From 67151c73f751739638f91472c18b2af5f3789894 Mon Sep 17 00:00:00 2001
From: bishe <123456789@163.com>
Date: Sun, 23 Feb 2025 23:15:25 +0800
Subject: [PATCH] without cnt

---
 checkpoints/ROMA_UNSB_001/loss_log.txt        |  22 ++++++
 checkpoints/ROMA_UNSB_001/train_opt.txt       |   8 +-
 models/__pycache__/networks.cpython-39.pyc    | Bin 51012 -> 51061 bytes
 .../roma_unsb_model.cpython-39.pyc            | Bin 18663 -> 19104 bytes
 models/networks.py                            |   2 +
 models/roma_unsb_model.py                     |  73 ++++++++++--------
 .../__pycache__/base_options.cpython-39.pyc   | Bin 7598 -> 7610 bytes
 options/base_options.py                       |   2 +-
 scripts/train.sh                              |   2 +-
 9 files changed, 73 insertions(+), 36 deletions(-)

diff --git a/checkpoints/ROMA_UNSB_001/loss_log.txt b/checkpoints/ROMA_UNSB_001/loss_log.txt
index d27cb3a..fd8dd2f 100644
--- a/checkpoints/ROMA_UNSB_001/loss_log.txt
+++ b/checkpoints/ROMA_UNSB_001/loss_log.txt
@@ -46,3 +46,25 @@
 ================ Training Loss (Sun Feb 23 22:33:48 2025) ================
 ================ Training Loss (Sun Feb 23 22:39:16 2025) ================
 ================ Training Loss (Sun Feb 23 22:39:48 2025) ================
+================ Training Loss (Sun Feb 23 22:41:34 2025) ================
+================ Training Loss (Sun Feb 23 22:42:01 2025) ================
+================ Training Loss (Sun Feb 23 22:44:17 2025) ================
+================ Training Loss (Sun Feb 23 22:45:53 2025) ================
+================ Training Loss (Sun Feb 23 22:46:48 2025) ================
+================ Training Loss (Sun Feb 23 22:47:42 2025) ================
+================ Training Loss (Sun Feb 23 22:49:44 2025) ================
+================ Training Loss (Sun Feb 23 22:50:29 2025) ================
+================ Training Loss (Sun Feb 23 22:51:47 2025) ================
+================ Training Loss (Sun Feb 23 22:55:56 2025) ================
+================ Training Loss (Sun Feb 23 22:56:19 2025) ================
+================ Training Loss (Sun Feb 23 22:57:58 2025) ================
+================ Training Loss (Sun Feb 23 22:59:09 2025) ================
+================ Training Loss (Sun Feb 23 23:02:36 2025) ================
+================ Training Loss (Sun Feb 23 23:03:56 2025) ================
+================ Training Loss (Sun Feb 23 23:09:21 2025) ================
+================ Training Loss (Sun Feb 23 23:10:05 2025) ================
+================ Training Loss (Sun Feb 23 23:11:43 2025) ================
+================ Training Loss (Sun Feb 23 23:12:41 2025) ================
+================ Training Loss (Sun Feb 23 23:13:05 2025) ================
+================ Training Loss (Sun Feb 23 23:13:59 2025) ================
+================ Training Loss (Sun Feb 23 23:14:59 2025) ================
diff --git a/checkpoints/ROMA_UNSB_001/train_opt.txt b/checkpoints/ROMA_UNSB_001/train_opt.txt
index d7766e4..4d2cd07 100644
--- a/checkpoints/ROMA_UNSB_001/train_opt.txt
+++ b/checkpoints/ROMA_UNSB_001/train_opt.txt
@@ -1,5 +1,5 @@
 ----------------- Options ---------------
-             atten_layers: 1,3,5                         
+             atten_layers: 5                             
                batch_size: 1                             
                     beta1: 0.5                           
                     beta2: 0.999                         
@@ -28,10 +28,12 @@
                 init_type: xavier                        
                  input_nc: 3                             
                   isTrain: True                          	[default: None]
+             lambda_D_ViT: 1.0                           
                lambda_GAN: 8.0                           	[default: 1.0]
                lambda_NCE: 8.0                           	[default: 1.0]
                 lambda_SB: 0.1                           
                lambda_ctn: 1.0                           
+            lambda_global: 1.0                           
                lambda_inc: 1.0                           
                    lmda_1: 0.1                           
                 load_size: 286                           
@@ -50,7 +52,7 @@
 nce_includes_all_negatives_from_minibatch: False                         
                nce_layers: 0,4,8,12,16                   
                       ndf: 64                            
-                     netD: basic                         
+                     netD: basic_cond                    
                      netF: mlp_sample                    
                   netF_nc: 256                           
                      netG: resnet_9blocks_cond           
@@ -78,7 +80,7 @@ nce_includes_all_negatives_from_minibatch: False
            serial_batches: False                         
 stylegan2_G_num_downsampling: 1                             
