+ #jE Rt^RIHt^RIt^RIt^RIt^RIt^RIt^RIt^RI t ^RI H t ^t ^t ^ t^t^ t^tRRRRRR R RR RR R /tR-RRlltRRltRRltRRltRRltRRltRRltRRltR.RRlltR R!ltR"R#ltR$R%/R&R'lltR(R)ltR/R*R+llt ]!R,8Xd] PD!] !44R#R#)0uhardware_check.py — Detect whether this machine can realistically run ComfyUI locally. Improvements over v1: - Multi-GPU detection: scans all NVIDIA / AMD GPUs, picks the best one (most VRAM) - Apple Silicon: detects Rosetta-via-x86_64 false negative; warns instead of misclassifying - Apple generation: defaults to None (unknown) instead of mis-tagging as M1 - WSL2 detection: identifies WSL2 + nvidia-smi situation explicitly - ROCm: prefers `rocm-smi --json` for new ROCm 6.x output - Disk space check: warns if /home or workspace volume has < 25 GB free - PyTorch verification (optional): tries to import torch and check device availability - Windows: prefers PowerShell `Get-CimInstance` over deprecated `wmic` - More accurate VRAM thresholds and verdict reasons Emits a structured JSON report. Exit codes match `verdict`: 0 → ok 1 → marginal 2 → cloud Usage: python3 hardware_check.py [--json] [--check-pytorch] ) annotationsN)Anynvidiaz--nvidiaamdz--amd apple-siliconz --m-seriesintel comfy-cloudcpuz--cpuc$V^8dQhRRRRRR/#)cmdz list[str]timeoutintreturnstr)formats"h/opt/hermes-venv/lib/python3.14/site-packages/../../../skills/creative/comfyui/scripts/hardware_check.py __annotate__r6s!i#cc\P!VRRVRR7pVP;'gRVP;'gR,# \\P \ 3dR#i;i)TF)capture_outputtextr check) subprocessrunstdoutstderrFileNotFoundErrorTimeoutExpiredOSError)r r outs&& r_runr#6sdnn 4    bSZZ%5%5266 z88' Bs-AAA A21A2cV^8dQhRR/#r rboolr)rs"rrr@s   rc\P!4R8wdR#R\P!4P49g)R\P!4P49dR#\ RR4;_uu_4pRVP 4P49uuRRR4# +'giR#;i \ dR#i;i) z;Return True when running under Windows Subsystem for Linux.LinuxF microsoftwslTz /proc/versionrN)platformsystemreleaseloweropenreadr!)fhs ris_wslr3@sG#h&&(..00EX=M=M=O=U=U=W4W /3 ' '2"'')//"33( ' ' ' s00C!B0% C0 C ;CC CCcV^8dQhRR/#r%r)rs"rrrMsDrc\P!4R8wdR#\P!4R8XdR#\.RO4P 4pVR8H#)zMReturn True when Python is running translated under Rosetta on Apple Silicon.DarwinFarm641)sysctlz-inzsysctl.proc_translated)r,r-machiner#stripr"s r is_rosettar=MsFH$W$ : ; A A CC #:rcV^8dQhRR/#r r dict | Noner)rs"rrrXs$${$rc\P!R4'gR#\.RO4pVP4'gR#.pVP4P 4FpVP R4Uu.uFq3P4NK pp\ V4^8dKB\V^,4pV^,p\V^,4p\ T4^8d T^,MRpTPRRRTRTR \TR , ^4R T/4K V'gR#\VR R 7p \ V4^8dWR&V #uupi \dKi;i)zIDetect NVIDIA GPUs. Returns the GPU with the most VRAM, plus list of all. nvidia-smiN,rvendorrindexnamevram_gbdrivercVR,#rGr)gs&rdetect_nvidia..ys1Y  eAh-C8D%(mG!Z!