From 9f066cb55bcebe9ffff41259a966663e9b875fe2 Mon Sep 17 00:00:00 2001 From: Yihao Wang <42559837+AgainstEntropy@users.noreply.github.com> Date: Sun, 10 May 2026 21:32:15 -0700 Subject: [PATCH] [Docs] Add MiniCPM-V 4.6 cookbook (#24876) --- docs_new/cards/logos/openbmb.png | Bin 0 -> 25727 bytes .../autoregressive/OpenBMB/MiniCPM-V-4_6.mdx | 565 ++++++++++++++++++ docs_new/cookbook/autoregressive/intro.mdx | 6 + docs_new/docs.json | 6 + .../minicpm-v-4_6-deployment.jsx | 379 ++++++++++++ 5 files changed, 956 insertions(+) create mode 100644 docs_new/cards/logos/openbmb.png create mode 100644 docs_new/cookbook/autoregressive/OpenBMB/MiniCPM-V-4_6.mdx create mode 100644 docs_new/src/snippets/autoregressive/minicpm-v-4_6-deployment.jsx diff --git a/docs_new/cards/logos/openbmb.png b/docs_new/cards/logos/openbmb.png new file mode 100644 index 0000000000000000000000000000000000000000..e6bb0e98002792c5cd34881f815df0049e9aa65e GIT binary patch literal 25727 zcmeFZXH=70_b-YK-5}!L3Mxeu3jzYtB{UULYQTh=P*gf0K}vv7ZGeLG5<|6+K!6Zh z0t5)Co8E*#0trQ>2_YaQARurN-#f;6#~tH8?l>RLr}KP!R#sN#nrp2&e{;^?%1d+L zT>(B3J{}$(0Yih^ARe9r#XLOw1&;9Uwg8SU)$;H--!Z&>(;|q;*;)^AHVha=QKtuy zcjx}bq}djmI=^f}$R}r9P1g2|G3EYZEo;%V|4|ir*@ek@n{MSJCya*p_Pi2!JQmXZ z>wdtg!0HP;|0e(4xCR!K(Hr-f1f2DoMarz!mZnAd+x`=81FcjJG}>%XOa9Q*zB zex9J{iEriNI_F$CzDBl!Disw8K424X?ff;}xEB7y!}6<7^0FhoBiHAL)~%<$JvC&6 znDFrIDy{KrXsK=k4um&1;KRg(YP_A|JxoU(1;qqBV?<1-sGNu09ALh!ZR=3U6)cFx zIhO_PZwNPpPj?7Ia?ZY76%ak#@$W_XJ#`{Lz%E7Yu6>>ubY7~hKoJK2`JA)`BP*33g6mmGVz4$%ODvj&Y>9n{nbw*HbmU%J& z_QO60K1eW`>Yl6`uP)=u=ERbPOp_czT=X$rVq6poy8?*e~IfOm*tx&GHvxKcK!1H?_%vBlW z54DG=rTs8*N|sYkfeUo96`IHr_wHtvU%7}^-M#V!lX{N1WjWmuQ!~ES@_NpOB!7AdVL5gqIaV_F_T+{XAkbS>brn)-`2B`p8 zE%9R1K$$TZnf9ygD-ez^k>M6==E0~YQ*@)e%^s_AF|$HS7^|}o+e%WAVvCF)rVi*X zAblZC<4US;ASPt%sBZ{_8)P3>>vx|OLH_#rK zi=E)E9Vi|O-SPIH1ueJvRpQNu1*%?gfpl~j$rNS3H$3!Sl4evyKU=^!*9LWxF|6g% z0`O}0mmVggXEtmOK9Gy2ewm3jb_f@Zmt`Uf`aSbq!QNPN`u(ww@a*ez9O~ZT{?hz9 zlhto<2`N(9cc-uII5P|3jF~O&6o)I!B3G7BU13ktqvOCn;N^_5^I*N{zCv2kDT3vM z$VE6LtMI%RYV4YH?2Jlmd>Y{8%@z+pTBU8>MFvoJSd7?_eQmSE0({C8oqhnWnn<6z zC+|@(?s7Rz529}K19Z;x+e&o+ACL|p;_dCAk6Sx6=A4q@@-6tiZ3G0G1)x*$t>(--31RuhvZx8*RQtx0Qc-o1$%YlvGhz$1SN|@_9@Dt+es4;=+vBY1A=5;B7(@+ zn7q5j?U5Mn^?nT-sMTUiI?7(O__J!yrOhLyD_U!T`q?^LGF*wBWv8vN?VtQx>yA3T z*G>{b3E#obGNVAUPUG|oP!4+T#^d4K*K3DO>!UWsj^u9&<_0eTEk8(0;@aTmD?Qw} z!=hVrfwx3V8aCO1$jbSX_iCB(A3WDrb=&(JB7)!TwA^)^Z3z37Ms5B4ISulFRMI>y zmMKtG?S}J$ejDsZ1^sO!=c$}8eLUkT+%v{yXg_XJR2SiRUtCi>B%>3g_m_)$tM6xu z;EamDx8({nv%kzvVpO~D{Z*O2V^q#4KmBtY{(^LxNPo(^0;3>&|s;jNjyEE-3!8X$kDNN`>23ICq$qr-`n%QGik-g%z z$K<5=rB1)Dsih}Eqjd% z3-l&kwip5^-g7zh+pB%aXc-W>tW>3Edu3(Dt!E`ASBpMAEm#>I_llpHL5>S%FC;8MULs}yOP#fBt5-Z9d@-`Kt& z%Sg{w_M1hn&K79r=S11r+)?!WSaSPkV8Y3C!v5dR_)$I0lbmgR*kf3_-;b~SldJhr z5CaJz+T*7|lo;Ykz23xRv-20L@(;Sjo*383Tu%?Q0a%!hoOsA8k&Sr^9FSvciflju z;y{0;qujqk#l{p%=|>`tN$C=-XN?&BsgH4;j1aj~15W5;6ESwL84TPzdc->L zOWJ}aVl1SGKfc!`ffX*GVa4?`mt8O(>WDI*F^vIQ#qb7Z7g!TNo2wtN%9|i>8t(gT zo*%Q*tXX_*?%Gda?S-D~+%%Ia%O!&%8);v2!?q|FwU-{*Zt@3u=l#22X;;0r^KIpV z(%)Uh22{bHbf#y%l%H2?_4m{lb!=;gI2lWc>cPozc3uUvm!?XvZ2yW#O_<>8>BI9C z=LZ~)+c-A8j7(kin)X}uDYifIw4%EODVQsJCEW5s+dlZ3_TMfKS^0E84Y z|0;|x6E@Af)puXNJ`I2zG{KO3uSacpi9c&;%xD*J_KUcdb>S;@A^>-_=+nU9a$)5u z?fGA?b7L8HXg1x|L}ZpSW@7Dn?-)*u-d^Hf+86I6b8Qdxs6QRLj;-2qW7W;qN*eOH z++k6goo!R6NS}&W!^@Ya;Z_a{l_NwaqaUai5>H0*MF5_Od1JfFrf5WAem`@7a4}TknZ3b{&u9J~|z!2$yn8I>u6ZLhl zchiK5z%yy)TbHs}wrH5sz-`#Mfu#wLim8-Xum+YK|C zhJNq>|wQYQ4SkMGWTf)S;{9j15c5ZBg*T?GZ?{Hg!+a>dmwa2 zqyfmoVzcT7;3SbSB+%19*ev z(V4$#m-!)MEP|36vk?zo7G15laO96CG5H8_P>Bqs6!Ox|A;cdJ&m!&QbN3gG{5+DR zwqw&_^+_qgrPlepa^}-;8yPK69j^PDdsn1a(o~;|2iK<$_b3of4ZMuSI4ee57?!`K z+6%RTJ@BLD@o>?1!`(Z74g?@p7=p$~LI#aGIJ2WCg$<561xVa|IGEDyX){t%o#MV2+ATdFAwtNkXL8tq9!ZW#a#-|% z!sZIeWQL5dTqUj(8t*vgsZ4gvxFsjF8;9&daD>eg6Eg|Te=jWE5#W7a*`@s5BlQSz zDl+6g+vI+V0Jp_OX8d&i$ssHM_&_<0T#ITTzb(t~pC&EY!Db`aYnA?mH~IKdUy(gC zC3+f!PLe@)@SC2M?~p9A_rU?HohFaVIlHtoA6l_u@M?m~<&y3ky2?N`K)phS(0-j@ z&25UlU|;0&iJoCJ5ZTDlCVS-;-n8_P83&Ad+)t67`J;%PN=w>@l?JhVBBaakLHQ%; zP(`_lk}J-Ce|^ZGt|mhLUS;<_)jL;cP~%>rYs~5eef|cV(yCo~|L3U&rbQkJHZa>R zY@dhYwQ85{gt%tP*ZTz=;(3g2^j75g0X%bxzuD^avdifX7vPbywpOKToE@l_I$~Kh z&Qa;hX_wTO(w`Oq&^~ty^lw)?orsY}3=j}ZAN5c(x?*QEFNPQMFhr!(*G+^Qj}H>k+ogMd)eBGgXT`BFfZZFkS4!WT?PKz2_-cZ#brnu&0pWgEi_@8K=ak zxESuN7WN1xa@hWFjfF@J^~(D5y8+H++4Jh?5ckh`C3R3U{mOiEhlhyIsxP)vZ|1!E zKw}27?1%r;8{^DxbpcP1$CfnI-XRvsfeG@-MmQ}!m*LgBD!ciZ-`<^9aS|hZb6M>w zMzrZs;R{o?h}IqHF=iF=3}o44r9OV08V_|f`1JDObt`*VM(lWZvn;wY*%3&`8%6(jl2njN)kDJ5%JKO}q;!GNMdzgwz3qbRwLGZ1eYoI_Nko#-hFVYZF+j&c7bd?I1 z%N%jOl$Jv{ZacL?(KM(%;EEQXVQ4Wo9sESmm8c9NCN<1p>rT;We`v{tK%`O41~YER z>Q^lPpD-GIHd0?fChUM_{ z)~Mq$e8aP}7yEI+)4nOMV=p68Fein5MOA+K)$cs|Eb*&5WJ4n2Cv|>|ohtHa*0+%R z>-AV2iQe#Q^>A+9wVV_ z9D6!JPj@3zY9l%#2$wQ_K1H^ZwbG@fTT|zJe?(@`{+st}5gX2-oBm_JhGGxXJIxE# z-UjD*+px!6;wkn?j#)d8p_~f zlyu@ty)?&|gEy=_v;EKts6ac|fl-toZ5S|2mMN~?J(HWKdY*$#Iu2B%1eS>98yJ&L zEd7+RucsuQ^{?XAA6{E2T4QWbjRln_hUle`q3!Luf2F(tn_*%*VsK^gF;Sc8{fj|3 zx9y$49qaST<7*nfL-2E@RPiU4%wuoLgm#$el3RB)2pyry3Rioc=PctM*Sn^Xqobgl84ba)or&>OCp<^E}~_qHRKo97kAdj4pFniCYVz>YdiSBJ;99 zgSL|hrv+*fs_9@oH&c~GtPw(+j#TL7NcU)1=95wmyV4AgbOgAbXHm;tgBid5xz2tLs?Z-%pPSL@TZm-RXQ8mafztvZ@-VI&Cd{0ODR)_TW5Qmk}rTb@N%YrjqLz-c*@3R4VG<~*SbE2(_w z`?vD4i4(4fSYzXaOEi zSU1|w9xC@Eh)jvS5srnlyv{Jf)+m{lgIeULqBBHwvCh4PpFppmt2WvHK#Nty+hcP_ z2ICes5mnD%X+*Uzs_4a#H`8IBBFsQoo@2jWMx`n;Ed^2)j9mhTmS=DD`d#I6A?*>#Adbo+69jF}~ zJsmOSR2m!Vu{P{nn@~mkSbwBroQKpVG)0$TEC;(-8zi;+T$|%K96I*n6zUtbstI?y<)F>k_Jb?qw4U_M1-qfXaO3)k4P^enkl zLQkSyh*izTlQP6Od3s}OQHy+5f%wxBu$I+|ie}}&!+mq^5WZv2Lk3r?^4)~?0;Qnr z&`X$5w_sKG7fRbL$d0u!PTM6tl z5;Zi_E2mnQC03E^YlaP0qK*tS-#jlJZBh0PF$po0;SfD(rK$qgSvTScZ$6sj`-&R} z%X_@fB6}}dY@Hd0+FAqz%Dk)1?4y5NDPR79*sw>u@uZCpSv;tp1B{KdWNw~V$PCpg z$%-c+3zwDqr-qP5LDM)=jTHidnl->7g0X+%kPMbOy#M43oIm-UdA`Eg=KveD6EwzAlECtp6&7(rq@YD+EeZ`dCHVK)fS5AI!Q^<9{l z^C-f2l?^#GVWNDt{JLXfd&Fxkel~5pXO#$V|67!g>t7vYk@^ z>=N4B`LHuqdMwovvYc^%QA!MQa+n`*BS^sK`dokml}SX?2<8rAkJK;4R%>UteycBhogq$Ex!W!@$HuSYimi z3-hr*-+vZoc7dClll6UPz4z1x&8(qc*Na!2sL9Rl&`O=1)G+8Z^tsp+Ty&l`+DI7~ zTK~){)z4OU{BYMb_DL6i*g_}Jg`9+)n34Mt&yvBWJ~LWf`i^W+-eOjOx9(98BMsQQ zuHGv}yuWc24|V>FXEG!u7xHA;gVB+@{p;oSNW)I#_S(+Eom^d1-UDuqmyL(=tCYKA z@f%jmCtj#Y|CGZgU@{jaJcau0w-)i#pzXArUo%eYo}UA$BQw3xRTrVsABC)(WsESq z#sa=!bs7UZixoZGhKOG_5vnJ{$dn?gW-=fJLX*XNPtr*iluz)aHIue5pVrW{q=k10 zjQB;eyk%k9xR)uuS$ejR)?K07OVRC}&ef(7e%bAGj?su?we}LyU@temy}AR*RWtMR zghj+M`9X)9W7aHjd(}cB0%=_}A69?