                    suffix:                               
-                      tau: 0.1                           	[default: 0.01]
+                      tau: 0.01                          
          update_html_freq: 1000                          
                   use_idt: False                         
                   verbose: False                         
diff --git a/models/__pycache__/networks.cpython-39.pyc b/models/__pycache__/networks.cpython-39.pyc
index 40fbb5c3111eb70879d74788f829d845264fd328..91353a9c2e5846b90c8a57791cb979eb0cf5682c 100644
GIT binary patch
delta 5643
zcmb7I33QZ25>C&ZV?tmO5@H6z6*EhKB*-O~9C9UwE1&_wWRjVLn8_scOd{ZffId)B
zu9S<gBFcIKaeXj=tl$F?@L1i4t}g2+pm>0a;)PG~VEd~-$AGN6%X=xRyQ`|}tFEr<
ze-5wKUs$cTx~$fiDE06Aq5CS{++qD=w%ALLW@m|obS(Rt#<@YA+ft>I^8&hHR8RzC
z-O;P`4KcoP?wFviK1wbQ=y_frGzX*Iv8#+~EmP^lDV_BJ1J>MzU~JIrHu73r!05IH
zb)BQ`6nBV#DHszp22DW=TNCd1RmK@njt+IBOi@u|rN<Qr734IT;=(bdE`Md2v#iGF
zruPRA7G_Era-Fcy#32hZ(&2npgNV^tM556C5?cOG)WwMyVY(zgAXlfv(D@;U^x`z_
z9Da-VjABQO5GA2WBhCoZhsuKdhz5-+6$?WfM~%}Ze2ke-K+b@ir7uR?V?P7?4ahl4
z8k3QD9_&|&gmtBU`L(vN<7i^E(^=;7dY#UD$UCOQ#2Odr>q2|-X(dx6qN4m3#aM*F
z6rW=0C{Kq>V{bHwpDBOBO7SfnpD;rFKuHq^=U#xxW{|@m-z(yXlkcnlW=E^&c%5!n
zz%@I%+*RdaR$ihd6Ymj^lVws4W#=W)`61C%vd~IrW4qJxNz=L{s=Z}jr5`rd*4*Zi
z4xJpVMly+BSez_gBlF}<B8eWGJXjo~_a=|ALH<?vYfw@aUA#diQcm94B+Q3R-dtae
ztlrvU!_GF2)A;-%Q&;R`r;<>bI6;T<uNVKMU-CapO5<q|c7&z6vD#^}7TlF$hgoy2
zM^@JbJYjRW*A)nOe6lMw7xWWn>D_{n;sZ*Zaz-SQZt7fdKIEKQAw(LzG_6ovq_pW<
zMFt(6K1P^B(KD=KU~lX+05)S>Znv|##$D(2$U&IR02z)zSSx-0fXi3raRQ9&MR|oi
z6Z&I08)P6z4&`h}7FJqcIMoDuauD?|?qNKx^uHwA%wJ3z9O^u4wy@KJS?85cuA8&P
z_0)88Zvkza5qg<)?B-mNN%~vrbB3w0!zABC_^*deHauK$VRMBippZ+M1(|5=h(^-3
zTL#3Bf@BVa^wQ&}Wrc}iKvPkUIa5s)IL!1$VUidC=cu;GpuDInS|Bn*XA~8qsc6m|
zF@bi^$r1T<X3k@#aWIxo>*gjW<$=8kqyS_x$VhsAZf`M#KAhV}`P`wnn^NmvoFQ(Y
zS;by8^I~yDr!h(}tgm-@97dT1C{t<pya{3&A8HA0nm4`E3?<^wYaph;ZXQRooKl><
zm6egs-pNzdv{To1Euuu!rFfGvm6Mkkk;2_fa=H>6Pk%2<4SneJiYyE{o#ixeoIbH*
zYkc*&?lIn)vMSjJ)A{f#oo;ufj_->xXW@9owT5<*;=|FEzS_Eg(_L9Dzf%gcqy2$J
zUUjY1m@glYYTM*8I4^zcdMUMSU-<)e;R|~+ByXX&O2>(%)TL~a4Q^He9^WDnpD3rd
zL2D`-AjZ<svd=^z?Qkc}8w+LJP(~wXVF(oA=;<CmgYs*Du8Kk)f+i{qgNX2a2o^NY
z@1x<KPlwM|Lh8`AVjfwi?T3t$^VFQfqMg`OthXly=if+jm1l~Lw6c7T4gTI6fz2w%
zwqg6?*8XB5*{d=`)fF9u0@ye|YxG3yzyOOC7mibD?DGVirF>p{up1xRcyNq<1QcRG
z_#o5h-25~Lj!5DRuY~v-7II<P$R;T#KT$N(_{!ar%OI})aP2&hh%IvkoaL&^)Jc!F
zEAM{;<glTV<LZbQgvBB`2TH;xD>#AoQSq%MHo(~)fxm;Ccp3g175+<&B`~S}qGTM1
z3&R9($Ou`AVJ~on;5vgtOO|dXYR(a(-L4z4-ku$Z;FaOhf&`sd&HHi@)h9|zjQ<`r
zS;WcxGM)4`B;!8G`7qUTeK{E_8GV>mR!_IVpO+%gcak@@p-102W$LDr#dN)|scI&S
zaTel;RC;9r93yB4#F=hs&Ghy}Z^_1RjIXX*OBH`af?saBCCfjdDK)cHHAWBmf?lk-
z$%ef?j&RnjJnLDzrhEI%HN(eL<yTm*Qo~`8Er9<~+p7!W@Ex@Iu_V$&`Xz%dLCE>P
z1zbgQ7Hpl$XNePZm^G|bu+CUHuiFlaZ=Mq3T;n4;jL=+xN^7q_lUJ6BYC0xUMJgYJ
zf%N|AHrW0t0-HrvUFnyRuT#^)#L%n$e+f}akJe2SZn{v{Pt2lT_1B6R%CGNg>IW}p
zQf+;Y6jaNwK|Q+aRsOMeTv^7`w)%82lm1qpA>!zp`aKCtpfd<gYo2q^htS)%6(s>`
ztX_ppz39{nLKXHbIjhyRP|3pk#EjdZ?Sl3J5WWqZ`D#8&)?)P_xCP++APpD>z}11c
zLFyH8bkrifj7BXUI%yH6xo~KhFkB3BJ4g_uL6O$TYuGT`9{z<GF6ApRhTd8HMA{vf
zt#6II7us=nD0KT2+cG6K8mJA*Q;Uf2DVMckR`0Paom@mE!IfeWeIA@=x(hb%rYQ|+
z!XNTBSj0F^g7&ohUKt7N{Q<YU8=|E^zZm2`HB>dctw_m{n1#4p#mUh?$CmUHE9laa
zt~L%wgikTb3Dsaj;~MjXdLyh|`j)}o2>Uakzh03-ty%%YR)cJ$Wy||o5Rt7$HQ2Mf
zw`ipImZysj^vm*|vD{iY(0!=Sikw)4R%^9(s#xI=MC(@AuHk^%)lxX#=XBTH=JUI%
zYrU1e3UzA_tgwqUp%W|a(iL$V;?@LNtv{hFX#hJw-{7tE2U<FK*vQRWy+~b)Jkqbd
zVYF5S^?4vpkYDLor)2R4y?sxS_=)<j94PJ~&&meVYAj~b+NF+A$9p}xK5Nw2k)kEj
zE7+H>Z5=plF3Z7HfIL83R@w3+z+<^5RyKeGhOm|0a(aAaHExfT8?n$2BGW)30)>S?
zakl*!N?AQkH}EQP2%DtK=dP&^o9ld)<#@7gff}EDcW^yu^V+Nto54N|QlQA}=xUe0
z$|+p|kGx7;!Un8_#TeNi8rfuCH$Z5y8bLGGmFON3^g({I5ux~&-dGn_`O@6<mA<hF
z;@ql`B=UYWER^do*%{;YiYss=YH@rD(;UYuLx=^ghLup__Rqm<4;1GWM~fpp2XZQI
zmbMJr@{pq&!Oku$wPl%n5ObSAw6{+U-TA;4Yqgqa{cPFH7v*7^w4p}}%p`bPr2uI)
zOBnQN+43pAze|dv>DP7R>E{hj6@s}NkL!o8Rl*S$MMICc6tq^O?P_p!;K-eB|1o#?
zg=eSVg_aB3n>b~jBC)wKb}J@;Pi|viZ_%Th+knB7$1VewCATYJc1n17u-Fiq`0)Kg
zY@id5^h@Fr#1Zu%q9G0|@AF=0$d(87VlBP(Sg)aN&WFw8WKC^NU4ZX=>+A8VR^Z&z
zl=FCJW3yt<Q2ygPOiyESIeXlJu6bfW>`&N`v!xlIGs#GZci?)<m$NP|){-eVv{cnp
zI`~AsIuYxWoAhD_ZQGWbyaN{5_c&!r_43zYaR<G-ZKR1)%ue054-{Hjq)_$tqTj?W
zn?Bv%8av~wN<E1j&0eu9w7&TrLnEg-I#Ho&1C7`%8#{3I{EHeoI%#>o3rmb71n&2f
zpX13*ic_9%02hd7psQsBY}+J#xua_;gJiQiu^xF;89T?R*AG7HhtZuoUr%5uuBLe4
zNFVp;05ycNThhJ@935y>A;yJPKTa5RBJNWBP`Rrm$oMItFFAJY>VT(7*sRL7$0c*+
zGjwiOj@Zw81!>UkRSAEEGyHzp4{jgtbr*fLyMMx7Oth1=B=YRGXVv8G*cnwswaZxf
zY}=l}J;T&*2z&QrsN+4gXPCZmH!N`6<vt+6ITSb&v|yjW%>Phi+T+Z%niC=j-S!Y$
zwZ6(}a|%mg5D!;7t;*1&&-E9#GVUh&{JCwWYPhD|yTcUPm)TlS;ME%7<42G%=u&d!
z3H&0(U$+_i2aA(5Z~s7r@c#X0MI=JxZVN)IhVUe2{x5`2LIb#3p9<}eIAOS24uw!U
zNOwB6*rI+5G8`zk9)y;VY(c~I5v@6JZ7VML@=7GG#1wh54cDw@8?D)+?##hHrsolv
zOggovGu0magXl>Q9!wLbXz#%cagI(Myf})#ocYPaO)MJnS#akDMOr`9xv4#YC9O|^
zdlKXVr5v)0^EBwtfL7mLQ1hQg`j(=XFQ}}VKqn7XiKBdK7pe5c*ZXjh>I^f<il~pi
zhrn|~JqnJ!YI#)6KAdi9VV556rh4Qfhx2Vml+=&dN=x53gFHf)4tG~S=yg9*CH_pS
zj@+5>0kE6~vB72{r6097;&%`JO+)f$|3nP=6R#NDe2^*-Jfh_?kVPOb!wf%5xTm73
nYp)H|D9QB(H%jtl;Cw+aB;3@FI3057QMCVPp16)IFAe`UTp^N%