^q h S D uWt^Q/ f   )$  t/ 0D 4y1}Z K/5   s:E '-E E! E!cV^8dQhRR/#r?r)rs"rrrs,,[,rc  \P!R4'gR#\.RO4pVP4P R4'EdG\ P !V4p.pVP4Fwr4VP R4'gKVPR4;'g5VPR4;'gVPR4;'gRpVPR4;'gVPR 4;'g^p\V4pVPR R R \V4P4R \VR, ^4RR/4K V'd%\VRR7p\!V4^8dW'R&V#\.RO4pVP4'gR#\$P&!RV4p\$P&!RV4p V 'd,\\V P)^44R, ^4MRp R R R V'd VP)^4P4MRR V RR/# \\3d^pELAi;i \ P"dLi;i)rocm-smiN{cardz Card seriesz Card modelzMarketing NamezAMD GPUzVRAM Total Memory (B)vram_total_memory_brDrrFrGrIrocmcVR,#rKr)cs&rrMdetect_rocm..s) rrOrQz,Card (?:series|model|Marketing Name):\s*(.+)z VRAM Total Memory \(B\):\s*(\d+))re--showproductname --showmeminfovram--json@)rernrorp)rRrSr#r; startswithjsonloadsitemsgetrrW TypeErrorrXrrYrZrVJSONDecodeErrorresearchgroup) r"datacardscard_idinforFvram_braname_mvram_mrGs r detect_rocmrs. << # # S TC yy{c"" ::c?DE!% ))&11/CC488L3ICC88$45CC9B"9:bbdhhG\>]bbab [F eCIOO-uVw%7;f ". 5&<=u:>',$  I JC 99;; YYF LF YY:C @FI+I+ I&A I+3#I+I($I+'I((I++JJcV^8dQhRR/#r?r)rs"rrrskrc 2\P!4R8wdR#\P!4R8wdR#\.R O4P 4p\ P !RV4pV'd\VP^44MRp^p\\.R O4P 4;'g^4pV'd\VR, ^4MRpRpRF pW`9gK TpM RRRT;'gRR VR VR V/# \dLUi;i)r6Nr7z Apple M(\d+)rmrDapplerF Apple Silicon generationvariantunified_memory_gb)r9-nzmachdep.cpu.brand_stringr9rz hw.memsizerr)UltraMaxPro) r,r-r:r#r;rzr{rr|rWrY)chipmr mem_bytesram_gbrvs rdetect_apple_siliconrsH$W$ < = C C ED /4(A$%QWWQZ4JI ;<BBDIIJ 1:U9(! ,sFG $ 9G % '''j7V      s.D DDcV^8dQhRR/#r?r)rs"rrrs  + rc\P!4R9dR#\P!R4'd+\ RR.4pRV9dRV9gRV9d RR R R R R /#\P!4R8XdG\P!R4'd+\ .RO4pRV9dRV9gRV9d RR R RR R /#R#)r(WindowsNclinfoz--listIntelArcXerDrrFz Intel Arc/XerGrm powershellzIris XezIntel Arc/Iris Xe>r(r)r -NoProfilezHGet-CimInstance Win32_VideoController | Select-Object Name | Format-List)r,r-rRrSr#r<s rdetect_intel_arcrs 44 ||HHh'( c>u|ts{gv~y#N NI%&,,|*D*D`a c>u|yC/?gv/BIsS S rcV^8dQhRR/#)r rfloatr)rs"rrrsUrc\P!4pVR8XdA\\\ .R O4P 4;'g^4R , ^4#VR8Xd~\RR4;_uu_4pVFYpVPR4'gK\VP4^,4p\VR , ^4uuuRRR4# RRR4VR8Xd\P!R4'dX\ .RO4p\P!R V4pV'd,\\VP^44R , ^4#\ .RO4p\P!R V4pV'd,\\VP^44R , ^4#R# \ dR#i;i +'giL;i \dR#i;i)r6rmr(z /proc/meminfor+z MemTotal:Nrrz(\d{8,})z(\d{6,})rrri)rrz:(Get-CimInstance Win32_ComputerSystem).TotalPhysicalMemory)wmicComputerSystemrwTotalPhysicalMemory)r,r-rYrr#r;rWr0rsrUr!rRrSrzr{r|)sysnamer2r\kbr"rs rtotal_system_ram_gbrs|ooG( T"@AGGINNQOSZ[]^_ _' os++rD{33 a1$R7^Q77 ,+,) << % %C +s+AS_8!<<KL IIk3 ' QWWQZG4a8 8 3  ,+   sS>F4%G9G4G GGG4 GG G GG G('G(c V^8dQhRRRR/#)r pathrrrr)rs"rrrsS5rc\P!V4p\VPR, ^4# \dR#i;i)rHrmrr)rR disk_usagerYfreer!)rusages& rtotal_free_disk_gbrs?!!