- name
SKILL.md
스킬 파일은 게시자의 저장소에 그대로 있습니다.
복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add artokun/comfyui-mcp --skill comfyui-launch-flags -a cursor -y복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add artokun/comfyui-mcp --skill comfyui-launch-flags -a claude-code -y복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add artokun/comfyui-mcp --skill comfyui-launch-flags -a codex -y복사한 명령을 터미널에 붙여넣습니다. 이 SKILL.md가 해당 에이전트에 설치됩니다.
npx skills add artokun/comfyui-mcp --skill comfyui-launch-flags -a grok -y복사한 메시지를 Grok Bot 채팅에 보냅니다. Bot이 GitHub에서 이 SKILL.md를 가져와 모든 Bot이 공유하는 스킬에 추가합니다.
Install this skill: https://github.com/artokun/comfyui-mcp/tree/HEAD/plugin/skills/comfyui-launch-flags게시자 제공
CLI flags passed to main.py control ComfyUI's runtime behavior
(e.g. python main.py --reserve-vram 2 --use-sage-attention). The three that
matter most for making a graph run rather than OOM or crawl are the
VRAM strategy, the attention backend, and the cache mode. This skill
is the decision matrix for choosing them.
⚠️ Verification note (August 2026). Every flag below was checked against upstream
comfy/cli_args.pyon current master. ComfyUI adds/renames flags often — when in doubt runpython main.py --helpin the target install and prefer that over this list.--enable-triton-backend/--disable-triton-backendARE ComfyUImain.pyflags on master (they used to be documented as SwarmUI-only; that is stale).--use-ck-attentionis kitchen INT8 attention — nosageattentionwheel. A June ComfyUI checkout still pins comfy-kitchen 0.2.10 and lacks--use-ck-attention;kitchenaction:"status" reports ComfyUI-side flag support, not only the kitchen version. Usekitchen/panel_kitchento see what this GPU can actually run.
How to apply today. The MCP's
restart_comfyui(withaction: "start") currently replays the exact argv of the previous run. It does not compose fresh flags. So set these when you launch ComfyUI yourself (thepython main.py …line, arun.bat/shell alias, or the SwarmUI backend args box), and the tool will preserve them on restart. Injecting flags through the tool is a tracked follow-up.
Symptom ▶ Flag(s) to try
─────────────────────────────────────────────────────────────────────────────
CUDA out of memory, long video (LTX 2 / WAN) ▶ --novram (+ --cache-none)
OOM, still want models resident when they fit ▶ --reserve-vram N then --disable-smart-memory
GPU slows to a crawl, spills into "shared GPU ▶ --reserve-vram 2..4
memory" (Windows WDDM) mid-run
RAM blows up switching between models, or a huge ▶ --cache-none
text encoder (FLUX 2 / Mistral) won't unload
Plenty of VRAM (48GB+), want max throughput ▶ --gpu-only or --highvram
Want faster sampling on NVIDIA ▶ --use-ck-attention if kitchen INT8 is available (skip the sage wheel); else --use-sage-attention
Z-Image produces BLACK / wrong output ▶ --use-pytorch-cross-attention (NOT sage)
Sage gives black output on some models ▶ --use-pytorch-cross-attention (or fix dtype)
ROCm, kitchen present, triton ≥ 3.7 ▶ --enable-triton-backend
VRAM strategy and attention backend are each mutually exclusive groups, so
pass at most one from each. You can combine one VRAM flag + one attention flag +
one cache flag (e.g. --novram --use-sage-attention --cache-none).
Modifiers (combine with the above):
--reserve-vram N reserves N GB for the OS and other apps. It is the fix for the
Windows failure mode where the GPU quietly starts using shared VRAM and
throughput collapses. Typical 2 to 4; bump to 10 for heavy video decode.--disable-smart-memory forces aggressive offload to regular RAM instead
of keeping models cached in VRAM. Reach for this when a run gets stuck or
OOMs intermittently. Slightly slower, much more reliable.--async-offload enables async weight offload streams (default on where
supported); --disable-async-offload turns it off if it misbehaves.Two gotchas worth memorizing:
--use-sage-attention; you get black or garbled output.
Launch Z-Image with --use-pytorch-cross-attention instead. See
z-image-txt2img.--use-pytorch-cross-attention, or (SwarmUI) set
Advanced Sampling → Preferred DType = Default (16-bit). Sage-on vs Sage-off
also produces slightly different images, so expect non-identical seeds.When a graph hard-crashes with
No module named 'sageattention'/triton: unavailable, the fix is the sdpa / no-compile fallback intriton-sageattention, not this flag.