}yGu>B@_8fmnQ@WHzsRr)7G=NCFda#wq*ZFz zvfXmU*~IO6%voKXw1^;RwiTa{3$s!#5!!x9d+6v!CNYOC)q4 zoU!$OMd6r3D%FJud>YiJlaoFYgOev0T>Nm8VR;<<7qkW|xMac30C~ zb+5SP#JA>Bgn9Q0KJ9L>c0yLXl2yqr9N0ywTXN&gKc!f+98W$+>G~8QSdHWgnM5g5 zU)KT;#;74FSLzL@$)O}69L`w7?9DtAODmxnO6gmtUdC)R`P${`MA`*>(_TghBDrCr zRpn)bluJ$$Us1BT*`7p9XRHN4f0gn?0Ke~dIjT_!IA}G87q?Rehdqyaprf=bUifIT zD7?BDj67vIa=yHKB8`(xs8wIdki9CTq3zk>%USs_XdttT4*RLi=E0} z@usf!xg8mndwU(>Zeu@ZR4!WXi*a6aIcaxASs~z7cEp@Zv)@{GIHyIB=SP4O(d%~& zZ?B?Pch*UtbwzwHRr9;;|4bdx&M+f2*_m^A&tmcf!Ct-HEkph;;6-BxB!-5b=3?_q z3=LI(eBqfoc}vilHgg%eUHPE?TZ&OkQ?4FEbwBt8@8dwdmgBpw5IV&?adZ}m`=;+; z@;4Wv3>8~UuYVclF>UnsOye{86gN?o1-p5<>Z#SYs4qwet}gqa0mf_kJ4bi2TeV}c zN84LzO)IAU3%N)P&D=rs=pr`HG*?{YyjH=0(k}-}cQUI%3A}rmsXIf*3UFSG{ulRT zgGwAOMc0V9ua zfh*OMnx9}CIKjSJ3%;Oyg3LW|jS9wKjIrZ=qtLdi6YM?cogfqG9(!%Sqo9m6g7k zmsXk6l#1l|;n_gaq3{dTKL@E^k6lgy@J3m9CQc}6hnDscLy6ztC**!#tS}SGTywrF z3y&(ecMNFZQizG{9ZVf#mB%va?thD2CNOXdl@n1?E^u~Of5+1yv#+jK%dA>@NUeO1KQ zEKgC_n{N*B2A~ke%6eRne)wKex}Nd=@U!sgz~95P2(Em659NzY&G!|ZkO6Z_^DIto zJgVaMB#c;SDdh(_oH0G|%ZC}* zPQI+rU}8h^%U$a<{D;*qfz_%Do3R_(UsP`GK`UA~HNzMUc{8>{57}59f%FzXXEYm@(L)5SHx_nk4{vZIHf{(e?AqUKJNcAoX-JN2GCgS{sy2M{D|}?BU$+Ll z(%W`NN?u>F275KZQxA8Q^f3c_h*sgv^ga5HQ$NTbMbM5Ps^eogTkjlcVYo!=zO%hD zAPpbM0Iq-IX5Nx^Y4tOgn%Sn%0wqQmMBTdg-b1SgT_M-@HL{{t4y;c~PG>5%zhpH``F2$x09hUoqI*=N zuRjx}TB59_z7i?`vr(v5<0`T0v+5H(yi(@cRrE8L*=Ls!@|yXeDW^j;m27mI#qBFf zo*1S*0tuP3A$6BwN?6^^0JL~Q?CA9DxDRIBi|LINi4mTFB;Z%O_#?H{`i?6tLk1#> zg24y&U?Srpi!4ni|GWxrzLrMTZ+`>4?9D2CZIV7jwzDm<6=VI7|1_smoOQyLtkflW z+!e-Ong5nH70d2tqN<@7SxaZEcIUNY=r%cL`hJ3- zTye%nlisJ99Tr|1A#0Ly7{7&kEJhsbTG|fv|skAZT}C6)uD>ZKFFf( zCZAZPVXcr*6sw#U1CN`~|2U&xm{eO8w@xXcR0>1bCOu?gfnDz{Ivr}~)!<~Mmr;ZW z1Us%(H`Au6wHOX3(?RU7WfQM4fpX2-YU;WB)Yp@5iHq~gj9E`e_I2&~16uJ|TSD0dRCJ6jd!(c&iXd1!ycMtO+`0g+8J+^5dbx+CY_qu`|8 zn7g9}pZ%D+tX6Kf`YKG{+j4Y=F5YB-_yr&xnx0e8DQ@e1x8p8Kgy3tW`1v{}G`u&-VX9uG~_75xC$)NTR^|HSZ_oY9G4Z*Vm~5m8K6*`aYj$m(9WqA2YF3%pM_r9(UaF2nFDBi>J7IKy zG_$i3?1dbyfm3?KxB*U(=LM-uQ3H}k#rRPt==RtB3S&%_5v>!`a0oOrR>g~uj?onD zx4SsFsdRBcv_|{WEZp?RwJwWesAIEz{tcRaRSP2n#9`=6$mNlq*ZYeyb9d;=jZ#^s zHZOZMNYPLoAX5c{r9Z*6%O_8`*omR4Mhg9unO?-wbk9n}PGIhAPsEDF)*@tH7T(wV zv=6(VT~3Vme^b@#oBidRkFU4PU_(M7iQ_u;1sNcV8Ls%Y{PD@a>}|`6n~Gv=Q9-?k zAD_TmjyscwH@YOmX2Q`i4+`R9inCB93ZM<$9IurY-5tA~h6G)Pg74>bLViIWRd!oISA?}Ke@ z=IRbKYzgLmH`}I_QGd;m>8m%eVl&m^;QacWiku&X&{x1GUmtmeE`|UUe5u2$S1{6k zU~5o*zT<2Kk!sPfek(BEZfuht9sHxz1lo{F=5&ZD*h@(4ju^gOv;8wgfjh3VG7wC# z8%r)uDNl(i^8xqMYcH{Kl(^f_t~c@)GM%Sh^`WHpIn3B5RS-3x>hDYHRk^B(cfU_d z;fOjYT&>ZTRiIgeZ>@naSAH7%8NGArp-*p*Zm0Y|U3>hKMp=+bz~t8V_ih^<4MD{U z)M~6|;3$JLUsC5k>uYwbOsyX5R)kf^6kU=NecQ?WyZWR@y{*Tx7B@k6Yd6dtC$6jv z>CJo<#|TMJUp6t$%|)!B9mBD?UTOeph;_g;xxF_@Qd2snGKpef`4>DEu zAU*cgisWB8-I>iR-j@7Na+&yP}j-oi12 zN0j~Ka)q?U_n$$Z+x*&=)2-O7nDf}d);f2Cl#l5ni6?97fj=y#v#LBhZ_QqChIkB& zFdi-4?nyYDjR-bOhBD{}86GuGbAK83NOCdQoMP)3R#jRNNhww`_GGeK|A)>3p}F%c zw>-eqCglo8v_MNCXz~eb^!}rPchVLtn<{w^)i%2_$V|U_ut8yn22eOuyK(QC7Z}i#JUiZb zl58icFCVDRRTeFMo)?#w0#Yfqr|Kk!Nb#rS8|oE4x48>hvYNM4! zcT*10%6PC>NZ^t`vzcQsiM_K+HrYN@7kO52G_ZsW2gkjIeMqR5hBIHG-YRq#U zxKY-rv!_$NXIe2elBQ?$@krP>JN8CxLkTANW6UjmjQ$z@v-nr3k$I(*oIy|iyhj5q z)Y}!d-2IBJ!bUq@`str7>EKjVx3v!M@)yM@zTAjJTbcc1ctX`ZkyqJ9+h>6`*`I#B@fugQvT&lP$bIhD3Kf>_*`Vpydqc#RKa{Up zi%2LSqoap}$(bPAoL40b)p zaB;V%TeqlL)eyDKWkqpacMI%~p?O6!_f0=+FP6g{oF<*AEqmu#JEjL?jo1h*)Tn;B z!?^Sxs={}O7rTb!6cgFEvC%`QYkL00<-1vhm86zbA9K~8U6!12mJbv43WDX*ojo#6 zKy}|x(mGU)-AToePaD2PqA3$_SqWqP&9BhIdmGf$xDZgu?^ZN1R1Hv0o$-4RJ3ISu zE@1P5LUUzg38kulbnHW)IYZaeM=wTlZW znX(9|Q`3WYD0TvP;bNV7>DgC}<@E1rKj_2aD>pB*t^aWEG?p4u=^4iM=~#!e2V@^0 zBaT(*Y0S=vS!GgGhqOwe9*D9jF5p;2mv=YQOuoSxG-8bx@4R}Sr?Kc?XK|jO>09h| zYv$;Mm&ZFwDc(Op(dl+m-;znWVdKH|?;+&%&qye?W}gkgMtM@~Pjsij{O8IKK+pbH zg!G>|`~F|v%isyf!)S%0TFVd!RYH{jUJyYqt;7Ey`)HwN8k*+>@LlMtQSk7Zf?C3JpIS~T&t9UM4A))g)(sNcZjdz zCh99^g#h>d!Vkh-D7+kxbD@{3?YuAJt6{C?N9!e{wHI`mH_|lTxfZx5x@;XxO;*?s zcfF!5WDAQyfPxJkFm*KVYxcOF`laP1HOb+x-26DLA75*emv%8hxWd@rtZ&hnzB0O| zolDU({~Cs+e{9%gh3_uC-9l-a62@+Hs;c|R#Z*4{MQ!V!aIVYy{+Gk{I>X5V zoX;yB0#gCrIAP;v--lGM;}p-JV9dBfVm0%sJ-^}K+mu4&rZ*A;QtkNN~9be zsnk}pCPZ?@e$0JBqkbAlI zaWw%zm!1Im*AB<;&t)ZYWfuG}LC-9juQvuvVPXxl zd522xt;kLWUs{L(?YjH7YI4%^0m&m50mAg`0n@Hsqp}=z(USGBkx< ze>ZTYeqnUC(s6p_SKEonU^zcdYM|t~xigsy zf*YfO^RuHIuK%lxLnq~@qpK_CSH5nsf=f|_W@fpvq~1i?$t86n2Dc%^U$j^%VR%dT z)eC0O7LF1DYBXt5n&om|9SUBcEV$@BL`$)9^KNN=0R`V{AymlF+F{S{rJ{gU*#i0m`(E8rtP z+ApdD%viG_Ax8a3mWiKCJoMZ90U+VN>;cGF#KM#C>968pUy?T*zT*9JJxM$~c84OO zd8*02OzVk_(>QZA1HirSd69!GzLCDgM1YRCzb_@ph5nh6`BDC#BZ7zMRh5FYfcy3) zkKx|ot^J=dq4j;)mc4MIlTGEldj>K_gp*6CZ(elfcb}*nY3@}}%lYh#%3duW^4zvd z291v%xUIk>JV2xkp027%!W)n>>|(5@C~kf4kBZ}oA-vF}qLgKJGkYuRm6F|LX$h%>v$Y3cHyfRPIfWqWz;n3+@<-r zc-GvDNLpn&9IhQXR5L0f1K&W6Q6okW_gy5ce%qz};HUe(*l#rX%ZApN$ku?Qt^a-V z3NXe9%^C&pzN-6`QrF)8U4LO&#NDAyRaBIxd*n+W1N70f%#}vC1 zyM>Wq)05K}Me3YrJGy!y^Sb3fOvG#+#S$c1{gA};ED?dS05(sh|JdeX8`-l?mKi5! zg?G+W;fE5Pk*hdcXy9oA2G0H`wD*R8Q??uKmq+Y5^0W?EtAfAb9e+zbr=ca2o$w?CiKV8O-eqaR@=D8Os zm-lB}CoyzNLb9nv)Sp0VlkFexURO8)h}HsEecxhzOc}5d!GK#BN_UDvX{ZMpVgA#T z3$Y#w8qYTCLJ~}IohJHNwVYX2DqRv=b;s29u8LQ|SF-9EK1023a5?NkKHW>2ZTMt` zDNs#~s~P_>>9Aur9yVl{gPQ6=1u2a3Yrj#bz(4f4^--=0Vm0}0vOhEa8=j*BA_uid zFDdPXOGJlFxMqty(e@LAuJGNe_8ShAgmq>u9p+PVsZN}bs-sUm`mXWb5^cE%JyBYw`AELjv3@EqgBj?0 z@zfaUQ@-h+ds-clilTaq?K2cryGmwySJKTIaG7yPlY&%w!SK=GG9QKOScw_Jqulrb7hjG2k*eq2Oc%N`j zS(sCi{-maBmtHe1U6DF0M6Qpj?5FV(o-Irs$l#zlT!LE^c19g%rYX||{n)+|=G@Rl z>2F5=fr;`NDaXW0i}Nt{tBMZakdt*$il^2;6c0dXP=)uvXS3lr(#_s$@UUc-V9;m^ zr(O%FGMcmBGh`ub6uFtZ=q0^AN%wH&w}5AR!AAP=1+bJvy|?GtFQXr3;Fv&IvMDG^ z2`nn4jb_{QmF1C|UGH%_5wQ~EO{QaySRo6By$vnI4f!%95HMe4`k_GMGJ3~_SkTJV*B9sCn= zhBcD{b>#CiCQ+M96B_$MBEFnGFme2Rg1O<$*DVrtjZ?cpR}i0BEv%23OOvD1!<=VV z?-W-wOkm*S=Jk~6Y0gny~n9XqNCZl;K^0^=%K0PxGuyB<*JKpwQNrPlq?2c?WJ-2gZgE z%$E5d|5F2{93CKe)H~S~(Cll~)YtPRbQulnYrFDq&49OlEnnbn8;p+7aARFTdDh$xKM>_u;6y5SX;(?25c5_P}*()0}Qax!7D_h=Uy=x1_({?blM%Ug| zNS+eIS0r-0BZ|GZ*PV7(Vm54xw#M*fGYy{%^L>$~q4PzzCfyuoyd}I>CcY>6 zv#E7Zd`IzSJEXw5T}kz|tk=+&nRo9u$^X`Hz>*|6au;6NtLDSPpeUqw4WOgltGk5o{sd3l}S)7DMP$sM)&88-7hDOgD0lNI%Ppv&w} zm6{VWf%$yi1SRc9EGk=JB}kL#q;A)6g=fLxY%c5sX)0p`f)FHWXwYdX;Os2=mZd#% z^Ie(Ru&tCj(KJnRr4pUvNqB&?F53ko16x6aRgtrre+xhYTEk8qV0XhR-XhrV3pQsn zOrba0W(^;kM!kC;ng)oA)9LB22v;89U``(0W2~s6Ya3Sb2RsUB1l!DmY0&ZkalEVu zNJ*d%sQG>{#sEJ!