinstruction replaced with the from-source install; SECURITY.md dependency-pinning/Dependabot claims and CHANGELOG "shipped" distribution claims corrected._version.py,tests/__init__.pywere 0.2.1);requires-pythonraised to >=3.10 to match the README badge.Part 2 (truthful CI: drop continue-on-error, fix/xfail failing CPU tests, delete dead self-hosted workflows) follows in a separate PR.
🤖 Generated with Claude Code
https://claude.ai/code/session_01142EwVwkXFRGnYu7YgHSYL