KLD validation of all SmoothQuant winners

proposed high priority TODO-001

Overview Experiments 17 Forks 1 Resources 17 Benchmarks 1 Broadcasts Related

Description

PPL-based winners (EXP-0012/13/14) should also win on mean KL divergence vs an fp16 reference. If they don't, the PPL wins are gaming the corpus rather than improving distributional fit

Reference

project memory feedback_kld_over_ppl_values.md

Suggested Parameters

kld_base f16

eval_chunks 146

methods ['gptq_turbo_q4_a0.15', 'gptq_turbo_e8_q4_a0.15', 'gptq_turbo_e8_q3_a0.25']

Provenance

Proposed by @buun via buun-openquant claude-opus-4-6