27B model validation of SmoothQuant + E8 stack

proposed high priority TODO-002

Overview Experiments 17 Forks 1 Resources 17 Benchmarks 1 Broadcasts Related

Description

The SmoothQuant-α + E8 lattice gain mechanism should replicate on a much larger model, but the absolute PPL gap may shrink (larger models have more redundancy → quant noise has more places to hide). Need to confirm the mechanism is model-size-independent before publishing

Reference

EXP-0014

Suggested Parameters

model qwen3.5-27b

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

eval_seq_len 2048

gpu rented-multi-3090-or-4090

Provenance

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