--fast enables experimental, potentially quality-degrading
optimizations. Accepts specific PerformanceFeature values:
fp16_accumulation, fp8_matrix_mult, cublas_ops, autotune. Bare --fast
turns them all on. Test output quality before committing to it.--fp8_e4m3fn-unet, --fp16-unet, --bf16-unet, --fp32-unet, …) for
forcing a compute precision. Usually the model or loader picks the right one, so
only reach for these to work around a specific dtype error.Long video OOM (LTX 2 / WAN, 24GB): --novram --cache-none
(add --disable-smart-memory if it stalls)
Windows shared-VRAM creep: --reserve-vram 3
FLUX 2 / huge text-encoder swaps: --cache-none
High-VRAM throughput (48GB+): --gpu-only (or --highvram)
Fast NVIDIA sampling (most models): --use-ck-attention (if kitchen INT8 is available)
--use-sage-attention (otherwise; needs the wheel)
Z-Image (any): --use-pytorch-cross-attention
ROCm + kitchen + triton ≥ 3.7: --enable-triton-backend
Cross-refs: video OOM specifics in
ltxv2-video / wan-t2v-video;
per-model VRAM math in troubleshooting and
model-compatibility.
The attention/compile accelerators are version-locked to your exact
torch + CUDA + Python. A mismatched wheel doesn't just fail to import; it can
break the torch install. A known-good, mutually-compatible stack for late-2025 /
2026 NVIDIA (including Blackwell / RTX 5000, sm_120) looks like:
Operational facts worth carrying:
TORCH_CUDA_ARCH_LIST=7.5;8.0;8.6;8.9;9.0;10.0;12.0+PTX spans RTX 20xx→50xx
and datacenter (A100/H100/B200). +PTX lets newer archs JIT.~/.triton / %USERPROFILE%\.triton and temp)
when you hit stale-kernel Triton errors after an upgrade.uv pip install over pip for the venv. Resolves and downloads are
dramatically faster. install_comfyui already supports this via preferUv.troubleshooting.--use-ck-attention, --enable-triton-backend, --disable-triton-backend, --fast); hardware gates in comfy/model_management.py (supports_fp8_compute SM ≥ 8.9, supports_nvfp4_compute / supports_mxfp8_compute SM ≥ 10.0); kitchen backends in the comfy-kitchen README https://github.com/Comfy-Org/comfy-kitchen--enable-triton-backend is retracted as of ComfyUI master.Pick the right ComfyUI startup flags for VRAM, attention, caching, and speed. The full decision matrix for OOM (--novram / --cache-none / --disable-smart-memory), shared-VRAM creep on Windows (--reser
---name: comfyui-launch-flagsdescription: Pick the right ComfyUI startup flags for VRAM, attention, caching, and speed. The full decision matrix for OOM…---# plugin/skills/comfyui-launch-flags/SKILL.md개요
Pick the right ComfyUI startup flags for VRAM, attention, caching, and speed. The full decision matrix for OOM (--novram / --cache-none / --disable-smart-memory), shared-VRAM creep on Windows (--reserve-vram N), model-switching with big text encoders (--cache-none), high-VRAM throughput (--gpu-only / --highvram), and attention-backend selection (--use-sage-attention for speed, --use-pytorch-cross-attention as the highest-quality / Z-Image-safe fallback). Also the acceleration-stack + Blackwell/RTX 5000 (sm_120) notes. Use when a graph OOMs (especially long video like LTX 2 / WAN), when the GPU spills into shared VRAM and slows to a crawl, when switching between models eats all RAM, when Z-Image produces black/garbled output under Sage, or when deciding which attention backend to launch with. Flag names verified against upstream comfy/cli_args.py; see Sources.
FAQ
게시자에게서 plugin/skills/comfyui-launch-flags/SKILL.md를 다운로드하거나 경로를 복사해 저장소를 엽니다. PlugMCP는 스킬을 목록에 올릴 뿐 파일을 호스팅하지 않습니다.
게시자가 https://github.com/artokun/comfyui-mcp에서 관리합니다.
Claude, Cursor, Codex, Grok Build는 게시자의 저장소에서 추가한 SKILL.md를 읽을 수 있습니다. Grok Bot은 채팅에서 스킬의 GitHub 폴더 링크를 보내면 설치합니다.