starred/turboquant_plus
1
0 Fork 0
mirror of https://github.com/TheTom/turboquant_plus.git synced 2026-10-22 11:02:42 -05:00
Code Issues Packages Projects Releases Wiki Activity
342 Commits 14 Branches 3 Tags
main
Commit Graph
8 Commits
Author SHA1 Message Date
TheTomClaude Opus 4.6
 		4a27f20956 Add quality & speed summary section to scripts README 
Top-level at-a-glance table with PPL, NIAH, speed ratios across all
tested hardware, compression ratios, and link to decode regression
analysis doc.

Co-Authored-By: tturney@psyguard.ai
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-03-26 23:43:52 -05:00
TheTomClaude Opus 4.6
 		d456030799 docs: add M1 Max and M2 Pro community diagnostic results 
First community diagnostic from @Ambisphaeric (M1 Max 64GB):
- Prefill: 93-98% of q8_0 (excellent, flat curve)
- Decode: 84% short → 39% at 32K (M1 constant LUT issue)

M2 Pro Mac Mini partial results (in progress, noisy from workloads).

Co-Authored-By: tturney@psyguard.ai
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-03-26 17:17:34 -05:00
TheTom a3a335d4a9 docs: add AMD/ROCm experimental warning — untested, FA broken on RDNA 2 2026-03-26 17:08:42 -05:00
TheTomClaude Opus 4.6
 		8533410eb1 docs: add fork table for NVIDIA/AMD — point to spiritbuun's CUDA fork 
Community testers need to know which fork to build. Apple Silicon
uses our Metal fork, NVIDIA/AMD uses spiritbuun's CUDA fork.
Also added flash attention requirement warning.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: tturney@psyguard.ai
 2026-03-26 16:58:26 -05:00
TheTom ce7af8b3f9 docs: add 'plug in charger' warning — battery throttling observed on M5 Max 2026-03-26 12:26:45 -05:00
TheTom f37f639ebd docs: add clone step to README quick start 2026-03-26 12:09:15 -05:00
TheTomClaude Opus 4.6
 		5bf0aef9d1 fix: 5 Codex blockers — zip safety, exit codes, multi-GPU, PII, skip-ppl 
1. Zip failure no longer deletes raw artifacts — preserves data if zip fails
2. main() returns 1 on unhandled exception (was always returning 0)
3. --skip-ppl no longer requires llama-perplexity binary
4. Multi-GPU nvidia-smi: sum memory/util, max temp across all GPUs
5. Model path logged as basename only (was leaking full path with username)
6. Build instructions now platform-aware (Metal vs CUDA)
7. Sharing instructions: GitHub issues + @no_stp_on_snek DM
8. Removed legacy bash diagnostic script (superseded by Python v5)

253 tests, 100% coverage.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: tturney@psyguard.ai
 2026-03-26 11:58:17 -05:00
TheTomClaude Opus 4.6
 		8c9d6d4726 feat: Python diagnostic tool v5 + venv launcher + README 
turbo_hardware_diag.py (2270 lines, stdlib-only core):
- 13 diagnostic sections matching bash v4 output format
- Cross-platform: macOS (Metal/Apple Silicon) + Linux (NVIDIA/AMD)
- Background monitor thread polling every 10s (CSV output)
- Real-time terminal: rich Live tables or ASCII fallback
- Anomaly detection (bad): decode degradation, thermal throttle, swap growth
- Notable detection (good): excellent ratios, near-parity at long context
- Outlandish detection (investigate): turbo3 faster than q8_0, PPL better
  than baseline, decode improving at deeper context
- Every probe wrapped in try/except — warnings, never crashes
- Output: .txt log, .json profile, .csv monitor, .zip package
- Tag format compatible with hw_replay.py parse_diag_output()
- Full argparse: --help, --model, --skip-ppl, --skip-stress, --verbose

turbo-diag launcher:
- Auto-creates isolated venv (.turbo-diag-venv/)
- Installs rich for terminal UI
- Never pollutes user's Python environment
- Falls back to plain Python if venv creation fails

scripts/README.md:
- Quick start, output format, how-to-read guide
- M5 Max baselines, anomaly flags explained
- Hardware replay system usage
- Troubleshooting, privacy statement

106 tests, 66% line coverage (main() integration path is the gap).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: tturney@psyguard.ai
 2026-03-26 10:43:12 -05:00