delta 5603
zcmb7I3s}@u7U#~3;V~mQsEC8$+fhV7P043Sz8|3Y07V#?fe;w*GJuak*e&JD@_Dwb
zO)YEFMXl6P+D&&QAKCV0wSBg=W0t06*7ml~vUc73JAaU=b=%hO^KkCH=bn3h=bp#?
z&o}G!C)VpNPKzZxMEx82-SfqR_gfxJ7F+4b<V3NMUQh0pGtQ@TnM-wYkyqyn@d;mq
zD{P&<F5J`K74Fm3gvb?MJ<scXp}sIz#5$u|i}V^?kv?7LkbA_hh1ci{_ZfXApP6?E
zmt~!Cc8J63FG$&HBJ+sh!bIan45j5G%6r;jM|Xn=*O^7M(Edl^uP#!A3)2mGo}3+{
z{RwHu^x^_t95r8Dp;4npi$(qwqc027`)aqmNX26piyHsSV>5J7moW1I$cG@8se5`{
z#1*h#f_y>~()&h#2KIAB0=mUj@(XRDA>A$-X<K@Mi6y_I4LjqyTu^KCL`X=LSuti|
zFj-^Fx5*3s;p6Wzi0`Rx;#%<)nJ0}F*J#3|;i*@lYA?tMkZ%-mM9TNnpWHC%c9!M4
zoL*;cSdp{T&8++vJv-?>@g${9P7xjH&ZPtBqlg~VJUOeYO`R$Ssht61Wks!9I&^Y4
zbVgIk`Yz&S8Zl+7h^9BE3>UAGWoo(&@-M+sgOW1qtOjW#Q5sE|(}&XTsgHNUdcahc
z>Zy?BO|81@tm{pxot9&Y!+Cbv<hP5n^!2nm#2HGOez8+Go(5rqSgJd3J5OU~G{)GW
zHnh?$%d5TaKxk2!)9ZD6WE`EFkt9B($eClrdo*?CWnrTsnG40I{*9T%LUf~Vv$Dmv
zG<{Z+=tI|Lr3<6~j@cG5tT#?dh0buN%avbV;i@ik%VC)912PH@v6OhKyiQM{J0D<V
zFY;ygib}%rV3468$&`}cMMTi4>`W8%$zfEG*VFix+W!yAnDbAQ28Slj%~jSsG52#t
zH_uBJX>@#EAA#LABlh(rF+WxGqoMO_QtnV=he^JHYe<7mRy<O1fzV>NS0R^jAY{U{
zs~Syb=MS-tfn+j-^wM2LU)Dy86iUfSiR-84W;jA+I`oYP$pFEf%T%`ZZ;FkjojFTI
zKYvW_QX$6D+64>51p076ikLz%3xAh79@?jZG$~<+nNJQFOUj+aZaD?>(?F(!Oa>WE
zpDpYorc-p@Km}`DUc8v9&gv`1P-9-1nz@=+-0@B&7|_=^-43HnggG;*e9=Ua$@jj9
zPA{6(ah4Ks=rx=(pmz#8+(nb~`zS3H`F*;~RMUC7kha_j5g|9MCZ$S?@_3WXQUVj`
zMq#W!x}Z!XVyL0d1jmsTF}}i6lj=$@t0*j$128=e_Qca(XY7PTj5$E>Qe2zYCn-J<
zR^q9w_U5}v%H`K;LvC1=cX^r0IW^|Y1VYKjsJl@4w1j#uJ{5cOTzL(ra5m3{<UBGK
zW{6cZrEszhW|jaJCz`NE%Pf^>adf<Jh)Abvg&)_?#+FEEuqxsR*L;|cS-xB#lXF0T
zC=fQwUBv*aG;p^>B#%Mgd~5=sAQU4oOu<4LxP%ZIwLVg;rE>R`Q8{W_-629lB2L+<
z9HVlPnrjUgnThrGIN^vLO`D7Qi!F4xXn_r;-xmbgBGVP*X<{PHDDCgxUVNJnvuWSm
ziFZ!K8H{nE;sTK>;5}|{{$jqqW}L!zH~}0xU=%ilf$+U{qqq`#JOjj){uJda*q1`{
zOS+5QR9kXr%3=tq|42)h`WuKSbb9lP+)i(`bZckvsbe9BW0V|MQ6LV8<uVtWghv)}
z^!}07FDbAAL{kvj4ss%+%}{7>)aOH$`rmrP1sFzQh;Wcj410s)Am|JZy;izF3e{>`
zIAlPwJsjY#pE^f#3HC&2R=Y*Wzz{7G{)dmCZ9Yf#i`1dKt_z1;F;unYEMNKEjQVXl
zTt3SNdrk(S?kF=<rcP0hf(Mqy&^?|NrP)x*(T00av6Bc)Y<N2kj%agRM7IZYTV)FO
zRF`Y<P!$x&KUS@-7&&=~8Y<=rttEXzS1YF5aMr~j8$;zn%C6kdqdi+qurXHI70|0l
za2T}EzMBk7dv^u=udtgRKf*2&ZWwe?LKZX5B?@O#Cf^$FO4VhCw*(3q9>?p=E8$F1
zLL5!)8Tk_+T7$h#+tk&Uk*JClUD@S28d^2W2HoEU!7<AvFvWnAC}ml+|NE-9g>cfD
z>MT)6eQT1$9GX+pPlQojO?T5ESe8vYYI???5C#l|s#L1zVvD#?+(l<=5=1urP}5h0
zQ?J_Qs5<Pq9oKdOB%9m)?POY((+P)$>6Ld(YQ?NmwK`GG;aXkki}(L&S%H{cgWXPS
zJPN|G&T*~gLu3V3TfkL<LypKg46DF-K@k74S`o)>ns*zhaz)zYWtiq7pkcyrImil-
zl^{Mv+Ptk{!>k4ZTrEgFhs`(&U%Au%s~@d51xLMh1%-5J<xE?H5*rKD24$&PSbNEa
zR?G=v8Ex{dRn%U$$g~FP8mXbqE|&VY)tN;G-+6oB;i(tUS9x8s5u%F$d<Dq;YN$#u
z-wG$)7|cRkKEScCl0<!ySWU_G-EHthWso`Ha+%+$N{;IPFEpGn#I1tvIoSJ<BH3D#
zf$2>Kxrz>~9%P0inv802d37Jvuq<m5gq4!k^ornS$^IQc3)iGXz@u98w9~_D9D_Gw
zd=RP+fmDOASM6#cVD;p?Dr!Ad&hpB#5>K(p&@a}+iFJN!W1}t}Zyv43KvrusXfKkX
zTIt|@39<YD(HaY6@<0}W{F7eq*hRcXrnNcZIu)!PrXJxBt*tYy!(u;rp~2x#zu&E^
zU$4fF7%go60_B{PY^D{WMVKoFc?d&J)DjFAfCQn(YA?)f0f*KYu#`EMyJf!HQ&{10
zOZhMsk|2@=!k&uYuFUuz;V`@ApS(UxH|!SJ119P8xGKs6q1B#}BD_!^heCEn0=ORZ
z+Q!7u+rVxIfp2od%AHlE`O@ii%Uif9V8BX1q)Y9cOCvT75t{2p(!HAs#AEcHKgI~h
zd`Y^^0TnRkHh-b7-vs^KfDk2;)G%9a#3Z+-2NgHN5v_T37pB>(Kkra(;q1J2M{!`W
zx7gF{bLBTnTZk#u4l8Y2(>uOB&TdAR{4Gv?6hy21<#ce%4ojJuX!}yx#;LN6Ry^EO
zgW3t!mMBnKy%Gk!Jz4JJFP>-f!f5d32{h=DeC399kG!R?-=KtpvPJU%d^4l1675!l
zUjqK;qv>XlIILR6pN7|8rKUA8{~b8L-5%wj+DWO~>mz;-TYz3RG3+<!%=R|eu=9x@
z!A_LB8FuFMFq(HHin1Q-C${*PKlU4;UT@YNNu9V-vD@9S#t%}<#|8W)J08`G4P<<(
zclON`34~_Iipq*=FPDe5H{va|z{b6b1Vp7wfL<m&!0l1fA5h&>`%Ms#_c9U-P1`vn
z;yTtj(DvXnrwd}~EnsT>$ks(hnET5u3RN7P-<hW_am=o*da;kr?vCxU584<=q*9|g
zJiB-wMeZ46;>fg9=AL2r#8OX)&a{0`&VTqbiDLJ*`qTKbQevary=B7hKehMohI&p8
z^sQ{wG*;oTWSGlE_h~hBbkwrNk0l0i0Nl@~4EvT_8Yc?p4VRT?u(v&GY?9&+bdP0>
ztad-vgRgDgfeiH|0=6_dc;IvtOL1Akt4Ee|N0;=BKUvF@uCQ~6Q8^b?K>dPY)QQL&
z)->sF>+O_t@K6W5TLPh~w7Z=$RX#&;hf>60K4t|KA6ggn2iU=no5SFq=VMk=kLJM<
z&tXC%w4k}Ctyzti<3Lm#)lws5GvDOX%_GG-bfvkky3eTRM(XPiVk3K69)|7k{0v8w
z=H7QO!|wg1p5>a(5#Yn#cBfi`Qt@3Z1wh<f6}jQn(wXN6ixP&rk?e<eo62CCmVC$P
z(&7HCH3vqm12ldk34<;sRlb8Cs3;)~$v0e_qYX!fDS#)Bd?bPaB34@gS~P&?F!O%_
z+=U&0)%Kj=$Y(?xAzV*ezX6qG4;r`LtbQmOcC^TH6uX3E4H~TXspV+DR#-6QR4}aQ
z7<uj{tj4`IYH997xh(@tN8p!!6!lzZ+R-vV^rTl>?BZRz($ZIaOi?dfAIqQ2{3_y3
zXNG(t>>j5`+ZQ@_zNfI11j0Ys>;!j}CLfCvpHlI$A+4shsOihWrp4%GOKZ6sLLH8m
zij#b8-x3`^J&<cuCur)di2Bys3Z5J9NpNh~H!$Y`y8lFisg>P-B3}GJe?2kHc0x&A
z!*N>dMjGS^N<P^`{n9t<WU1Iee?ECn)OqZ`0MZdUZItz5Ts?m4;NLSOf96lbkiYO2
zf-44D0x}1r0c07-OHjkF8E(EP?%J~iWl3_0K{`p!296jwM8eFVdaFM9KYekkNTsw>
Gqy7!L;Zkh?