^SCY)_qlG;Y(lIEXZWw$!ge_lnrTR>>xn$;jT-(|zFHSx|3#;q zBH`rxnb;cNB@|S8#dat8-8_4%FN2n#H$1O{($FR zi61Woq9p^Rf^&w7UE42f)_9z`qZfkvXL^1-yT0f2kzI|3Y}YL;Edg|c?D%a+*>E;* z7IOvvpdchmIJ1h459GtXiwfe8jM=4`H6drmMaCR(agyror#-RPJ$HuvGf%9KT6NYk zhPGKt+%R-^Tysa?y1(n&`Ub_)!hjqgjKxeXpuJH}gUFijE!T&=CH|uP@o6kE$n?}5 zy0DvTf6SLUW!G2Ak&{;Zjzv0Y55UK{S^4b{^6b}#G4o5;Nv!lRCu}e)D0lg6PvE?= z_UN}IY6~lH_$T#Kw<~3fh>Rb{y!~)EQXppLNSg~`oKvksyTne-5%B^Dy!ntvRTju@ zyUH9t28uc<-61JTp;xyF=MDiw@T+jGHi}*D%)5wMda)i70!z@ z85v)KTCcLyE}qwCL!id5Xn;P&H|b&GDUcY*sFiRPTHnKeY&=egQJ&Hs!$Hku@6N?8 zWQOJg)o=B4VB25>{HhJMPu!RXb`j zZ90|h{wJ#)_ZQx8hp7cl6485VYhYFMa;g7X3tExmE=v$AueGy^edM;8zzURLN&n!4 zjq+=a))3d~){-u(qwnTFEU5(PS-*nznMj*^;8!5eNG6yXUc*c5?f3TmTUHm!>HpAN zW#rd25uDGZny0J^-GeHGTYV=9`L|N@*k(lm5D|xh zrd$0ZhPHCR@-Hlr2Wyrj6RNUYy{tn=(e*t+PUJ_ndX|{ri*Yn~{(H6wLk&+-y(ePE z``j$;UzB(WsjyI1PH1%1IlC~0KxJ#5H>;Hr@I?jj_h{yBeUqQrB%&9J$49C@2g=6H zqv|HlM2}c~_Rzc;q9{CW_KNkzZnbm&AEC~^RfXxj$@m-^?>QAiK1=h?~}Xa zuRY`H=5^f}s^LDA8CVh;hxz!P_ehGj>V8W5uwCs(F^AeOK9nCm)>(J8EJ1z{Z`t3s z#PDHTDav7mW!+7B-eH)!NXJZHIN?QwaJC|$}OG5jtNfgZsFD|lt?~3nF=cR+Nrd_In&M( zaHX%G!R2#U!bIsnhg`YQmL)Y7$_zEM`KqeE6ZW|q8S4!ths=p0-X9m(^A~aa1UI6J zG`R2DB}?zaQ4C%5tL(>U;<K7lO#4?Tw#is1jv7hFXzGWM#m@c0p zOT!**YwnRe78<;MXKb@60h;sWbASkU)#2Mizjs^r2@yTZdpXO;><5Lbw)7vzov!g= zZ(+`-80Q#?da9b+7CoMh%0D&iN_q(BiRt1W&L6P#t0klG(_u54{kru3r=9PNYHE4c zM?Jz(K#m>(1r_Be7?2`Wf`Ec3AWcF|2q?WHy#`bS6qMep^qL?g2?PQnhZ;I0frO@r zlo(J+KoSBs==rVx|E{~%`EWnq_d{0ptex32li4%xywCGKIsje)A?pDhtE>T8^Cys7 zyZ|z7zR6?TZKEr+C0bwQb+BX$fmzvC+_c)oQVEelYuR52riRRgVftXw%wW`cOT1xB zb6We2)stdnxN-2gVQ@0a9A@4p=a3y{2l&qf6t?ip86~+nS5e;q4d&sgP&EVS-_erx z%U71D?-+JWr;PTUV2uO|30S3u06aJ=;QdC!LUzEHag(vRFBe^9U9VPkUO}(Q0zg$o zhSI|>iOwl(y3>U&yWLN|d*^xqm=+M)#$&rdW#PGNj{42$bnQ_~ zPPXKQOU~6+aapSRTyUZ5$1lU4i?AcVOR=E8_wY|V1;E+78lb&C8?qJ!AG!GYicQ@@r{-|>FyTBQ*?hs~keU91%v#0>x z!jL^5T=nw0&tfBKOUPW67QD!H!)(Qm`Mg`f9s~P2XZYsTQ~Xx+LD1lU#z>r$s|444 z&rZ`OQ=|;emS1vnlm%vrR_Ns-c`-4rXnvvt7^3~e^ zuIQK{pb0I)5d6c3uE#`xpf#?-cvrY|BLZ%cK46?#;gs;_D`*3_p7(x>ZbR|8si4~LC(hiL@Qd~7_k(HQ<})+ZMMQJ?!xcG^7hnpN6nAkPP> z9Y>nx3@sb^?a>#d>DKo$vCR}75TNK*h(2ufh;Mp` ztDgFPOmMq?wFRAs%X4~P0K1^;)87lZfDV8Z@ZR~r9Jgdf5jhH{@i&~Lw~LMy{bofx#WW`J+?#y2f_vlad}Q&)>o9alooUUY6jCZPuYG{O zg9(=PJnyW!3cloO=?@3zy_VHgBD{eo#D)TxY82OB8zQm^8~E7cutLnB>jZyx>9D3d zcYrJMh`_DH!x-SZh@{o3Yg( zNRcyhSO;}oL5@N1tz0$9)px|mwF3F4di%%ScwaVI-UzLnapAJrcuF(xX63`Opr{B9!h zDC>#+{$x7%q^DWWDSAJru4{yW=#;`Gj3F~1_wzX; z?F}7%egJ2;Q;F|2xJ5qU^K;{0Zl{TQ{^s<|@Kw_zQ8?OHn=!Qd@2oDF_R&?p{)D+q zjOrRwWg&sjTqd+t;ng=S980?wXMf}`GO5tNZK$j&L1j)Lo@q_{_VJzj=ttBnOgm`# zq<2Rr;Z0%@|4f1f7mXvUwvi^9ZU``A{-2lRzCNKoQu zK<8a{ekgnCSm2B%Ugwznd<#i|uL>s;7=;lV!fS-_?e^`m3nHz9VUCo`dCmek>(@lp zCSTqY#Dr{6Em}r)&`{|%EAEApRVkN5;^rc0xAs34Vk`F9^{;k%5@pwIRMm=z`QQef zlZ4-5ziQF$W3m5{MP{t)#VfLph3J#we;%@)f0x$8!4m3e>wW zoE6{{nOxE}GJ`SfJrWC-<{5#2YrzZlxdpJ;|6&wa)jGeX7 ziRPHdiWL{N4G8#}+3A2lZ}$QRIMjP%e8PlQWm4Gt<_lMPS}49SX%f&o*NNkOG1vJH z1K;8(aR1=w!y0YRyi%5~i@rZLJM2Fnzpe{5f@REqu>Cyv19swoZ~lAj#;2>3^e}H1 z>_@q`PGt2`PC_vc$UzEbWrY;bpE9h+Db~APuO%iSH$AWSFPSuZjM}A+kuE8#PQ1(w zZArKcC&9Z;Th=Yx?zSq1K&+M9xAP8%5d!)r3UoIjHUWnvpHL2HQC|@(r}ka=?z=3 zSL4GUKpMkK>cA6jgr=>5jQmhrP}1Bc8cGti)Ru?^CVe&ub(hXfQRutocJu};LeQz< z`ZXuAg(4|9l+0ynuhP0=2Xa`p?_pU`=BlN_B^)yR_#f8?Rlg>q^!aI+1bMb_6n~4g z`F5aZ5#IMztDrT6te!VmE3PKMDh?G!PwVo>lY}Jw+_;u-#=Bz2pWq^A1z%@3zV<`y zqpkBS-`3V;U&##~MY|uA=IM<2H7C)n&b60!rD0&*HXx9;$ zh5PYay+aneg5w{0g_Q|aAWdkl5HU8uAWXU}S94CH7V8G76a_s%>3tj(I-_fH9hCej zTX6|PayB9IWTG^cWmqRnEb>iu#SBwiPFhy@{Sp7r-6P1R$VteT%!0x;4*yJ8nT0suf8B?!B;0p2=feqecx|7&Jig! z_iP;V(6(WySsM_cJ(eE+Vx+JR={hm7ljBjkIvs30Z(X^$Pl#0T0l*N$( z%8%d)!h^tgdQJRgJ3*c9^2P~ezJR8XF`q3Vk+5Z1qMhQ3f|{a2@Y?XK;XLbWLB#5q z9$GIRqL^(6_3Z6ErYpYAn1pJms*m-K6s&&x1E@H0S4HuL#DwXBb-4P3!?e5>i!)|% zpqEO{a=L)?sL=GSrUiLc?^IK0b5sap?m@*y7G>@ay2Hy{OXx4^7Vn3NVpEXQ1AsKB zqRudG?~8W=y@S=k7EjtKk!ey%2ii);rgJLu&Jjeal?KkDfi3=egr}Ej#Y{GdBG!b! zUHHDd3x;`Q+~WuPQLxGv@#E><*C35izV^m(k_8(B4S^P7Y8y_7!9-V$l%n=n;zqO* z+wh_hhE48AQr2b4<&fF2a;gkSBgd!pbm21(>9us%OE^OQ)Y>Pz`!(ld(i>`%Ri);C z_<-yY9@spoa=71`$*l2@H`gxVCNX@|VgZp@f_L~2oqa}EECwr$5!Swd@K1&794Ko_ za?Jd`h!Q=XTC7t>_`bI%D9%vc|0d5OYpAiKslb#>B%z(~cpQMedLRB8Yx)qD{dGt2 zsJT1VP1T5A;m%q#DqNvW3u2)99kdm#*%phm)qtCV!jlOG78=0W6^-!SvOQvA$D~m~ z{+jOf?ABs4$BK;RMQ%kY6z;4z&xXephq0~F`p}S>%Y+`*3lgk{Z_PjijofN6pcrHp zWB0t0YBR5`6_j_qDOoOJgU)x*CX5aB;5U&C>kiiknfMf|U!0xw9%Po=OTQT{1$rWz zu((gNZ66&`Z-M)6Cll`F6U<9fksdh$;Yu@5=1CmbQA&G_nAjA;rczvx^D*mohwH2_=1$^0ZwS%HQsezCVg=zD=V8dz4SXJv{(`c{%#erVU^ zE-yK*ew|#{Tn|cTekrHaI1~s>#wLxcT5aDNDTqYKTqs^Kq7pFJ0mY6=^k+!n1w=BJ;^{Bj)tnwUwJ=CnmXoOi|^Ua58tL5S%l@VITOV3Jb z$eIiYjYwe>Yv6Th$GNSJX_IK!zZ!nw@vreW%z{oyDS6w0D|RqIjqwL)f|N{LnqCYr znv>$rBP{t6_2shfSRh75pIc@u(|n)OvcBvblRF;^Fdi{Q01# z^?5>&>_N^JPDh^a`tVST{`6wW0pH6zccbt_&m0z|JDnN7{ak3CP%dXu4UAckCo0$D6}DDm}tsy_gHr0#0NO4_PW3?y#63T+r=jlgH$o6jZxS&C_&lsMvSe>4yE&g0x-xA~f=)*CD5L#MSdpaUbs5-DZX!FSWZ&_&n z3B2Ta;#?p`4w|jRp8}_!LY()c>c7p~S1?CuRattNyT#wCZ2=G@PbAv*DtonCN_HzL z=1Ayq)3t?eI-Zp3tB~_sOKw|{-`&D$|3XUR?C%Uea4Tq6a#b9s-K*_+Hx_2rVl3N!aH+f|a=Z3Qf~ z)r{P=@md5SG%S$7o5COkM@J|4qSXEZK+hU06f@uQ<0ni5}+N8~Rs zf95p5<4t@+kGu~6Kx}mGm*FYb3^EASw)%MM6P7Mv++f_daz+QW%M;jFF5+nv<;h$C zf2LUyUE40RWBS{YKsxH2p#XlNw6Gq5 zpS`o*c#dp76Msw&0FX*kx*Be5S>5LVa9q>Z-nx6DaHz=r-0KMCL}Xl*frj~!s(Y#z zNotxPvw-loec&N43riFb4o%#cJ=42XG?*9rE(Q#xi&{xYDOgH;-i=i2`6*h*kym1d zG*NsmZSNuh(e|$Ox5}?piHX)f3sQo}dhX+V&LJx+q@7Wy643YhN+J4Qfg(NT^`feH zYZ}RttYGkCc5jZV;c5r6@dgzOf~qT8rE)np=|>JiM+;(pfcR2Tw`xJ$e!Dl1V8x-H zJjW>AsD7?YU5Qo2sn4QKhWs-g;{1qdaaqn0nvl7cZ#)>2pfVbk6DMGw`<(rEvkyCn zj_%39df%Lh05QJQd&$w>W_+`+0`EgYnlS9p6>X?+UJJ-KAf)*k<-rB&ScB3- zp&l+7denf=N%twT+C|*O&=t%olSo(v^VTSp(-g|AFCaTDj@9w6z;7=ze=%=|^bqmS zAB25qI_}OG#%SqTGskm~FEH|PVe+##cktx#A+R?6D`p*U^1)NC7A=2_==0(8$5utd zC#ZS}wJqC$5I2QKKU21!2&dNTZ8D!pawA%vdd?0w<;?)4MI90cRWzpC_;JIo zPh3xZ9mWhbnV#;~GbnZHnl!Ft=O0~CEcV**eK@^L4x3wiCPNT#FQpK40qCrr@rDGu zN=4`?TefB$dsMuvSA4cU_pKpnsFu!nG9N2(X{OV?7`6%lBj3VO6v~{|Y@F!HVzbMY8{E_73de43L zNZkwCm3Id0N>)s4tdpgs-mhBXw+2Iw)qDq=Aw7lO0GDqHsvZtfspN^mc#TRl!a1E- z3cuD^xn@FlI=08tW5ysk%225nNXo06NS;yEp1Ls+fM9Frtxfe2kmA_#VRHv+TrYOG zSADs85O@Wd(0!NnIlDL#Vfp|-J#jWnWai3wUCcx*p*PBqC>}>n8PIbD@DpVb!Wjyd zg!VfK!UH1G;}Z?hgO(Rp49o#I#uU$2Qo_=uo8%_+03$`Ok0|~#%6CnX=8ZVxN^c*# zvIA<}LRb3U2OQ|Uhz5Z=TzYfdl_w=6e;~)N4Bz8b2kOeBsMur!f-Edd#4AQlvn)b` zHRzpO=l}|#I%aET(n;S2^{h-6GtUomhLw7&vsFb4)rgHeOK@Y-MQ!nDfMx?vG)ah% zc6z$}(em8apn8Ykmv?qBsb8BDQ}D!E(HnkyUn`e|Uy(-r| zq-=Zx`_;!@XhW@O`sNr*wU@108*;CMrO{TwwM`fxnd=5Pn-djOZcte2%>kwJJD-^h z2OUgSH2b zq=h9;TCK^n@7&2<#txR+NQa>4u^~#yf168Wdui02YK5c^t!)#a zM*RwI=6Eex{CCtn`mN!PlaZa$_uP~ddQQ`K=1v~mew;&D`fb5}#Ll1$(9mhzF|yk^ z*xTurI6EPSco2Gfod14GRM+|q`joUv^WV={htl`-e(m(T9QAoaInG)~LmE%Ku>IEV zn#U3RAav>1_RV3FCaiO!?vb_aWYM}!_A zu&2%3WwuLlVE>=*!Me{rpPK%y@P9s&{quh-{6|mXpml8f2ax=?!v79k{&VMl zFZ`D-{0I5`mwx@z+Jol&pZ(tm{O?BKT{mn0(2c?J@}c>c4F^@|Y8&1`Xg+-QUk=iq Apa1{> literal 0 HcmV?d00001 diff --git a/docs_new/cookbook/autoregressive/OpenBMB/MiniCPM-V-4_6.mdx b/docs_new/cookbook/autoregressive/OpenBMB/MiniCPM-V-4_6.mdx new file mode 100644 index 000000000..394597199 --- /dev/null +++ b/docs_new/cookbook/autoregressive/OpenBMB/MiniCPM-V-4_6.mdx @@ -0,0 +1,565 @@ +--- +title: MiniCPM-V 4.6 +metatags: + description: "Deploy OpenBMB MiniCPM-V 4.6 (Qwen3.5-style hybrid GDN backbone + NaViT vision encoder) on NVIDIA GPUs with SGLang — multimodal text + image + video, slicing for high-resolution images." +tag: NEW +--- + + +The public MiniCPM-V 4.6 release weights are not yet on HuggingFace; benchmark numbers below were captured during SGLang port verification on an internal test checkpoint and will be re-run once the public weights drop. The License field is also pending verification against the public model card. + + +## 1. Model Introduction + +MiniCPM-V 4.6 is the next-generation multimodal model from [OpenBMB](https://huggingface.co/openbmb), the team behind the MiniCPM-V series. The model combines a **Qwen3.5-style hybrid LLM backbone** (Gated Delta Net + full attention) with a **NaViT-packed vision encoder** that handles arbitrary aspect ratios and high-resolution slicing natively, plus end-to-end video support. + +**Key Features:** + +- **Hybrid LLM backbone**: Qwen3.5-style mix of Gated Delta Net (linear-attention) layers and full-attention layers, providing long-context efficiency without giving up modeling power. +- **Native variable-resolution vision**: NaViT-packed vision encoder with mid-ViT merger and per-image window attention. Images of any aspect ratio are processed without forced letterboxing. +- **High-resolution slicing**: Source image plus a configurable grid of slice tiles (up to 9 tiles in the open test variant) lets the model reason over fine detail in 1280×720+ images. +- **Video**: Frame-by-frame multi-modal data items routed through the same vision encoder; any number of frames per request. +- **Reasoning Parser**: switchable thinking mode (Qwen3.5 lineage), exposed via `chat_template_kwargs.enable_thinking` per request and SGLang's `--reasoning-parser qwen3` on the server side. +- **Tool Calling**: Qwen 2.5–style `` JSON format, surfaced as OpenAI-compatible `message.tool_calls` via SGLang's `--tool-call-parser qwen`. Composes with thinking mode and with image / video inputs. + +**License:** TODO — verify on HuggingFace model card. + +## 2. SGLang Installation + +Pull the nightly Docker image (rolling tag, tracks `main`): + +```bash +# CUDA 13 (Hopper / Blackwell, default) +docker pull lmsysorg/sglang:dev +docker pull lmsysorg/sglang:dev-minicpm-v-4-6 + +# CUDA 12 (Ampere or older drivers) +docker pull lmsysorg/sglang:dev-cu12 +docker pull lmsysorg/sglang:dev-cu12-minicpm-v-4-6 +``` + +For the general SGLang installation guide (PyPI, source, Docker) see the [official SGLang installation guide](../../../docs/get-started/install). + +## 3. Model Deployment + +### 3.1 Basic Configuration + +**Interactive Command Generator**: Use the configuration selector below to generate the appropriate deployment command. The `Reasoning Parser` and `Tool Call Parser` toggles add `--reasoning-parser qwen3` and `--tool-call-parser qwen` respectively; see §4.4 for usage details. + +import { MiniCPMV46Deployment } from '/src/snippets/autoregressive/minicpm-v-4_6-deployment.jsx' + + + +### 3.2 Configuration Tips + +- **Mamba Radix Cache**: Qwen3.5's hybrid Gated Delta Networks architecture supports two mamba scheduling strategies via `--mamba-scheduler-strategy`: + - **V1 (`no_buffer`)**: Default. No overlap scheduler, lower memory usage. Required for AMD MI GPUs. + - **V2 (`extra_buffer`)**: Enables overlap scheduling and branching point caching with `--mamba-scheduler-strategy extra_buffer --page-size 64`. Requires FLA kernel backend (NVIDIA GPUs only). Trades higher mamba state memory for better throughput. Strictly superior in non-KV-cache-bound scenarios; in KV-cache-bound cases, weigh the overlap scheduling benefit against reduced max concurrency. `--page-size` must satisfy `FLA_CHUNK_SIZE % page_size == 0` or `page_size % FLA_CHUNK_SIZE == 0` (`FLA_CHUNK_SIZE` is currently 64). +- The `--mem-fraction-static` flag is recommended for optimal memory utilization, adjust it based on your hardware and workload. +- Context length defaults to 262,144 tokens. If you encounter OOM errors, consider reducing it, but maintain at least 128K to preserve thinking capabilities. +- To speed up weight loading for this large model, add `--model-loader-extra-config='{"enable_multithread_load": "true","num_threads": 64}'` to the launch command. +- **CUDA IPC Transport**: Add `SGLANG_USE_CUDA_IPC_TRANSPORT=1` as an environment variable to use CUDA IPC for transferring multimodal features, significantly improving TTFT (Time To First Token). Note: this consumes additional memory proportional to image size, so you may need to lower `--mem-fraction-static` or `--max-running-requests`. +- **Multimodal Attention Backend**: Use `--mm-attention-backend fa3` on H100/H200 for better vision performance, or `--mm-attention-backend fa4` on B200/B300. +- For processing large images or videos, you may need to lower `--mem-fraction-static` to leave room for image feature tensors. +- Multi-image and high-resolution images: the image processor produces one source patch plus per-slice tile patches; each is its own `MultimodalDataItem`. No special server-side flag needed. +- Video: decoded frame-by-frame through the same image-style slicer. No extra flag needed; pass `video_url` in the OpenAI chat completion request. +- **Chunked Prefill**: For high-concurrency vision benchmarking with many large/sliced images, pass `--chunked-prefill-size -1` to disable prefill chunking. The default chunked-prefill path can mis-split a request across an image boundary in `mm_utils.embed_mm_inputs` and crash the server; disabling chunking sidesteps this at the cost of higher TTFT under concurrency. For interactive serving leave the default on. + +## 4. Model Invocation + +Deploy the model on an H200: + +```bash Command +sglang serve --model-path openbmb/MiniCPM-V-4_6 \ + --trust-remote-code \ + --dtype bfloat16 \ + --mem-fraction-static 0.15 \ + --host 0.0.0.0 --port 30000 +``` + +### 4.1 Basic Usage (Image) + +```python Example +from openai import OpenAI + +client = OpenAI( + base_url="http://localhost:30000/v1", + api_key="EMPTY", +) + +response = client.chat.completions.create( + model="openbmb/MiniCPM-V-4_6", + messages=[ + { + "role": "user", + "content": [ + { + "type": "image_url", + "image_url": { + "url": "https://www.ilankelman.org/stopsigns/australia.jpg", + }, + }, + {"type": "text", "text": "Describe this image in one sentence."}, + ], + } + ], + max_tokens=200, + extra_body={"chat_template_kwargs": {"enable_thinking": False}}, +) + +print(response.choices[0].message.content) +``` + +**Output Example:** + +```text Output +A black SUV drives past a Chinese-style gate with a red stop sign and traditional architecture, while storefronts and street signs line the sidewalk. +``` + +### 4.2 High-Resolution / Sliced Images + +The image processor automatically picks a slice grid (up to 9 tiles) for high-resolution inputs. A 1280×720 source produces grid `[2, 3]` ++ 7 patches with `tgt_sizes=[(24, 44), 6×(28, 36)]`, byte-for-byte matching the HF reference implementation. + +```python Example +from openai import OpenAI + +client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY") + +response = client.chat.completions.create( + model="openbmb/MiniCPM-V-4_6", + messages=[ + { + "role": "user", + "content": [ + { + "type": "image_url", + "image_url": { + "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/idefics-few-shot.jpg", + }, + }, + {"type": "text", "text": "Describe this image in one sentence."}, + ], + } + ], + max_tokens=200, + extra_body={"chat_template_kwargs": {"enable_thinking": False}}, +) + +print(response.choices[0].message.content) +``` + +**Output Example:** + +```text Output +The Statue of Liberty stands tall against a cloudy sky, holding a torch aloft and a document in her left hand, symbolizing freedom and enlightenment. +``` + +### 4.3 Video Input + +```python Example +from openai import OpenAI + +client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY") + +response = client.chat.completions.create( + model="openbmb/MiniCPM-V-4_6", + messages=[ + { + "role": "user", + "content": [ + { + "type": "video_url", + "video_url": {"url": ""}, + }, + {"type": "text", "text": "Describe what happens in this video in one sentence."}, + ], + } + ], + max_tokens=200, + extra_body={"chat_template_kwargs": {"enable_thinking": False}}, +) + +print(response.choices[0].message.content) +``` + +**Output Example** (run against an 8-frame synthetic test mp4 of shifting colored squares): + +```text Output +The video shows a grid of colored squares moving in a random pattern. +``` + +### 4.4 Advanced Usage + +#### 4.4.1 Reasoning Parser + +Pass `--reasoning-parser qwen3` to the server (toggle "Reasoning Parser" on in §3.1, default) so SGLang splits each response on the `` / `` boundaries: the pre-`` block goes to `reasoning_content`, the post-`` text to `content`. Per-request, the chat template's `enable_thinking` flag toggles whether the model actually emits reasoning. + +- **Thinking mode** (default, `enable_thinking=true`): assistant prompt ends with `\n`; the model writes reasoning, closes with ``, then the answer. `reasoning_content` and `content` are both populated. +- **Instruct mode** (`enable_thinking=false`): the chat template injects an empty `` placeholder so the model emits no thinking tokens; `reasoning_content` ends up empty. + +```python Example (thinking mode) +from openai import OpenAI + +client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY") + +response = client.chat.completions.create( + model="openbmb/MiniCPM-V-4_6", + messages=[{"role": "user", "content": "Reply with the single word 'hi'. No explanation."}], + max_tokens=200, +) + +msg = response.choices[0].message +print("reasoning_content:", msg.reasoning_content) +print("content :", msg.content) +``` + +```text Output +reasoning_content: Got it, let's see. The user wants a reply with "hi" and no explanation. So I need to just say "hi" as the response. ... +content : hi +``` + +```python Example (instruct mode) +response = client.chat.completions.create( + model="openbmb/MiniCPM-V-4_6", + messages=[{"role": "user", "content": "Reply with the single word 'hi'. No explanation."}], + max_tokens=200, + extra_body={"chat_template_kwargs": {"enable_thinking": False}}, +) + +msg = response.choices[0].message +print("reasoning_content:", msg.reasoning_content) +print("content :", msg.content) +``` + +```text Output +reasoning_content: +content : hi +``` + +#### 4.4.2 Tool Calling + +Pass `--tool-call-parser qwen` to the server (toggle "Tool Call Parser" on in §3.1) so SGLang extracts `` blocks from the model output into the OpenAI-style `message.tool_calls` field (with `finish_reason="tool_calls"`). The model speaks the Qwen 2.5 tool-call format (`\n{...}\n`); the `qwen` parser is the right one. Tool calls compose with both reasoning modes and with image / video inputs. + + +Do **not** use `--tool-call-parser qwen3_coder` for MiniCPM-V 4.6 — even though the Qwen3.5 cookbooks use it. `qwen3_coder` expects an XML-style inner format (`v`), but 4.6 emits Qwen2.5-style JSON (`{"name":..., "arguments":...