$'UZZ7+Q// s25 AAcV^8dQhRR/#r?r)rs"rrrsKrc^RIpRRRTP/p\TPP 44TR&TR,'d<TPP 4TR &TPP^4TR &\TPPP 44TR &T# \dpRRRRT 2/uRp?#Rp?ii;i \d RTR&Lfi;i \d RTR &T#i;i) zJOptional PyTorch availability check. Only run when --check-pytorch is set.N availableFreasonztorch not importable: T torch_versioncuda_availablecuda_device_count cuda_device_0 mps_available) torch Exception __version__r&cuda is_available device_countget_device_namebackendsmps)rers rcheck_pytorch_cudarsL T**D'!%ejj&=&=&?!@   ! !(- (?(?(AD$ %$)JJ$>$>q$AD !& $U^^%7%7%D%D%F G_ K# LUH0Fqc.JKKL '!& ' & %_ K&sAB:A0C0C1: CC CCC.-C.1DDc 0V^8dQhRRRRRRRRRRR R /#) r gpur@rr free_disk_gbr*r&rosettarztuple[str, str, list[str]]r)rs"rrrs@Q&Q&+Q&uQ&EQ&4Q&Z^Q&c}Q&rc.pV'dVPR4RRV3#V'd(V'd VR,R8XdVPR4V'd(V\8dVPRV R\ R24V'd;V^8d4V'd,VPR4R 8wdVPR V R 24Vf(VPR 4VPR 4RRV3#VR,R 8XdVPR4pVPR4pVPRR4pV'dRV 2MRp V'd V RV 2, p V\8d5VPV RV R\ R24VPR4RRV3#V\8dVPV RV R24RRV3#VPV RV R24RRV3#VR,R8XdVPR4RRV3#VPR R4p VR!,p V \ 8d5VPV R"V R#\ R24VPR$4RRV3#V \ 8d$VPV R%V R&24RVR,V3#V \8d$VPV R%V R'24RVR,V3#VPV R%V R(24RVR,V3#))uDetected Python running under Rosetta on Apple Silicon. ComfyUI MPS support requires native ARM64 Python — install via `brew install python` or arm64 Miniforge, then re-run.cloudrrDru[Detected WSL2 + NVIDIA — confirm `nvidia-smi` works in your WSL distro before installing.zFree disk space (z GB) is below the zo GB recommended minimum. ComfyUI core (~5 GB) plus one SDXL model (~6.5 GB) needs space; Flux Dev needs ~24 GB.rz System RAM (zm GB) is low. ComfyUI swaps model weights through host RAM; <8 GB causes severe slowdowns. 16+ GB recommended.zTNo supported accelerator found (NVIDIA CUDA / AMD ROCm / Apple Silicon / Intel Arc).uRCPU-only ComfyUI works but is unusably slow for modern models — use Comfy Cloud.rrrrmMr z with u! GB unified memory — below the z GB practical minimum.zBSD1.5 may work; SDXL/Flux will swap or OOM. Recommend Comfy Cloud.u9 GB — SDXL works but slow. Flux/video likely too tight.marginalru* GB unified memory — good for SDXL/Flux.okruZIntel Arc detected — ComfyUI IPEX support is experimental; Comfy Cloud is more reliable.rGrFz has only u GB VRAM — below the z5Most modern models won't load. Recommend Comfy Cloud. (u; GB VRAM) — SD1.5 works, SDXL tight, Flux/video unlikely.uA GB VRAM) — SDXL comfortable, Flux possible with optimizations.u6 GB VRAM) — can run everything including Flux/video.)rXMIN_FREE_DISK_GBrwMIN_MAC_RAM_GB OK_MAC_RAM_GBMIN_VRAM_GB_USABLE OK_VRAM_GB GREAT_VRAM_GB) rrrr*rnotesgenrmemgen_strrprFs &&&$$ rclassifyrsE  E  u,, ss8}0 rs '77  ~-?