diff --git a/models/__pycache__/roma_unsb_model.cpython-39.pyc b/models/__pycache__/roma_unsb_model.cpython-39.pyc
index bd981a861591db2901c5931be472be7dc126039d..ac7f924b59dc6de3192c8da5e0c2f482b13a1437 100644
GIT binary patch
delta 4095
zcmZ`6YfPKh^}FBa$L9wI3>fnWI6weHco;}{BoGL%q#<dyqOG$wxrXaF{xCMV{#Y7h
z%c{~xR0%2DZKO!0sohMg`lnJvr>yIuY5mn!+oY;#SNX@*O<S{8UAJ~pwTGtdoNMEJ
zky_9_-?`_Ud(OS*p2zjwACTNt;tE$(*d_Rj9(g@F!EU;~Ur&C@pRE68mQa~iP}8=F
zNNGuvs7psmrGZkV%bG%~shif&T3WYlDM@;C6RoFSgJgq2ve6*f1d`2BiSB!m&=%Ud
zZ7yl>>E<lmPusSY5}_T8chCa{m7PFQ4H8|t8CG-~B>V=WJqDw_;vfeNl6^u~zkxEK
zXH9gF1`MPj5E`Z<2Ff8Yd6<qGC}ThwrxRt0g&v_tw+S91XU$9Bl{0ci0dHEsXEL4K
zmXY(Y<e%Ey4_EN<liPQ1zv7!+KJ%FGBR<&hDRuSr#l!0>G#r`>eJi%S<-HJzMbGKJ
zvq|QovD7Myt;Z5!J;}Bz%aYM}awQz!>Nbcj3X<<(oo%k_n=Bmn%_bA6SW1s1Rxj>0
zH9o4y1+x42zDvYDg|T3A6F1@Z)--u>_vzLlLay^qeDhg%t}8F)x?%tRJV{HT9=%sP
z2siHr=gJ0ppWd$x<Oc7P4e0LNP?~7Nd18<piNg6qhd`rqPWoO;#st`*oI^jHCYQ=`
zdGOyAa{f;CE7XnxZA=@7WuDwbUIGJ0U>t=pc~+*LZE$vIsw~s<wKB)bGVNb0(>o1j
zreGX{F&#zKGx~8bH49_zk__{Cec=+GXps-KB}kCx+s-@w1rGfOdUdOZw{&^<aQhRo
zMEGm%&DNk<Tr2;ey^TD}|I_{eIl&h?4v?pKykmqs&hs60B*SlY{Bfn~^wL9f#tUb4
zppkh16x2jS7kUaNJ+{7HaIUgQSdWC(lQa@9*ns;Gi$L_SDwLKJiGpP*u#`-t3JwS$
zJ;Gwi#8QC&^1ukUbaszlLn-r0L=OjwpP_;yxdEDETM-soG*0G>lX<Rn4#C!+?;Nsx
z4mwf`{JqXz`-og6yA`ieOKzxLGU?s5`NznkF9{<$3&#@dRlwP6NT^Ol^bm`j-;A+H
zDimd5%HBe{4J;?cyx>|*u5WC@>E=RlI6}~ZFf_*=K{*pd^agtyP=2$=Yi4gE_I}SS
znb~dZJuJiKzR~vx`GCLEQp0ce*~!Z8n|(i$$p&8^yxII$5N4+kpr6=Z0QjwJ1o4mg
zL|}k?D%NEJPI8$)6F5n}+`SX{pDd@%JUmjzKN~ro-GRV><hY_F<GP~t<a+WFwX9)4
z^a7{%io$kuTkas>ozxyBc?_A(5~Zt5>C*eC17uuU>vIxy0@jY$^QKG1LQ;o37;Wk5
zJO*Ju6c?nmX9dwMhyx`h{beNsMkTP#L0YjUUoo*SP+g`56=Z4w1~RB@DB}UxpQ#5p
zknw_v;dG72`VoQG3LG*Q=yiF5YjWA)v<FaF2*i3Ja9H48fscZ!F&Lmg8>cQ>DYA2-
zL<Lm_>JdT3jogzwDrmT^d$dVGlQgiSjdzxnVwR#QG<i?>7_~vIt>VueeuZ4)4~#A>
z`3bv<7t8^iVHE(kY-hu3k<jtM3E!6EpszRyO!(MufVu1^1$z`?R8Ou&5-I;QdkUxp
zMGtSX7x-JF4_bZ)b9R-tj2$E|@l#{1<RiW|)=~Xq6vOO{u-~J^Wu6-wB6<GqSZ_Cr
zBGXzjyN3kWbrk#%!5<LZ;`_(vEI&tdhp&wf)E8>kH}zN?@+BOH{g(Dr;5WxldNKbB
zWI1SxC00`mD^NjRNwNQO_r$3tu}86u4^f-Y4y{{aR>G@29?K5_PhIBEPMmINBW<!u
zRM}1%i3&ZzEjN=|_;2Q4PMpl*yTJ@&7Dnn<0KnUlMl#SFG?|hHDW~M63<SAmD$lU{
znQ<m+y+vSV2oh^rhS0GHOaaUan3}dwIWOmIP}p&1-J8`*RH>D+X;O>1WfzM~ut*h{
z88Dkzpi+}zxovM&TwYp@nFbm(M~NzhFyovMDCVv1*>HSyGY-Ynw>WdghowfeUj1AI
z+Hcm-e;N2pT3~+zxa_YehK1AEx|tI_yaaq!ftVeE0|8b(_6CA+0Kv*)K?{j^fT~?4
zh<iqYy(E!^iQitVP#(|}bd_SL6bgIuKzV+Ds03Zbd6{3a!6L%$AU-yN<!L}nUBIOx
z+KOz~e3>8w*a19)0G)`v13-}>G8zGq6qUG$3j)cmz>s?oBR$X=c=eQ*-#F^aV!Z1T
z#o!^D43!T9M?tJu`E-*8fubQ_$yxGJRHh0%wB!saaUf(g^*)Xa9E=2e8do_W#FU(q
zLVQpQuFO?HctSu5{s9yqAOJ*%lve6tP-~3@Unzv;)&(J<@>7$(mMw5xDju2%C6ck!
zMJN<bpS@HmpltB9$;JmOMOL%Ra3)rXKqOeXu@nOV9=Diukes%7GMd_4k36<f?1u0z
zl-LDahcUtK@V6%itKS8*86D~bL{gdww@nRat$^bYN%#b;gRO#>C|NTxD1kZQMJhK)
zkQLyCoGDGH0$E|kq(VuNLzV=jA{-PF4j_6Ge2F@)5a!fv#S8(e6TEJsX7IXPq-M1W
zAiK(B%x_7n#O%&hrO|k`a66cT1#SaX0bkcp3tWbb3cXSWuUFEr2UX*Sy?76T(p3D`
z!B)*;yRbo%q~SKRl2!*D6qjNqJG~NKS{_a{D#?2%@sv(72j;P%K|9l&*nP#U0cNhe
zTrWZSu~2m!z6mH;nW9G;feU2`{Q=NgD*^&$O)zV$l^P*Fifm7j6>0`XGsKFGo2T0r
z@8VtIpQsSaOhHsda1sGtrKlSIeWGYUau?tGP;ZKrLDY>m_|CL{?q2waxDp|CFNW+Z
zVSV*nD22~7@uU(#R9}vy7tqw_{NL05nQ{jRTH`~-chsKURXxGp2g%aYCukE%j)jGJ
zQB?ziN`7vp%l#gRbzwYWx=V_aRPDYz^E#<%LJ_Qlv1#Bg`xW+geqpw)V+Tk~MWzjb
zIGCt~*hJYrq_yzdv%c&XIESB|Qo*qiVbJBmIy_O?B#;XV{NQ1t3*JI0agzPGR5Y|w
z8-wjvin<7I<4`<&F~U+`X}A>@2A!gurZ&JxIL_LMR4E1R#ReZ*P3Q~~dTxiDyrIq$
z(!^~Gt(G>lx5N7vn$7GNunYd}g=x~suPvN5w*YIMf3mPL^)zlTfZ*2%ev9Bw2*f+{
zGsM1w0Lvu9)Mn^U)`hxT2)};9!*h!@^L^@s>QVQrF4e2r)$?kD>QLL%4we96e{;c{
t*j(SZ$b`rKjO%}fKs@wS_*uAmu0N4rxA{L8FFLZw!Aik?!qX?#{tqzIIFbMW

