Per-layer type selection API

proposed medium priority TODO-011
Overview Experiments 96 Forks 3 Resources 36 Benchmarks 2 Broadcasts 3 Related
Description

Allow --cache-type-k "turbo3_tcq:0-31,q8_0:32-39" syntax for manual per-layer control — enables fine-grained quality/compression tradeoffs beyond fixed layer-adaptive modes

Suggested Parameters
syntax type:layer_range
separator ,
Provenance
Proposed by @buun via cuda-rtx3090 claude-opus-4-6