}`) inside the same `` wrapper. The result is `finish_reason="tool_calls"` but an empty `tool_calls` array, with the raw markup left in `content` — broken in both directions. + + + +```python Example +from openai import OpenAI + +client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY") + +tools = [ + { + "type": "function", + "function": { + "name": "get_weather", + "description": "Get the current weather for a city.", + "parameters": { + "type": "object", + "properties": { + "location": {"type": "string"}, + "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}, + }, + "required": ["location"], + }, + }, + }, +] + +response = client.chat.completions.create( + model="openbmb/MiniCPM-V-4_6", + messages=[{"role": "user", "content": "What is the weather in San Francisco? Use the tool."}], + tools=tools, + max_tokens=200, + extra_body={"chat_template_kwargs": {"enable_thinking": False}}, +) + +choice = response.choices[0] +print("finish_reason:", choice.finish_reason) +for tc in choice.message.tool_calls or []: + print(f" {tc.function.name}({tc.function.arguments})") +``` + +```text Output +finish_reason: tool_calls + get_weather({"location": "San Francisco", "unit": "celsius"}) +``` + +To get the final natural-language answer, feed the tool's result back as a `tool` role message and call the API again with the same `tools` list — the model emits `finish_reason="stop"` with the answer in `content`. + +## 5. Benchmark + + +**TODO — re-run all benchmarks once the official MiniCPM-V 4.6 release weights are public.** Numbers in this section were captured during SGLang port verification and should not be interpreted as representative of the public release. + + +**Common Test Environment (all benchmarks below):** + +- Hardware: 1× NVIDIA H200 (141 GB), single GPU (no TP / DP) +- Docker Image: `lmsysorg/sglang:dev` (transformers 5.6.0, sgl-kernel 0.4.2.post1) +- Precision: BF16 + +**Common Server Launch Command:** + +```bash Command +CUDA_VISIBLE_DEVICES=0 python -m sglang.launch_server \ + --model-path openbmb/MiniCPM-V-4_6 \ + --trust-remote-code \ + --dtype bfloat16 \ + --mem-fraction-static 0.5 \ + --mamba-scheduler-strategy extra_buffer \ + --chunked-prefill-size -1 \ + --host 0.0.0.0 --port 30000 +``` + +(`--chunked-prefill-size -1` is required for the vision throughput run; see §3.2.) + +### 5.1 Accuracy Benchmark + +#### 5.1.1 MMMU Benchmark + +- Benchmark Command +```bash Command +python3 benchmark/mmmu/bench_sglang.py --port 30000 --concurrency 48 --max-new-tokens 2048 +``` + +- Test Result + +Numbers will be filled in once the official MiniCPM-V 4.6 release weights are public. + +### 5.2 Speed Benchmark + +We use SGLang's built-in `bench_serving` tool with random text prompts (1000 input / 1000 output tokens) to characterize text-only serving performance. + +#### 5.2.1 Latency Benchmark + +```bash Command +python3 -m sglang.bench_serving \ + --backend sglang \ + --model openbmb/MiniCPM-V-4_6 \ + --dataset-name random \ + --random-input-len 1000 \ + --random-output-len 1000 \ + --num-prompts 10 \ + --max-concurrency 1 \ + --request-rate inf +``` + +```text Output +============ Serving Benchmark Result ============ +Backend: sglang +Traffic request rate: inf +Max request concurrency: 1 +Successful requests: 10 +Benchmark duration (s): 7.47 +Total input tokens: 6101 +Total input text tokens: 6101 +Total generated tokens: 4220 +Total generated tokens (retokenized): 3554 +Request throughput (req/s): 1.34 +Input token throughput (tok/s): 816.44 +Output token throughput (tok/s): 564.73 +Peak output token throughput (tok/s): 690.00 +Peak concurrent requests: 4 +Total token throughput (tok/s): 1381.17 +Concurrency: 1.00 +----------------End-to-End Latency---------------- +Mean E2E Latency (ms): 746.20 +Median E2E Latency (ms): 590.05 +P90 E2E Latency (ms): 1446.13 +P99 E2E Latency (ms): 1709.38 +---------------Time to First Token---------------- +Mean TTFT (ms): 138.12 +Median TTFT (ms): 103.70 +P99 TTFT (ms): 330.79 +-----Time per Output Token (excl. 1st token)------ +Mean TPOT (ms): 1.44 +Median TPOT (ms): 1.44 +P99 TPOT (ms): 1.45 +---------------Inter-Token Latency---------------- +Mean ITL (ms): 1.44 +Median ITL (ms): 1.45 +P95 ITL (ms): 1.49 +P99 ITL (ms): 1.57 +Max ITL (ms): 5.79 +================================================== +``` + +#### 5.2.2 Throughput Benchmark + +```bash Command +python3 -m sglang.bench_serving \ + --backend sglang \ + --model openbmb/MiniCPM-V-4_6 \ + --dataset-name random \ + --random-input-len 1000 \ + --random-output-len 1000 \ + --num-prompts 1000 \ + --max-concurrency 100 \ + --request-rate inf +``` + +```text Output +============ Serving Benchmark Result ============ +Backend: sglang +Traffic request rate: inf +Max request concurrency: 100 +Successful requests: 1000 +Benchmark duration (s): 47.07 +Total input tokens: 502493 +Total input text tokens: 502493 +Total generated tokens: 500251 +Total generated tokens (retokenized): 469844 +Request throughput (req/s): 21.24 +Input token throughput (tok/s): 10675.32 +Output token throughput (tok/s): 10627.69 +Peak output token throughput (tok/s): 25911.00 +Peak concurrent requests: 130 +Total token throughput (tok/s): 21303.01 +Concurrency: 97.24 +----------------End-to-End Latency---------------- +Mean E2E Latency (ms): 4576.94 +Median E2E Latency (ms): 4331.97 +P90 E2E Latency (ms): 8634.07 +P99 E2E Latency (ms): 9636.44 +---------------Time to First Token---------------- +Mean TTFT (ms): 206.50 +Median TTFT (ms): 184.72 +P99 TTFT (ms): 624.23 +-----Time per Output Token (excl. 1st token)------ +Mean TPOT (ms): 8.73 +Median TPOT (ms): 9.16 +P99 TPOT (ms): 13.63 +---------------Inter-Token Latency---------------- +Mean ITL (ms): 8.75 +Median ITL (ms): 0.05 +P95 ITL (ms): 29.95 +P99 ITL (ms): 108.91 +Max ITL (ms): 448.40 +================================================== +``` + +### 5.3 Vision Speed Benchmark + +We use SGLang's built-in `bench_serving` tool with random images. Each request has 128 input text tokens, one 720p image, and 1024 output tokens. + +#### 5.3.1 Latency Benchmark + +```bash Command +python3 -m sglang.bench_serving \ + --backend sglang-oai-chat \ + --host 127.0.0.1 \ + --port 30000 \ + --model openbmb/MiniCPM-V-4_6 \ + --dataset-name image \ + --image-count 1 \ + --image-resolution 720p \ + --random-input-len 128 \ + --random-output-len 1024 \ + --num-prompts 10 \ + --max-concurrency 1 \ + --request-rate inf +``` + +```text Output +============ Serving Benchmark Result ============ +Backend: sglang-oai-chat +Traffic request rate: inf +Max request concurrency: 1 +Successful requests: 10 +Benchmark duration (s): 10.26 +Total input tokens: 767 +Total input text tokens: 750 +Total input vision tokens: 17 +Total generated tokens: 4220 +Total generated tokens (retokenized): 4220 +Request throughput (req/s): 0.97 +Input token throughput (tok/s): 74.77 +Output token throughput (tok/s): 411.39 +Peak output token throughput (tok/s): 654.00 +Peak concurrent requests: 2 +Total token throughput (tok/s): 486.16 +Concurrency: 1.00 +----------------End-to-End Latency---------------- +Mean E2E Latency (ms): 1024.04 +Median E2E Latency (ms): 897.99 +P90 E2E Latency (ms): 1584.25 +P99 E2E Latency (ms): 1781.78 +---------------Time to First Token---------------- +Mean TTFT (ms): 416.94 +Median TTFT (ms): 403.18 +P99 TTFT (ms): 477.49 +-----Time per Output Token (excl. 1st token)------ +Mean TPOT (ms): 1.44 +Median TPOT (ms): 1.44 +P99 TPOT (ms): 1.45 +---------------Inter-Token Latency---------------- +Mean ITL (ms): 1.44 +Median ITL (ms): 1.44 +P95 ITL (ms): 1.48 +P99 ITL (ms): 1.56 +Max ITL (ms): 2.89 +================================================== +``` + +#### 5.3.2 Throughput Benchmark + +```bash Command +python3 -m sglang.bench_serving \ + --backend sglang-oai-chat \ + --host 127.0.0.1 \ + --port 30000 \ + --model openbmb/MiniCPM-V-4_6 \ + --dataset-name image \ + --image-count 1 \ + --image-resolution 720p \ + --random-input-len 128 \ + --random-output-len 1024 \ + --num-prompts 1000 \ + --max-concurrency 100 \ + --request-rate inf +``` + +```text Output +============ Serving Benchmark Result ============ +Backend: sglang-oai-chat +Traffic request rate: inf +Max request concurrency: 100 +Successful requests: 1000 +Benchmark duration (s): 360.01 +Total input tokens: 79925 +Total input text tokens: 78283 +Total input vision tokens: 1642 +Total generated tokens: 510855 +Total generated tokens (retokenized): 430289 +Request throughput (req/s): 2.78 +Input token throughput (tok/s): 222.01 +Output token throughput (tok/s): 1419.01 +Peak output token throughput (tok/s): 19620.00 +Peak concurrent requests: 105 +Total token throughput (tok/s): 1641.02 +Concurrency: 99.69 +----------------End-to-End Latency---------------- +Mean E2E Latency (ms): 35888.57 +Median E2E Latency (ms): 35321.48 +P90 E2E Latency (ms): 41017.37 +P99 E2E Latency (ms): 60343.22 +---------------Time to First Token---------------- +Mean TTFT (ms): 35096.32 +Median TTFT (ms): 34301.37 +P99 TTFT (ms): 59966.25 +-----Time per Output Token (excl. 1st token)------ +Mean TPOT (ms): 1.63 +Median TPOT (ms): 1.45 +P99 TPOT (ms): 10.15 +---------------Inter-Token Latency---------------- +Mean ITL (ms): 1.58 +Median ITL (ms): 0.12 +P95 ITL (ms): 0.23 +P99 ITL (ms): 0.77 +Max ITL (ms): 2086.12 +================================================== +``` diff --git a/docs_new/cookbook/autoregressive/intro.mdx b/docs_new/cookbook/autoregressive/intro.mdx index 3c172c4df..da4ca2085 100644 --- a/docs_new/cookbook/autoregressive/intro.mdx +++ b/docs_new/cookbook/autoregressive/intro.mdx @@ -91,6 +91,12 @@ metatags: href="/cookbook/autoregressive/InternVL/InternVL3.5" img="/cards/logos/internvl.png" /> + { + // STATUS: Preview / pending upstream merge. + // + // Only **H200 + BF16 + TP=1** is actually tested. Other NVIDIA platforms + // below are listed in chronological generation order for convenience + // but are **not yet verified**: + // - A100 (Ampere, sm_80): FA3 falls back to flashinfer. + // - H100 (Hopper, sm_90a): same arch family as H200, same kernels. + // - H200 (Hopper, sm_90a): TESTED; default. + // - B200 (Blackwell, sm_100a): sglang auto-picks trtllm_mha; added + // explicitly here for safety. + // B300 / GB300 (sm_103a) require the CUDA-13 image variant (`-cu130`) + // and are not exposed in this preview generator. + // + // mem-fraction-static values below are conservative estimates for the + // released model size; re-tune once parameter count is published. + // + // Required flags (any hardware): + // --trust-remote-code tokenizer / preprocessor loading + // --dtype bfloat16 released ckpt config.json has torch_dtype:None; + // without forcing bf16 the GDN causal_conv1d + // triton kernel fails on bf16/fp16 branch merge. + const options = { + hardware: { + name: 'hardware', + title: 'Hardware Platform', + items: [ + { id: 'a100', label: 'A100', default: false }, + { id: 'h100', label: 'H100', default: false }, + { id: 'h200', label: 'H200', default: true }, + { id: 'b200', label: 'B200', default: false }, + ], + }, + reasoning: { + name: 'reasoning', + title: 'Reasoning Parser', + items: [ + { id: 'enabled', label: 'enabled', default: true }, + { id: 'disabled', label: 'disabled', default: false }, + ], + }, + toolcall: { + name: 'toolcall', + title: 'Tool Call Parser', + items: [ + { id: 'enabled', label: 'enabled', default: false }, + { id: 'disabled', label: 'disabled', default: true }, + ], + }, + mambaCache: { + name: 'mambaCache', + title: 'Mamba Radix Cache', + items: [ + { id: 'v1', label: 'V1', default: false }, + { id: 'v2', label: 'V2', default: true }, + ], + }, + }; + + // Per-hardware tp / mem-fraction-static recommendations (BF16 only). + // Conservative defaults; re-tune once the released parameter count is known. + const modelConfigs = { + a100: { tp: 1, mem: 0.7 }, // 80GB, Ampere + h100: { tp: 1, mem: 0.7 }, // 80GB, Hopper + h200: { tp: 1, mem: 0.5 }, // 141GB, Hopper + b200: { tp: 1, mem: 0.4 }, // 180GB, Blackwell + }; + + const generateCommand = (values) => { + const { hardware, reasoning, toolcall, mambaCache } = values; + + const hwConfig = modelConfigs[hardware]; + if (!hwConfig) return `# Error: Unknown hardware platform`; + + const { tp, mem } = hwConfig; + const isBlackwell = hardware === 'b200'; + + let