• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

stillwater-sc / universal / 23672108138
85%
master: 84%

Build:
Build:
LAST BUILD BRANCH: main
DEFAULT BRANCH: master
Ran 28 Mar 2026 12:09AM UTC
Jobs 1
Files 629
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

27 Mar 2026 11:40PM UTC coverage: 84.4% (+0.02%) from 84.384%
23672108138

push

github

web-flow
feat(ucalc): quantize command with RMSE/QSNR metrics (#644)

* feat(ucalc): quantize command with RMSE/QSNR metrics (#627)

Add `quantize <format> <data>` command for evaluating quantization
quality across number types. Supports both inline vector literals and
CSV file loading for production-scale tensor analysis.

Features:
- Inline vectors: quantize fp8e4m3 [0.3, -1.2, 0.007, 2.5]
- File loading: quantize fp8e4m3 -f weights.csv
- Metrics: RMSE, QSNR (dB), max absolute error, mean error (bias),
  clipped count (overflow), flushed count (underflow)
- Per-element table for small datasets (<= 64 elements)
- All output formats: plain, --json, --csv, --quiet
- Quiet mode: "RMSE QSNR(dB) elements" one-liner for format comparison

New files:
- data_loader.hpp: CSV reader and vector literal parser
- scripts/12_agent_quantize_comparison.ucalc

21/21 CTests pass on gcc.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(ucalc): resolve CodeRabbit review items for quantize command

Three fixes:

1. Strict numeric parsing in data_loader.hpp: parse_double_strict()
   validates that the entire token is consumed by std::stod, rejecting
   partial matches like "1.5xyz". Reports line number on failure.

2. Distinct file_not_found exception type so the quantize command maps
   file-open failures to EXIT_FILE_NOT_FOUND (4) instead of
   EXIT_PARSE_ERROR (1).

3. NDJSON clarification in script comment (nitpick).

Also fixes: comment lines containing semicolons (e.g., "# parse
line-by-line") were split by semicolons before the comment check,
causing the post-semicolon fragment to be evaluated as an expression.
Comment lines are now skipped before semicolon splitting in batch mode.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

44444 of 52659 relevant lines covered (84.4%)

6125603.99 hits per line

Jobs
ID Job ID Ran Files Coverage
1 23672108138.1 28 Mar 2026 12:09AM UTC 629
84.4
GitHub Action Run
Source Files on build 23672108138
  • Tree
  • List 629
  • Changed 3
  • Source Changed 0
  • Coverage Changed 3
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #23672108138
  • 74cd81a6 on github
  • Prev Build on main (#23670073104)
  • Next Build on main (#23678591174)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc