Prepare isolated Dirk Qwen3.8 benchmark

experiments/dirk-qwen38/README.md

@@ -0,0 +1,49 @@
+# Dirk Qwen3.8-27B experiment
+
+Isolated A/B test environment for
+`peculiar-ragdoll/Dirk-Qwen3.8-27B-GGUF`. It deliberately does not add a
+production router profile and never stops or restarts production services.
+
+## Candidate
+
+- Main model: `Dirk-Qwen3.8-27B-UD-Q4_K_XL.gguf` (about 17.6 GB)
+- Vision projector: `mmproj-F16.gguf`
+- Runtime: existing `mike-ai/llama.cpp:local`
+- Test endpoint: `127.0.0.1:5004`
+- Results: `/data/benchmarks/dirk-qwen38/`
+
+The candidate is intended to reduce unnecessary reasoning and total token use;
+it is not expected to improve raw decode speed. The production Qwen model is
+therefore the mandatory A/B reference.
+
+## Safety boundary
+
+`run-case.sh` refuses to start while any production `mike-ai-llama-*` model
+container is running. It does not stop production itself. The model server is
+bound to loopback only and cannot be reached from the LAN.
+
+## Prepared matrix
+
+Run the following only after the GPUs have explicitly been declared free:
+
+```sh
+./run-case.sh 80000 90,10 text
+./run-case.sh 160000 90,10 text
+./run-case.sh 160000 85,15 text
+./run-case.sh 160000 80,20 text
+./run-case.sh 192000 85,15 text
+./run-case.sh 262144 80,20 text
+```
+
+The largest stable context is determined first. Vision is checked only after a
+text winner exists:
+
+```sh
+./run-case.sh 160000 85,15 vision
+```
+
+For every case, record uncached prefill, cached prefill, decode throughput,
+GPU memory, context recall, tool calling, code quality and total tokens needed
+to finish the task. Do not promote Dirk unless it matches the base model on
+technical correctness and improves real Hermes task completion.
+

experiments/dirk-qwen38/download-model.sh

@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+
+MODEL_DIR=${MODEL_DIR:-/data/models/qwen3.8-27b-dirk}
+BASE_URL=https://huggingface.co/peculiar-ragdoll/Dirk-Qwen3.8-27B-GGUF/resolve/main
+MODEL=Dirk-Qwen3.8-27B-UD-Q4_K_XL.gguf
+PROJECTOR=mmproj-F16.gguf
+
+install -d -m 0755 "$MODEL_DIR"
+
+download() {
+  local name=$1
+  local target="$MODEL_DIR/$name"
+  curl --fail --location --retry 8 --retry-delay 5 --continue-at - \
+    --output "$target.part" "$BASE_URL/$name"
+  mv -f "$target.part" "$target"
+}
+
+[[ -s "$MODEL_DIR/$MODEL" ]] || download "$MODEL"
+[[ -s "$MODEL_DIR/$PROJECTOR" ]] || download "$PROJECTOR"
+
+(
+  cd "$MODEL_DIR"
+  sha256sum "$MODEL" "$PROJECTOR" > SHA256SUMS
+)
+
+printf 'Prepared model files in %s\n' "$MODEL_DIR"
+

experiments/dirk-qwen38/run-case.sh

@@ -0,0 +1,105 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+
+CONTEXT=${1:-}
+SPLIT=${2:-}
+MODE=${3:-text}
+
+if [[ -z $CONTEXT || -z $SPLIT || ! $CONTEXT =~ ^[0-9]+$ || ! $SPLIT =~ ^[0-9]+,[0-9]+$ ]]; then
+  echo "Usage: $0 CONTEXT TENSOR_SPLIT {text|vision}" >&2
+  exit 2
+fi
+if [[ $MODE != text && $MODE != vision ]]; then
+  echo "Mode must be text or vision" >&2
+  exit 2
+fi
+
+MODEL_DIR=${MODEL_DIR:-/data/models/qwen3.8-27b-dirk}
+MODEL=/models/qwen3.8-27b-dirk/Dirk-Qwen3.8-27B-UD-Q4_K_XL.gguf
+PROJECTOR=/models/qwen3.8-27b-dirk/mmproj-F16.gguf
+IMAGE=${LLAMA_IMAGE:-mike-ai/llama.cpp:local}
+NAME=mike-ai-llama-dirk-test
+RESULT_DIR=${RESULT_DIR:-/data/benchmarks/dirk-qwen38}
+
+[[ -s "$MODEL_DIR/Dirk-Qwen3.8-27B-UD-Q4_K_XL.gguf" ]] || {
+  echo "Dirk model is missing; run download-model.sh first" >&2
+  exit 1
+}
+if [[ $MODE == vision && ! -s "$MODEL_DIR/mmproj-F16.gguf" ]]; then
+  echo "Vision projector is missing" >&2
+  exit 1
+fi
+
+mapfile -t blockers < <(
+  docker ps --format '{{.Names}}' |
+    grep -E '^mike-ai-llama-' |
+    grep -vE '^mike-ai-llama-dashboard$|^mike-ai-llama-dirk-test$' || true
+)
+if ((${#blockers[@]})); then
+  printf 'Refusing to start: production model container(s) still running: %s\n' "${blockers[*]}" >&2
+  exit 1
+fi
+
+docker rm -f "$NAME" >/dev/null 2>&1 || true
+install -d -m 0755 "$RESULT_DIR"
+
+args=(
+  --model "$MODEL"
+  --alias qwen-dirk-test
+  --ctx-size "$CONTEXT"
+  --flash-attn on
+  --cache-type-k q4_0
+  --cache-type-v q4_0
+  --cache-prompt
+  --cache-ram 8192
+  --threads 6
+  --threads-batch 6
+  --batch-size 64
+  --ubatch-size 32
+  --parallel 1
+  --jinja
+  --reasoning auto
+  --reasoning-budget 8192
+  --host 127.0.0.1
+  --port 5004
+  --metrics
+  --fit off
+  --n-gpu-layers all
+  --no-mmap
+  --temperature 1.0
+  --top-p 0.95
+  --top-k 20
+  --device CUDA0,CUDA1
+  --main-gpu 0
+  --split-mode layer
+  --tensor-split "$SPLIT"
+  --spec-type draft-mtp
+  --spec-draft-n-max 3
+  --spec-draft-type-k f16
+  --spec-draft-type-v f16
+)
+
+if [[ $MODE == vision ]]; then
+  args+=(--mmproj "$PROJECTOR" --no-mmproj-offload)
+fi
+
+label="ctx${CONTEXT}-split${SPLIT//,/-}-${MODE}"
+docker run -d --rm \
+  --name "$NAME" \
+  --gpus all \
+  --network host \
+  --read-only \
+  --tmpfs /tmp:rw,noexec,nosuid,nodev,size=256m \
+  --security-opt no-new-privileges:true \
+  --cap-drop ALL \
+  --pids-limit 1024 \
+  --log-opt max-size=20m \
+  --log-opt max-file=2 \
+  -v /data/models:/models:ro \
+  -v "$RESULT_DIR":/results \
+  --label mike-ai.experiment=dirk-qwen38 \
+  --label mike-ai.case="$label" \
+  "$IMAGE" "${args[@]}"
+
+printf 'Started isolated case %s on http://127.0.0.1:5004\n' "$label"
+

experiments/dirk-qwen38/stop-case.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+docker rm -f mike-ai-llama-dirk-test >/dev/null 2>&1 || true
+echo "Dirk test container stopped. Production was not changed."
+

experiments/dirk-qwen38/wait-ready.sh

@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+
+deadline=$((SECONDS + ${READY_TIMEOUT:-900}))
+until curl -fsS --max-time 3 http://127.0.0.1:5004/health >/dev/null; do
+  if ((SECONDS >= deadline)); then
+    docker logs --tail 100 mike-ai-llama-dirk-test >&2 || true
+    exit 1
+  fi
+  sleep 2
+done
+curl -fsS http://127.0.0.1:5004/props
+printf '\nDirk test server is ready.\n'
+