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
ab5ae173c2
Merge pull request 'PolarQuant Implementation & Phase 2 Integration Plan' ( #18 ) from feature/polarquant-implementation into main
2026-03-30 23:49:52 +00:00
9816cd16e8
Merge pull request 'Benchmarking Suite: Objective Quality and Performance Testing' ( #19 ) from feature/benchmarking-suite-1774905287056 into main
2026-03-30 23:41:37 +00:00
e81fa22905
Merge pull request 'feat: Sovereign Evolution Redistribution — turboquant' ( #20 ) from feat/sovereign-evolution-redistribution into main
2026-03-30 23:41:11 +00:00
51a4f5e7f5
feat: implement Phase 19 - Hardware Optimizer
2026-03-30 23:27:28 +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
5f9f316f2c
Add implementation plan
2026-03-30 21:06:51 +00:00
2bd7354eed
Add ggml-metal-turbo.metal implementation
2026-03-30 21:06:50 +00:00
3705c332ac
Add llama-turbo.h implementation
2026-03-30 21:06:49 +00:00
2bcd36f7c5
Add llama-turbo.cpp implementation
2026-03-30 21:06:49 +00:00
Timmy
10f720b500
Full KT report: Phase 1-3 complete
...
12/16 issues resolved. turbo4 validated. Ollama deferred (llama-server
is production path). Per-layer adaptive found built-in. QJL assessed,
not needed at current compression targets.
Ref #1
2026-03-30 17:05:23 -04:00
Timmy
441f4ee765
Phase 1 Report: PolarQuant MVP complete
...
turbo4 KV: 73% memory savings, -1.1% prompt speed, -11% gen speed.
Metal shaders verified. PolarQuant checklist 5/6 PASS.
128K context on 36GB hardware is viable.
Closes #4 #5 #6 #7 #8
2026-03-30 16:12:01 -04:00
Timmy
cefaa6e778
Add build spec v2.2 and README
...
TurboQuant KV cache compression for M4 Max local inference.
Spec by Strago, triaged into 16 issues across 4 phases.
Ref #1
2026-03-30 13:11:45 -04:00
0b62c72737
Initial commit
2026-03-30 17:08:45 +00:00