delta 3559
zcmZ`+X>41^5q|sLBgIQMMbSEGNtR5@mPJ{%4j;00*p{O>ZE6&$(w3mLw4!H;QkG9i
z<C11k+ll|EP`yct6fr`^F6yF909$R4!axf2PwS#V&p%}kv_TxFr;$HxQ6NsgS<(`X
zrUZ80?Cdu?v$M0a%j^s6&R1DQti0T#;n%(E=4{8}^@^wJ+4J&3{i7M(GI%+!;FY|J
zSFbpkmeMSZ*Vrbn<$i9im_<pQy@3prAnSn)&T4$qtBg1B#uY~q-()*@GjCZjiv3ob
z@eto!GGiMgu1pD}U8!`GK(>?^?JP0c1y!AIEkTB9GQO=u(hbM9^BpCU9$<Fz-V#Z~
zo^<d&zN<vk4<~l>J?j!DAK-&4j8>5?%W0q0(|X!KjGTy>&WBcX8u^qK9+sn<?yY^J
zA!<0!I3nA3ZC`V?oMX;m3c4v}Yl?kmtt!;V+}G@yY)gt+e)-7eE*6qwn<p||)>Z_&
zhS#m=*)S&GmSZU`+70A(Aa~?giL_@H2BSNH@{MbsYe}2nc4obHZ;G8;$K~MgDtqz|
zcEsw-?#gK;`u&jYw)Q|PkR8Zrn13(>GK4mKO6P$UtS-8D9Td6``X9``bxiyJ#q5Wf
zy=eQ;_RrGH2ka5dbP(;(IUW5`d+eNpRx~cJw<TC!hT5NG|B{#5uj=<SIoQ!;<k`4<
zyrYf1AkTIjV-xaL#};-$-tFkG{}N&=9s<dG5;O7WeC(OHNQ#sk-tvwd?X0a6bp!|y
zz@3QOpAhj+&z+EGJNv6e6$$l3BJZ4vOf4pp@<!*@eJ_#FaWZblB85-iv~O|Ao?DnZ
z6Bp4Vk<#Ay1HFletnKQ9TZg;)oOdxza$KJ6>UQ<Ae!a>F8dui3bk?=@-S8l*xu&ex
zA~u&0-vcjh$a~uY<<s5`Q{!?|cVj5;oQG%8aq$>wm`OXnR0^7qQ{BM_)Q-rx?lHE1
z?X~VbI(&I|$7Af*a(_=V`~2GDJy&%$e<^ak>Ca@6f~I01-USyf6a6H(E#HppWbexl
zBdzSbZ0?(6KU|yb`$%_iPs-t!BmGVC-0ppuEH+cK!m9=xHmq&gZ8?oQPtzWDW2Af4
z5Lw%e%+;(;?ww^h+UCxpq-$N$g#%RwOa)q7syau>v7;z!SF#!<>j94p=f2bWswv(|
zX&I{`T?=7*+7H^Brb!~{0O*c%J!nrlh-vy#wdxS<QoLXB{fe*4G137d01kTK>ecve
z#RnBn2V?*Z`qm&-NOdlT(94h*Q<aE^(HlsSBO6dk`+*d>wIQQpK-w@=BBet@QZl)L
zJ;2>4E0t%a9$ij^PA!V3V}gg4Vn|hc;AIr=>cPW9-(uoL+PLRIjtd{L<wS0QoCveL
zXEts}?ZwmaL^3=gE<l<$?AWr%$}5A9IP(O3QT}^yJG&;g47IX%<it=%^*0G-BJ1Pg
zw}g03o*C+6ugceky4&+*r(&n$(b3-fOhmjwnBNikJ&`vr4Uao-5_ef18{Szz?O#~7
z=jNj#9-D_=(VJ<xIy@Qt2Vi-2V%nTb%p}D>si&VzivP$Dh97EF-YEw*CdXx>36sg{
zK*0GG@X7Nsx%c6^PS&Qom`nGt2IgWltj6#&zx>VK$&86Psg0o}1qgN+&0<+Sr(yR|
zX0t{L)G~9Z?C3dCI!DT8ohXf3v`md_maBk(#H}!z<>opxjFeGCdVm1h^3-eG@RP2W
zQq_w+G?$xU`Tf^(cV7S6?Tc4$U;Os%m#*FU`uWh2L&rn!t-Nyk>h;4X#9x64`@~=H
z5f_O}5xGI+BO-4R83LL17Ro}Dwn9aI+Px&=T!lD)e~{uWj=v3j*i@%GZxTLVfGX8$
zislxYR*_EPehPArB1;8QGpbkxtI=yfJ)m;u!Jz!&fl%hl7_$*YtQRq}jI5c{W)U~8
zujpx39C2X8`M`*d5sIiwMckdzb4J#~jU3}9X=c4SjqcX<afWfoyjserN8^rK4PWem
z4^#UPS*IKs3CZ+G!$kQ;H51RlJE2lVg-|Ob4U)?}zc`y*UWk8TsZe-TT%RU1#Z~-V
zzBkfa`7^LS9T8CPB|~-{+?64=3{9oc8<-FFwJbxTs1&mBG@|6O%(P)U@HEh888=oK
zJs2R>Etg+Y_q2|*F;APq!XiB>EY<QNi4dr)FXiA4q+NMY!>v-NBH`q?-!st)P7i`g
zHCYu+vC6~&n5u@P2F*nDb?!v;okb~CNdLN&GE=weRK5pNPV~GfC!z~;z}Lfp`_eK<
zgS-rT!f3vrmUg8wq`ZlgINPKJX+xBxvGApI#|jG$nwnK`6OqW414r&P1U2re#ewr{
z1Dbw+A&td_JO%SjK+$YChQ(epdj6bVuch5nDK~cyfk~L$Oe2Ia)q-KVI&`ZFw9>CK
z1HD%CT2v1w75&D-NHm0A2r1){9~^4S2<#zpgvj4O?x3mP3E{$5OcL=CQKxdVD3nSP
zP!SbWxC$Aj-VvsVCAjzy!n~2R`Lt&~ws4ZiqB!3ldw@NXkGAfno41gw8yV_38M9}e
zh$iX5q2305N?w(sS%P1bKN<~BP*t2Ro9~NKAy~IIV<*HfF%tF^ZgUT*Do<;OYamif
zq>e~HM#j1tegS-dlD!9vX82G<ovcCkv5GZo>?W%X5|m;vcMzx(VMF{~4o<XnT!u*C
zsjmtAW6(qy5f>2^o+jd&<=Kf)hTcK~k5Mh}U5X1_1u=VWF(LLtoHrKm0#aASn}kv^
z4w0&=`o&t0^;8z>mn(7qh4*oWTc9}YQ<X8f1S6<^ZA`1wraguGFFKR3MXTI?_+@tB
z%B{nU1?5kUv^qDFz030MktV#PX2hHw7w~T?BCC!bWLxAXk3Q^Z2IQo?a`fc>XUWY7
zkslNJ6_Gy@QEw_$xj#c)Dx!kwh@kk2PO{M~M<=&Ubh`#!0oP_%g)8XtxR&tsxVE}F
zL;|>QQ{ItSURZiYsG$9cRKHI|<(SJ}6f;lsBog8W^7`a6-m^49HA%c7cON_bUn!-S
A=l}o!

diff --git a/models/networks.py b/models/networks.py
index 74343e6..7519e80 100644
--- a/models/networks.py
+++ b/models/networks.py
@@ -331,6 +331,8 @@ def define_D(input_nc, ndf, netD, n_layers_D=3, norm='batch', init_type='normal'
         net = PixelDiscriminator(input_nc, ndf, norm_layer=norm_layer)
     elif 'stylegan2' in netD:
         net = StyleGAN2Discriminator(input_nc, ndf, n_layers_D, no_antialias=no_antialias, opt=opt)
+    elif netD == 'basic_cond':  # more options
+        net = NLayerDiscriminator_ncsn(input_nc, ndf, n_layers=3, norm_layer=norm_layer, no_antialias=no_antialias)
     else:
         raise NotImplementedError('Discriminator model name [%s] is not recognized' % netD)
     return init_net(net, init_type, init_gain, gpu_ids,
diff --git a/models/roma_unsb_model.py b/models/roma_unsb_model.py
index 2c275a7..8dbc273 100644
--- a/models/roma_unsb_model.py
+++ b/models/roma_unsb_model.py
@@ -200,6 +200,8 @@ class RomaUnsbModel(BaseModel):
         parser.add_argument('--lambda_NCE', type=float, default=1.0, help='weight for NCE loss: NCE(G(X), X)')
         parser.add_argument('--lambda_SB', type=float, default=0.1, help='weight for SB loss')
         parser.add_argument('--lambda_ctn', type=float, default=1.0, help='weight for content-aware temporal norm')
+        parser.add_argument('--lambda_D_ViT', type=float, default=1.0, help='weight for discriminator')
+        parser.add_argument('--lambda_global', type=float, default=1.0, help='weight for Global Structural Consistency')
         
         parser.add_argument('--nce_idt', type=util.str2bool, nargs='?', const=True, default=False, help='use NCE loss for identity mapping: NCE(G(Y), Y))')
         parser.add_argument('--nce_layers', type=str, default='0,4,8,12,16', help='compute NCE loss on which layers')
@@ -220,7 +222,7 @@ class RomaUnsbModel(BaseModel):
         parser.add_argument('--lambda_inc', type=float, default=1.0, help='incremental weight for content-aware optimization')
         parser.add_argument('--eta_ratio', type=float, default=0.1, help='ratio of content-rich regions')
 
-        parser.add_argument('--atten_layers', type=str, default='1,3,5', help='compute Cross-Similarity on which layers')
+        parser.add_argument('--atten_layers', type=str, default='5', help='compute Cross-Similarity on which layers')
         
         parser.add_argument('--tau', type=float, default=0.01, help='Entropy parameter')
         parser.add_argument('--num_timesteps', type=int, default=5, help='# of discrim filters in the first conv layer')
@@ -258,7 +260,7 @@ class RomaUnsbModel(BaseModel):
             self.visual_names += ['idt_B']
 
         if self.isTrain:
-            self.model_names = ['G', 'D', 'E']
+            self.model_names = ['G', 'D_ViT', 'E']
 
             
         else:
@@ -269,23 +271,25 @@ class RomaUnsbModel(BaseModel):
         self.netG = networks.define_G(opt.input_nc, opt.output_nc, opt.ngf, opt.netG, opt.normG, not opt.no_dropout, opt.init_type, opt.init_gain, opt.no_antialias, opt.no_antialias_up, self.gpu_ids, opt)
         
         
-        if self.isTrain:            
-            self.netD = networks.define_D(opt.output_nc, opt.ndf, opt.netD, opt.n_layers_D, opt.normD, opt.init_type, opt.init_gain, opt.no_antialias, self.gpu_ids, opt)
+        if self.isTrain:
             self.netE = networks.define_D(opt.output_nc*4, opt.ndf, opt.netD, opt.n_layers_D, opt.normD, opt.init_type, opt.init_gain, opt.no_antialias, self.gpu_ids, opt)
             
             self.resize = tfs.Resize(size=(384,384), antialias=True)
             
+            self.netD_ViT = networks.MLPDiscriminator().to(self.device)
+            
             # 加入预训练VIT
             self.netPreViT = timm.create_model("vit_base_patch16_384", pretrained=True).to(self.device)
             
             # 定义损失函数
+            self.criterionL1 = torch.nn.L1Loss().to(self.device)
             self.criterionGAN = networks.GANLoss(opt.gan_mode).to(self.device)
             self.criterionNCE = []
             for nce_layer in self.nce_layers:
                 self.criterionNCE.append(PatchNCELoss(opt).to(self.device))
             self.criterionIdt = torch.nn.L1Loss().to(self.device)
             self.optimizer_G = torch.optim.Adam(self.netG.parameters(), lr=opt.lr, betas=(opt.beta1, opt.beta2))
-            self.optimizer_D = torch.optim.Adam(self.netD.parameters(), lr=opt.lr, betas=(opt.beta1, opt.beta2))
+            self.optimizer_D = torch.optim.Adam(self.netD_ViT.parameters(), lr=opt.lr, betas=(opt.beta1, opt.beta2))
             self.optimizer_E = torch.optim.Adam(self.netE.parameters(), lr=opt.lr, betas=(opt.beta1, opt.beta2))
             self.optimizers = [self.optimizer_G, self.optimizer_D, self.optimizer_E]
             
@@ -320,10 +324,10 @@ class RomaUnsbModel(BaseModel):
         
         self.netG.train()
         self.netE.train()
-        self.netD.train()
+        self.netD_ViT.train()
         
         # update D
-        self.set_requires_grad(self.netD, True)
+        self.set_requires_grad(self.netD_ViT, True)
         self.optimizer_D.zero_grad()
         self.loss_D = self.compute_D_loss()
         self.loss_D.backward()
@@ -337,7 +341,7 @@ class RomaUnsbModel(BaseModel):
         self.optimizer_E.step()
         
         # update G
-        self.set_requires_grad(self.netD, False)
+        self.set_requires_grad(self.netD_ViT, False)
         self.set_requires_grad(self.netE, False)
         
         self.optimizer_G.zero_grad()
@@ -443,7 +447,7 @@ class RomaUnsbModel(BaseModel):
 
         # ============ 第三步：拼接输入并执行网络推理 =============
         bs = self.real_A0.size(0)
-        z_in = torch.randn(size=[2 * bs, 4 * self.opt.ngf]).to(self.real_A0.device)
+        z_in = torch.randn(size=[bs, 4 * self.opt.ngf]).to(self.real_A0.device)
         z_in2 = torch.randn(size=[bs, 4 * self.opt.ngf]).to(self.real_A1.device)
         # 将 real_A, real_B 拼接 (如 nce_idt=True)，并同样处理 real_A_noisy 与 XtB
         self.real = self.real_A0
@@ -455,9 +459,10 @@ class RomaUnsbModel(BaseModel):
                 self.real = torch.flip(self.real, [3])
                 self.realt = torch.flip(self.realt, [3])
         
-        
+        print(f'fake_B0: {self.real_A0.shape}, fake_B1: {self.real_A1.shape}')
         self.fake_B0 = self.netG(self.real_A0, self.time, z_in)
         self.fake_B1 = self.netG(self.real_A1, self.time, z_in2)
+        print(f'fake_B0: {self.fake_B0.shape}, fake_B1: {self.fake_B1.shape}')
 
         if self.opt.phase == 'train':
             real_A0 = self.real_A0
@@ -507,23 +512,35 @@ class RomaUnsbModel(BaseModel):
             #self.mutil_fake_B0_2_tokens = self.netPreViT(self.warped_fake_B0_2_resize, self.atten_layers, get_tokens=True)
 
             
-    def compute_D_loss(self):
-        """计算判别器的 GAN 损失"""
-        
-        fake = self.cat_results(self.fake_B.detach())
-        pred_fake = self.netD(fake, self.time)
-        self.loss_D_fake = self.criterionGAN(pred_fake, False).mean()
-        
-        self.pred_real = self.netD(self.real_B0, self.time)
-        loss_D_real = self.criterionGAN(self.pred_real, True)
-        self.loss_D_real = loss_D_real.mean()
-        
-        self.loss_D = (self.loss_D_fake + self.loss_D_real) * 0.5
-        return self.loss_D
+    def compute_D_loss(self):      #判别器还是没有改
+        """Calculate GAN loss for the discriminator"""
+
+        lambda_D_ViT = self.opt.lambda_D_ViT
+        fake_B0_tokens = self.mutil_fake_B0_tokens[0].detach()
+        fake_B1_tokens = self.mutil_fake_B1_tokens[0].detach()
+
+        real_B0_tokens = self.mutil_real_B0_tokens[0]
+        real_B1_tokens = self.mutil_real_B1_tokens[0]
+
+
+        pre_fake0_ViT = self.netD_ViT(fake_B0_tokens)
+        pre_fake1_ViT = self.netD_ViT(fake_B1_tokens)
+
+        self.loss_D_fake_ViT = (self.criterionGAN(pre_fake0_ViT, False).mean() + self.criterionGAN(pre_fake1_ViT, False).mean()) * 0.5 * lambda_D_ViT
+
+        pred_real0_ViT = self.netD_ViT(real_B0_tokens)
+        pred_real1_ViT = self.netD_ViT(real_B1_tokens)
+        self.loss_D_real_ViT = (self.criterionGAN(pred_real0_ViT, True).mean() + self.criterionGAN(pred_real1_ViT, True).mean()) * 0.5 * lambda_D_ViT
+
+        self.loss_D_ViT = (self.loss_D_fake_ViT + self.loss_D_real_ViT) * 0.5
+
+
+        return self.loss_D_ViT
 
     def compute_E_loss(self):
         """计算判别器 E 的损失"""
         