cmd = `sglang serve --model-path openbmb/MiniCPM-V-4_6`; + if (tp > 1) { + cmd += ` \\\n --tp ${tp}`; + } + cmd += ` \\\n --trust-remote-code`; + cmd += ` \\\n --dtype bfloat16`; + if (isBlackwell) { + cmd += ` \\\n --attention-backend trtllm_mha`; + } + cmd += ` \\\n --mem-fraction-static ${mem}`; + if (reasoning === 'enabled') { + cmd += ` \\\n --reasoning-parser qwen3`; + } + if (toolcall === 'enabled') { + cmd += ` \\\n --tool-call-parser qwen`; + } + if (mambaCache === 'v2') { + cmd += ` \\\n --mamba-scheduler-strategy extra_buffer`; + } + cmd += ` \\\n --host 0.0.0.0 --port 30000`; + + return cmd; + }; + + const getInitialState = () => { + const initialState = {}; + Object.entries(options).forEach(([key, option]) => { + if (option.type === 'checkbox') { + initialState[key] = (option.items || []) + .filter((item) => item.default) + .map((item) => item.id); + return; + } + if (option.type === 'text') { + initialState[key] = option.default || ''; + return; + } + let items = option.items || []; + if (option.getDynamicItems) { + const defaultValues = {}; + Object.entries(options).forEach(([innerKey, innerOption]) => { + if (innerOption.type === 'checkbox') { + defaultValues[innerKey] = (innerOption.items || []) + .filter((item) => item.default) + .map((item) => item.id); + } else if (innerOption.type === 'text') { + defaultValues[innerKey] = innerOption.default || ''; + } else if (innerOption.items && innerOption.items.length > 0) { + const defaultItem = innerOption.items.find((item) => item.default); + defaultValues[innerKey] = defaultItem ? defaultItem.id : innerOption.items[0].id; + } + }); + items = option.getDynamicItems(defaultValues); + } + const defaultItem = items && items.find((item) => item.default); + initialState[key] = defaultItem ? defaultItem.id : items && items[0] ? items[0].id : ''; + }); + return initialState; + }; + + const [values, setValues] = useState(getInitialState); + const [isDark, setIsDark] = useState(false); + + useEffect(() => { + const checkDarkMode = () => { + const html = document.documentElement; + const isDarkMode = + html.classList.contains('dark') || + html.getAttribute('data-theme') === 'dark' || + html.style.colorScheme === 'dark'; + setIsDark(isDarkMode); + }; + checkDarkMode(); + const observer = new MutationObserver(checkDarkMode); + observer.observe(document.documentElement, { + attributes: true, + attributeFilter: ['class', 'data-theme', 'style'], + }); + return () => observer.disconnect(); + }, []); + + const handleRadioChange = (optionName, value) => { + setValues((prev) => ({ ...prev, [optionName]: value })); + }; + + const handleCheckboxChange = (optionName, itemId, isChecked) => { + setValues((prev) => { + const currentValues = prev[optionName] || []; + if (isChecked) { + return { ...prev, [optionName]: [...currentValues, itemId] }; + } + return { + ...prev, + [optionName]: currentValues.filter((id) => id !== itemId), + }; + }); + }; + + const handleTextChange = (optionName, value) => { + setValues((prev) => ({ ...prev, [optionName]: value })); + }; + + const command = generateCommand(values); + + const containerStyle = { + maxWidth: '900px', + margin: '0 auto', + display: 'flex', + flexDirection: 'column', + gap: '4px', + }; + const cardStyle = { + padding: '8px 12px', + border: `1px solid ${isDark ? '#374151' : '#e5e7eb'}`, + borderLeft: `3px solid ${isDark ? '#E85D4D' : '#D45D44'}`, + borderRadius: '4px', + display: 'flex', + alignItems: 'center', + gap: '12px', + background: isDark ? '#1f2937' : '#fff', + }; + const titleStyle = { + fontSize: '13px', + fontWeight: '600', + minWidth: '140px', + flexShrink: 0, + color: isDark ? '#e5e7eb' : 'inherit', + }; + const itemsStyle = { + display: 'flex', + rowGap: '2px', + columnGap: '6px', + flexWrap: 'wrap', + alignItems: 'center', + flex: 1, + }; + const labelBaseStyle = { + padding: '4px 10px', + border: `1px solid ${isDark ? '#9ca3af' : '#d1d5db'}`, + borderRadius: '3px', + cursor: 'pointer', + display: 'inline-flex', + flexDirection: 'column', + alignItems: 'center', + justifyContent: 'center', + fontWeight: '500', + fontSize: '13px', + transition: 'all 0.2s', + userSelect: 'none', + minWidth: '45px', + textAlign: 'center', + flex: 1, + background: isDark ? '#374151' : '#fff', + color: isDark ? '#e5e7eb' : 'inherit', + }; + const checkedStyle = { + background: '#D45D44', + color: 'white', + borderColor: '#D45D44', + }; + const disabledStyle = { + cursor: 'not-allowed', + opacity: 0.5, + }; + const subtitleStyle = { + display: 'block', + fontSize: '9px', + marginTop: '1px', + lineHeight: '1.1', + opacity: 0.7, + }; + const textInputStyle = { + flex: 1, + padding: '8px 10px', + borderRadius: '4px', + border: `1px solid ${isDark ? '#4b5563' : '#d1d5db'}`, + background: isDark ? '#111827' : '#fff', + color: isDark ? '#e5e7eb' : '#111827', + fontSize: '13px', + }; + const commandDisplayStyle = { + flex: 1, + padding: '12px 16px', + background: isDark ? '#111827' : '#f5f5f5', + borderRadius: '6px', + fontFamily: "'Menlo', 'Monaco', 'Courier New', monospace", + fontSize: '12px', + lineHeight: '1.5', + color: isDark ? '#e5e7eb' : '#374151', + whiteSpace: 'pre-wrap', + overflowX: 'auto', + margin: 0, + border: `1px solid ${isDark ? '#374151' : '#e5e7eb'}`, + }; + + return ( +
+ {Object.entries(options).map(([key, option]) => { + if (option.condition && !option.condition(values)) { + return null; + } + const items = option.getDynamicItems ? option.getDynamicItems(values) : option.items || []; + return ( +
+
{option.title}
+
+ {option.type === 'text' ? ( + handleTextChange(option.name, event.target.value)} + style={textInputStyle} + /> + ) : option.type === 'checkbox' ? ( + (option.items || []).map((item) => { + const isChecked = (values[option.name] || []).includes(item.id); + const isDisabled = + item.required || + (typeof item.disabledWhen === 'function' && item.disabledWhen(values)); + return ( + + ); + }) + ) : ( + items.map((item) => { + const isChecked = values[option.name] === item.id; + const isDisabled = Boolean(item.disabled); + return ( + + ); + }) + )} +
+
+ ); + })} +
+
Run this Command:
+ 
{command}
+
+
+ ); +};