dank/serverAppConfigs
1
0
Fork 0
Code Issues 6 Pull Requests Actions Packages Projects Releases Wiki Activity

frank-small: pin VLLM_ATTENTION_BACKEND=TURBOQUANT

Browse Source 
Gemma4's config hook force-selects TRITON_ATTN for the mixed 256/512
head dims whenever no backend is selected, and TRITON_ATTN rejects the
turboquant KV dtype (measured: 'kv_cache_dtype not supported', crash
loop). The TURBOQUANT backend accepts any head size and the
turboquant_* dtypes, and the Gemma4 MTP speculator carries the
target's selected backend through to the draft layers.
This commit is contained in:
Frank
2026-09-19 00:07:57 -07:00
parent 5e581666f2
commit 966b70db1a
2 changed files with 13 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/frank-small.env
+8
View File
@@ -85,6 +85,14 @@ LISTEN_PORT=8000
MAX_MODEL_LEN=262144
GPU_MEM_UTIL=0.95
KV_CACHE_DTYPE=turboquant_k8v4
# Gemma4's config hook force-selects TRITON_ATTN for the mixed 256/512 head
# dims, but only when no backend is selected — and TRITON_ATTN rejects the
# turboquant KV dtype (measured: ValueError "kv_cache_dtype not supported",
# crash loop). Pin the TURBOQUANT backend explicitly: it accepts any head
# size, the turboquant dtypes, and has the Ampere path, and the Gemma4 MTP
# speculator carries the target's selected backend through to the draft
# layers, so the drafter follows.
VLLM_ATTENTION_BACKEND=TURBOQUANT
MAX_NUM_SEQS=16
MAX_NUM_BATCHED_TOKENS=16384
SPEC_NUM_TOKENS=4
tests/frank_small_unit.bats
+5 -1
View File
@@ -98,7 +98,7 @@ QUADLET_LARGE="$REPO_ROOT/dankclaw/frank-large/quadlet/frank-large.container"
! grep -q 'HALOGEN_VISION_TOWER' "$QUADLET_LARGE"
}
@test "frank-small serves the full window with turboquant k8v4 KV on a dedicated card" {
@test "frank-small serves the full window with turboquant KV on a dedicated card" {
local maxlen kvdtype util seqs
maxlen="$(sed -nE 's/^MAX_MODEL_LEN=([0-9]+)$/\1/p' "$ENV_FILE")"
kvdtype="$(sed -nE 's/^KV_CACHE_DTYPE=(.+)$/\1/p' "$ENV_FILE")"
@@ -113,6 +113,10 @@ QUADLET_LARGE="$REPO_ROOT/dankclaw/frank-large/quadlet/frank-large.container"
# (head_dim 256) and full-attention (head_dim 512) layers, and fp8 is
# hardware-dead on sm_86 anyway.
[ "$kvdtype" = "turboquant_k8v4" ]
# The TURBOQUANT backend must be pinned explicitly: Gemma4's config
# hook force-selects TRITON_ATTN (mixed 256/512 head dims) whenever no
# backend is selected, and TRITON_ATTN rejects the turboquant dtype.
grep -q '^VLLM_ATTENTION_BACKEND=TURBOQUANT$' "$ENV_FILE"
# The 3090 is dedicated to this service, so the engine may take most of
# it; the KV pool, not the seq ceiling, caps full-context concurrency.
[ "$util" = "0.95" ]