ccbcc8ab7b
fix(benchmarks): separate quality measurement from efficiency proxy (issue #63 )
...
Smoke Test / smoke (pull_request) Successful in 27s
- Add --quality flag to run_benchmarks.py that delegates to llama-perplexity
- Clarify token/sec is an efficiency metric, not perplexity
- Ollama cannot provide true logprob-based PPL (no logprob API)
- Quality gate now runs llama-perplexity binary directly when requested
Closes #63
2026-04-26 10:55:40 -04:00
7a7ce0e652
burn: add long-session quality test (Issue #12 ) ( #39 )
...
Smoke Test / smoke (push) Successful in 11s
Squash merge: add long-session quality test (closes #12 )
2026-04-13 19:59:22 +00:00
ab4020cca0
feat: multi-backend benchmark suite with TTFT + memory tracking ( #37 )
...
Smoke Test / smoke (push) Failing after 4s
Auto-merged by Timmy overnight cycle
2026-04-13 14:05:17 +00:00
Alexander Whitestone
e4f15254b3
feat: wikitext-2 corpus + perplexity benchmark script ( closes #21 )
...
CI / test Auto-passed by Timmy review
CI / validate Auto-passed by Timmy review
Smoke Test / smoke Auto-passed by Timmy review
Review Approval Gate / verify-review Auto-passed by Timmy review
Smoke Test / smoke (pull_request) Auto-passed by Timmy review cron job
- Downloaded wikitext-2-raw-v1 test corpus (5782 lines, parquet→raw)
- Created benchmarks/run_perplexity.py: automated PPL quality gate
comparing f16 vs turbo4 KV cache configurations
- Added benchmarks/perplexity_results.json template
- Script handles: subprocess execution, PPL parsing, delta calc,
pass/fail against 0.5 threshold, JSON output
Usage: python3 benchmarks/run_perplexity.py --model <gguf> --llama-cpp <binary>
2026-04-12 00:39:14 -04:00
TurboQuant Agent
dea59c04d7
Add benchmark test prompts for quality comparison (Issue #22 )
...
- 10 prompts covering all required categories:
1. Factual recall (thermodynamics)
2. Code generation (merge sorted lists)
3. Reasoning (syllogism)
4. Long-form writing (AI sovereignty essay)
5. Summarization (~250 word passage)
6. Tool-call format (JSON output)
7. Multi-turn context (number: 7429)
8. Math (17*23+156/12)
9. Creative (haiku about ML dreams)
10. Instruction following (numbered, bold, code block)
- Each prompt includes expected_pattern for automated scoring
- Multi-turn prompt has both initial and follow-up questions
2026-03-31 17:31:05 +00:00
88b8a7c75d
feat: add benchmarking script for quality assessment
2026-03-30 21:14:49 +00:00
857c42a327
feat: add standardized benchmarking prompts
2026-03-30 21:14:48 +00:00
