serverAppConfigs

Files

T

Frank e0329caef8 Add FASTGREEDY mode to serve script: reasoning off + argmax fastpath on for fast-greedy bench

FASTGREEDY=1 flips the serve script to the card's default profile:
- reasoning off (no logit budget overhead)
- LLAMA_TARGET_GREEDY_ARGMAX_FASTPATH=1 (exact-Q8 GPU argmax)
- deterministic sampler (temp 0, top_p 1, top_k 0, min_p 0)

This mirrors the card's 'fast greedy (default)' benchmark config.
Production (FASTGREEDY=0) unchanged.

2026-08-22 12:54:03 -07:00

bench-kairic-ab.sh

bench: add prod2048 config (runner batch sizes on chat profile); docs: ubatch 4096 results

2026-08-22 03:17:41 -07:00

bench-spec-stack.sh

bench: add Kairic A/B config harness; make v3 spec-stack harness self-contained (survives the Kairic env swap)

2026-08-22 02:53:32 -07:00

deploy.sh

docs: document frank-default-large v3 upgrade (Output-K8 runtime, full-W8 views, BF16 vision); update AGENTS/deploy payload sizes

2026-08-21 20:18:57 -07:00

frank-default-large.env

Add FASTGREEDY mode to serve script: reasoning off + argmax fastpath on for fast-greedy bench