gemma4d-kv-cache-offload

name	gemma4d-kv-cache-offload
description	Use for Gemma4D KV cache, prefix cache, SSD cold tier, cache manifests, cache-key correctness, quantized/compressed KV experiments, and restore parity.

Gemma4D KV Cache and Offload

Trigger

Use for active KV, RAM prefix cache, SSD cold cache, cache manifests, q8/q4 KV, Planar/Iso/TurboQuant experiments, and restore correctness.

Rules

• No mid-decode SSD fetch in MVP.
• SSD cache accelerates future prefills, not live dense weight paging.
• Cache key must include model, quantization, tokenizer, template, adapter, KV layout, KV mode, and engine versions.
• Restore must be tested against fresh prefill.
• Compression modes require quality metrics and fallback.

Required output

• Cache mode.
• Block size.
• Namespace fields.
• Restore test result.
• TTFT/memory/cache hit evidence.