@P?QRe e &1*):g)E 6(#K K  { b   `  u,, 8}ggl#'')$gg)3/"AcU)  7)} $G   LL)6#&GGWWmn  LL] ^M50 0   LL)6#&_` 5 5 yse+UVW_e++ 8} qr7E)) 779c "D v;D    fJtf$;RR.pRVRVRVR VR R R R R RRRRRRRRVR\4/ p V'd\4V R&V #\!WsWFVR7wrpRVRVRVR VR VR VR VRV RV R\"P%V 4RVR\4/ p V'd\4V R&V #)~Nr6r7u0Intel Mac detected — no MPS backend available.zFComfyUI will fall back to CPU which is unusably slow. Use Comfy Cloud.osarch system_ram_gbrr*Frrverdictrrecommended_install_pathrcomfy_cli_flagr install_urlspytorch)r*r)r,r-r:rrrr expanduserr=r3rbrrr _install_urlsrr_COMFY_CLI_FLAGrw) rrrrrrr*rrreportr install_paths $ r build_reportrrsooG    D "F%bgg&8&8&=>LlG (C    =    !      {w(*tww > T  ' D V L 5 u 4 w & d U MO    2 4F9  #+ \G$ G5 g  s7 s7"L/--l;  F.0y MrcV^8dQhRR/#)r rrr)rs"rrrstrcRRRRRRRR/#)desktopz+https://docs.comfy.org/installation/desktopmanualz2https://docs.comfy.org/installation/manual_install comfy_cliz0https://docs.comfy.org/comfy-cli/getting-startedrzhttps://platform.comfy.orgrrrrrrs$@FG-  rc V^8dQhRRRR/#)r argvzlist[str] | Nonerrr)rs"rrrs3 3 3 33 rc ^RIpVPRR7pVPRRRR7VPRRR R7VPV4p\ VP R 7pVP 'd$\\ P!V^R 74EM\R VR , RVR, R24VPR4'd \R4VPR4'd \R4\RVR, R24\RVR, R24VR,'dVR,pVR,R8Xd+\RVR, R VPR!^4 R"24M\RVR, R VPR#^4 R$24VPR%4'd7\VR%,4^8d \R&\VR%,4 R'24M \R(4\R)VR*, R+VR,, 24VR-,'d\R.VR-, 24VPR/4'dVR/,pVPR04'duR1VPR24 2pVPR34'dVR4VPR5R64 R2, pVPR74'd VR8, p\V4M\R9VPR:4 24VR;,Fp\R8Xd^#^#)?Nz3Check whether this machine can run ComfyUI locally.) descriptionrq store_truezEmit machine-readable JSON only)actionhelpz--check-pytorchz5Also probe `torch` for CUDA/MPS availability (slower))r)indentz OS: rrr)r*zEnv: WSL2rz3Env: Rosetta (x86_64 Python on Apple Silicon)z RAM: rz GBz Free disk: rz GB (~/)rrDrz GPU: rFu — rz GB unified memoryrGz GB VRAMrQz (z GPUs total; using best by VRAM)zGPU: (none detected)z Verdict: ru → rrz, run: comfy --skip-prompt install rrz PyTorch: rrz + CUDA (r?rz + MPSuPyTorch: not available — rru • rr) argparseArgumentParser add_argument parse_argsrrrtprintdumpsrwrV) rrr]argsrrLptr\ns & rmainrs,abANN8L7XNYNN$\OQ << D (:(: ;F yyy djj*+ F4L>F6N+;1=> ::e   # $ ::i G H F?34C89 F>238<= %==u A{g% AfI;eAEE:Mq4Q3RRdef AfI;eAEE)Q4G3HQR55$$Qz]);a)?LQz]);(<<\]^ . / F9-.fV66*++i(D'EQGGD66/**H$Dd 5bffX6F5GHIA F1#, !iD  iJ& r__main__)).)N)#__doc__ __future__rrtrr,rzrRrsystypingrrrrrrrrr#r3r=rbrrrrrrrrrr__name__exitrrrrs,#        j 7\ T4 7  $N,^>  B0Q&h<%<~3 l zHHTVr