+        print(f'resl_A_noisy: {self.real_A_noisy.shape} \n fake_B0: {self.fake_B0.shape}')
         XtXt_1 = torch.cat([self.real_A_noisy, self.fake_B0.detach()], dim=1)
         XtXt_2 = torch.cat([self.real_A_noisy2, self.fake_B1.detach()], dim=1)
         temp = torch.logsumexp(self.netE(XtXt_1, self.time, XtXt_2).reshape(-1), dim=0).mean()
@@ -534,14 +551,8 @@ class RomaUnsbModel(BaseModel):
     def compute_G_loss(self):
         """计算生成器的 GAN 损失"""
         
-        bs =  self.real_A0.size(0)
-        tau = self.opt.tau
-        
-        fake = self.fake_B0
-        std = torch.rand(size=[1]).item() * self.opt.std
-        
         if self.opt.lambda_GAN > 0.0:
-            pred_fake = self.netD(fake, self.time)
+            pred_fake = self.netD_ViT(self.mutil_fake_B0_tokens[0])
             self.loss_G_GAN = self.criterionGAN(pred_fake, True).mean() * self.opt.lambda_GAN
         else:
             self.loss_G_GAN = 0.0
@@ -555,7 +566,7 @@ class RomaUnsbModel(BaseModel):
             # eq.9
             ET_XY    = self.netE(XtXt_1, self.time, XtXt_1).mean() - torch.logsumexp(self.netE(XtXt_1, self.time, XtXt_2).reshape(-1), dim=0)
             self.loss_SB = -(self.opt.num_timesteps - self.time[0]) / self.opt.num_timesteps * self.opt.tau * ET_XY
-            self.loss_SB += self.opt.tau * torch.mean((self.real_A_noisy - self.fake_B) ** 2)
+            self.loss_SB += self.opt.tau * torch.mean((self.real_A_noisy - self.fake_B0) ** 2)
         
         if self.opt.lambda_global > 0.0:
             loss_global = self.calculate_similarity(self.real_A0, self.fake_B0) + self.calculate_similarity(self.real_A1, self.fake_B1)
diff --git a/options/__pycache__/base_options.cpython-39.pyc b/options/__pycache__/base_options.cpython-39.pyc
index 55ab7e1a65961dcda63124aa2f74d66d46317a27..7658be9386c7c6039866b5ccc650f594579220ce 100644
GIT binary patch
delta 327
zcmZ2yy~~<6k(ZZ?0SGv)cBik~$g9RF$(58?oS7V-oS&DX$zCK1l)J?W7TWB<X~)PY
zF}a>gfl+Gm0<IKB>B+3ziy7rMujGEq$!IWnx@e-LK1e-RL1IyHYJ6f*da)))5j#+*
zNPn}S*ls39&B<pa_9=iBai^!2#HXaDC6?xt6lsIBv8Ck~<tCOCSx#OpnZjtZSwQM0
zqo*cFw=IzP#i&!n2&6RyitIpQ7C_<_TXAVxT4qI&D@eo%M0kRT5Re>udO>M?W=e4p
z8%RiJvXP9IkPVQj$p&!}h+PymIZtMhfB}dJHWj3^C~mTXtfjmWi0K9*m_S4nh%g5c
PRv>~Is78EpzpOj}7!Xc}

delta 306
zcmdmGz0R69k(ZZ?0SG*Q?n<A%kynkAmo+J|I5Sz3tw<CoyxE7-j*(G(ayOR(qvYf@
zTq%rFlLfgKGs<q>%Kes;QGfDc(L_l-pu$^R1&Kw)squ+L>BX8HMeIPKBE8MBV!N3b
zH74Jb*rxzi#GRg65}%TqmROooQltga#+H^}l$%&mWHEWWWD2A8W*MoQjGh`G-8Mkt
z7o$!QBaqe<D6$2KnFEPiY{jK%X_*y8E+7#{5a9tLf<bcZ=>?_nnJL9ZY#<@+$#ybY
zLe@Yg&^ur!f!IZ%ldEJF3Fw2EU{gUli()71$XZGnf|#x#f(b-Kf(SDZVL7>8)*1i=
Cnn|Jn

diff --git a/options/base_options.py b/options/base_options.py
index f9de39b..b20e1b4 100644
--- a/options/base_options.py
+++ b/options/base_options.py
@@ -35,7 +35,7 @@ class BaseOptions():
         parser.add_argument('--output_nc', type=int, default=3, help='# of output image channels: 3 for RGB and 1 for grayscale')
         parser.add_argument('--ngf', type=int, default=64, help='# of gen filters in the last conv layer')
         parser.add_argument('--ndf', type=int, default=64, help='# of discrim filters in the first conv layer')
-        parser.add_argument('--netD', type=str, default='basic', choices=['basic', 'n_layers', 'pixel', 'patch', 'tilestylegan2', 'stylegan2'], help='specify discriminator architecture. The basic model is a 70x70 PatchGAN. n_layers allows you to specify the layers in the discriminator')
+        parser.add_argument('--netD', type=str, default='basic_cond', choices=['basic_cond', 'basic', 'n_layers', 'pixel', 'patch', 'tilestylegan2', 'stylegan2'], help='specify discriminator architecture. The basic model is a 70x70 PatchGAN. n_layers allows you to specify the layers in the discriminator')
         parser.add_argument('--netG', type=str, default='resnet_9blocks_cond', choices=['resnet_9blocks','resnet_9blocks_mask', 'resnet_6blocks', 'unet_256', 'unet_128', 'stylegan2', 'smallstylegan2', 'resnet_cat', 'resnet_9blocks_cond'], help='specify generator architecture')
         parser.add_argument('--n_layers_D', type=int, default=3, help='only used if netD==n_layers')
         parser.add_argument('--normG', type=str, default='instance', choices=['instance', 'batch', 'none'], help='instance normalization or batch normalization for G')
diff --git a/scripts/train.sh b/scripts/train.sh
index 93a5f96..dea6a51 100755
--- a/scripts/train.sh
+++ b/scripts/train.sh
@@ -28,6 +28,6 @@ python train.py \
     --num_patches 256 \
     --flip_equivariance False \
     --eta_ratio 0.1 \
-    --tau 0.1 \
+    --tau 0.01 \
     --num_timesteps 10 \
     --input_nc 3