fix: post-PR-A compose-path fixes for gpu-mode.sh + 2 patch READMEs

PR-A (#231) inserted the <quant>/ layer; two compose-path references
outside the doc sweep were left stale:

- scripts/gpu-mode.sh (the rig mode-switcher, symlinked into
  /usr/local/bin) cd'd into the dual/ topology dir and -f'd bare
  filenames (turbo.yml, docker-compose.yml, ...) that PR-A moved under
  <quant>/. Repointed DUAL_27B_DIR / GEMMA_DUAL_DIR at the autoround-int4
  quant dir (mount-safe — same cd-into-compose-dir invocation switch.sh
  uses), added GEMMA_DUAL_AWQ_DIR for the awq mode, and renamed the two
  ex-docker-compose.yml defaults (qwen -> fp8-mtp.yml, gemma -> bf16-mtp.yml).
  All 13 mode targets verified to resolve; bash -n clean.
- vllm-pr40798 / vllm-pr40914 patch READMEs cited dual/int8-tq3.yml (the
  pre-PR-A name). Commit 6182922 split int8-tq3 into the tq3-* family; the
  READMEs describe "TQ3 KV + MTP n=3" = tq3-mtp.yml (whose tombstone header
  cites #40914) -> dual/autoround-int4/tq3-mtp.yml.

Surfaced (not fixed here — PR-B's registry-derivation eliminates it):
launch.sh LAUNCH_VARIANT_COMPOSE[vllm/gemma-mtp] points at a non-existent
dual/autoround-int4/fp8-mtp.yml; gemma dual default is bf16-mtp.yml (the
registry is correct). test-launch-compat doesn't sweep launch.sh paths.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

models/qwen3.6-27b/vllm/patches/vllm-pr40798-rebased/README.md

@@ -1,7 +1,7 @@
 # vLLM PR #40798 overlay — TurboQuant max-workspace pre-allocation
 
 Vendored 2026-05-11 to unblock `turboquant_3bit_nc` KV + MTP on Qwen 3.6 27B
-(`dual/int8-tq3.yml`).
+(`dual/autoround-int4/tq3-mtp.yml`).
 
 ## Source
 
@@ -59,5 +59,5 @@ reports `MERGED`. Then bump the nightly pin past the merge commit.
 ## Verified on
 
 - vLLM nightly: `1acd67a7`
-- Compose: `dual/int8-tq3.yml` (Qwen 3.6 27B AutoRound INT4, TQ3 KV, MTP n=3, 262K, 2 streams)
+- Compose: `dual/autoround-int4/tq3-mtp.yml` (Qwen 3.6 27B AutoRound INT4, TQ3 KV, MTP n=3, 262K, 2 streams)
 - 2026-05-11

models/qwen3.6-27b/vllm/patches/vllm-pr40914-k1-only/README.md

@@ -80,7 +80,7 @@ rebased before merge so it doesn't revert #41434).
 ## Verified on
 
 - vLLM nightly: `1acd67a7` (includes merged #41434)
-- Compose: `dual/int8-tq3.yml`
+- Compose: `dual/autoround-int4/tq3-mtp.yml`
 - Qwen 3.6 27B AutoRound INT4, TQ3 KV, MTP n=3, 262K × 2 streams
 - 2026-05-11
 

scripts/gpu-mode.sh

@@ -9,8 +9,12 @@ set -e
 # both deprecated 2026-05-10 — supporting services moved into services/).
 CLUB3090_DIR="/opt/ai/github/club-3090"
 COMPOSE_BASE="$CLUB3090_DIR/services"
-DUAL_27B_DIR="$CLUB3090_DIR/models/qwen3.6-27b/vllm/compose/dual"
-GEMMA_DUAL_DIR="$CLUB3090_DIR/models/gemma-4-31b/vllm/compose/dual"
+# Post-PR-A (<quant>/ layer): dual composes live under <topology>/<quant>/.
+# Point each var at the quant dir so `compose_at` cd's into it — mount-safe,
+# the same invocation switch.sh uses (project dir = compose-file dir).
+DUAL_27B_DIR="$CLUB3090_DIR/models/qwen3.6-27b/vllm/compose/dual/autoround-int4"
+GEMMA_DUAL_DIR="$CLUB3090_DIR/models/gemma-4-31b/vllm/compose/dual/autoround-int4"
+GEMMA_DUAL_AWQ_DIR="$CLUB3090_DIR/models/gemma-4-31b/vllm/compose/dual/awq"
 
 # Estate planner state file (v0.7.0+). Instances booted via launch.sh --estate
 # or --estate-file are tracked here and persist via Docker `restart:
@@ -71,11 +75,11 @@ stop_service() {
 # Project-specific helpers
 start_27b_dual_mtp() {
     printf "  ${GREEN}▲${NC} Starting 27b-dual-mtp..."
-    compose_at "$DUAL_27B_DIR" "up -d" docker-compose.yml && echo "done" || echo "failed"
+    compose_at "$DUAL_27B_DIR" "up -d" fp8-mtp.yml && echo "done" || echo "failed"
 }
 stop_27b_dual_mtp() {
     printf "  ${RED}▼${NC} Stopping 27b-dual-mtp..."
-    compose_at "$DUAL_27B_DIR" "down" docker-compose.yml && echo "done" || echo "skipped"
+    compose_at "$DUAL_27B_DIR" "down" fp8-mtp.yml && echo "done" || echo "skipped"
 }
 
 start_27b_dual_dflash() {
@@ -127,11 +131,11 @@ stop_comfyui() {
 # --- Gemma 4 31B dual-card serving variants ---------------------------------
 start_gemma_mtp() {
     printf "  ${GREEN}▲${NC} Starting gemma-mtp..."
-    compose_at "$GEMMA_DUAL_DIR" "up -d" docker-compose.yml && echo "done" || echo "failed"
+    compose_at "$GEMMA_DUAL_DIR" "up -d" bf16-mtp.yml && echo "done" || echo "failed"
 }
 stop_gemma_mtp() {
     printf "  ${RED}▼${NC} Stopping gemma-mtp..."
-    compose_at "$GEMMA_DUAL_DIR" "down" docker-compose.yml && echo "done" || echo "skipped"
+    compose_at "$GEMMA_DUAL_DIR" "down" bf16-mtp.yml && echo "done" || echo "skipped"
 }
 
 start_gemma_dflash() {
@@ -163,11 +167,11 @@ stop_gemma_dflash_int8() {
 
 start_gemma_awq() {
     printf "  ${GREEN}▲${NC} Starting gemma-awq..."
-    compose_at "$GEMMA_DUAL_DIR" "up -d" awq.yml && echo "done" || echo "failed"
+    compose_at "$GEMMA_DUAL_AWQ_DIR" "up -d" bf16-mtp.yml && echo "done" || echo "failed"
 }
 stop_gemma_awq() {
     printf "  ${RED}▼${NC} Stopping gemma-awq..."
-    compose_at "$GEMMA_DUAL_DIR" "down" awq.yml && echo "done" || echo "skipped"
+    compose_at "$GEMMA_DUAL_AWQ_DIR" "down" bf16-mtp.yml && echo "done" || echo "skipped"
 }
 
 # Stop every Gemma serving variant before starting a new one