From e0de909d62965c9d259ae515bb0135e49f93cbfe Mon Sep 17 00:00:00 2001
From: c <lianghongkey@gmail.com>
Date: Thu, 19 Dec 2019 16:07:29 +0800
Subject: [PATCH] Revert "Refine kernal Random to w*h*16 kinds possiable."

This reverts commit 675cd8567e8ba0643e8c63d60324ac84b5b1f218.
---
 FilterEvaluator/Evaluator.py                |   1 +
 FilterEvaluator/EvaluatorUnsuper.py         | 136 ++++++++++++--------
 FilterEvaluator/bestweightEntropySearch.npy | Bin 416 -> 416 bytes
 FilterEvaluator/checkpointEntropySearch.pkl | Bin 1836 -> 1836 bytes
 4 files changed, 80 insertions(+), 57 deletions(-)

diff --git a/FilterEvaluator/Evaluator.py b/FilterEvaluator/Evaluator.py
index bd13e97..937b9bd 100644
--- a/FilterEvaluator/Evaluator.py
+++ b/FilterEvaluator/Evaluator.py
@@ -16,6 +16,7 @@ import random
 import cv2
 
 
+
 CurrentPath = os.path.split(os.path.realpath(__file__))[0]+"/"
 print("Current Path :" + CurrentPath)
 
diff --git a/FilterEvaluator/EvaluatorUnsuper.py b/FilterEvaluator/EvaluatorUnsuper.py
index 4cd111d..608a878 100644
--- a/FilterEvaluator/EvaluatorUnsuper.py
+++ b/FilterEvaluator/EvaluatorUnsuper.py
@@ -16,7 +16,6 @@ import random
 import cv2
 from tqdm import tqdm
 import threading
-from itertools import combinations
 
 
 CurrentPath = os.path.split(os.path.realpath(__file__))[0]+"/"
@@ -27,19 +26,6 @@ import Model as Model
 from tools import utils, Train, Loader
 
 
-
-
-
-
-# iterable = combinations(range(72), 4)
-# count = iterable.count()
-# pool = tuple(iterable)
-# n = len(pool)
-# indices = sorted(random.sample(range(n), 2))
-# fdafda = tuple(pool[i] for i in indices)
-
-
-
 def GetScore(netmodel,layer,SearchLayer,DataSet):
     netmodel.eval()
     sample = utils.SetDevice(torch.empty((SearchLayer.out_channels,0)))
@@ -198,12 +184,10 @@ def UnsuperLearnFindBestWeight(netmodel, layer, weight, dataloader, databatchs=1
 
 
 # search best weight from random data
-# Test total sample size = stepsize * stepsize * interation
-# Random kernel number = in_channels * kernel_size[0] * kernel_size[1] * 16
-def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, stepsize=20, interation=1000):
+def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, stepsize=1000, interation=1000):
     interationbar = tqdm(total=interation)    
     forwardlayer = layer -1
-    samplesize = stepsize*stepsize
+
     netmodel.eval()
     tl = netmodel.features[layer]
     outchannels = tl.out_channels
@@ -227,55 +211,93 @@ def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, st
     for i in range(outchannels):
         shift.append(1<<i)
     shift = utils.SetDevice(torch.from_numpy(np.array(shift).astype("uint8")))
-
-    newweightshape = list(newlayer.weight.data.shape)
-    samplekernelsize = newlayer.in_channels*newlayer.kernel_size[0]*newlayer.kernel_size[1]*16
-    newweightshape[0] = samplekernelsize
-    newweight = np.random.uniform(-1.0,1.0,newweightshape).astype("float32")
-    newweight = newweight.reshape((-1,newweightshape[-1]*newweightshape[-2]))
-    newweight = np.swapaxes(newweight,0,1)-np.mean(newweight,-1)
-    newweight = np.swapaxes(newweight,0,1).reshape(newweightshape)
-    newlayer.weight.data=utils.SetDevice(torch.from_numpy(newweight))
-    outputs = newlayer(datasnetout).transpose(0,1)
-
-    indexs = np.random.randint(0, samplekernelsize, samplesize*interation*outchannels).reshape(interation,samplesize, -1)
-
-    # da = np.random.randint(0,5,2*9).reshape(9,2)
-    # te = np.sort(da,1)
-    # tg = np.unique(te ,axis=0)
-    # list(combinations(range(newweightshape[0]), 8))
-
+    # shift = torch.from_numpy(np.array(shift).astype("uint8"))
+    
     bestweight = []
     bestentropy = [100.0]
+    bittedset = []
+    bittedLock = threading.Lock()
+    bestLock = threading.Lock()
+
+    class CPU (threading.Thread):  
+        def __init__(self):
+            threading.Thread.__init__(self)
+        def run(self):
+            hasdata = 0
+            bittedLock.acquire()
+            if len(bittedset)>0:
+                bitted = bittedset.pop()
+                hasdata = 1
+            bittedLock.release()
+            if hasdata > 0:
+                entropys = []
+                for i in range(len(indexs)):
+                    histced = bitted[:,i].histc(256,0,255).type(torch.float32)
+                    histced = histced[histced>0]
+                    histced = histced/histced.sum()
+                    entropy = (histced.log2()*histced).sum()
+                    entropys.append(entropy.numpy())
+                argmin = np.argmin(entropys)
+
+                bestLock.acquire()
+                if entropys[argmin] < bestentropy[0]:
+                    bestweight = newweight[indexs[argmin]]
+                    bestentropy[0] = entropys[argmin]
+                    print("finded better entropy")
+                bestLock.release()
 
     for j in range(interation):
-        # 400 8 4096 5 5
-        output = outputs[indexs[j]]
-        # 102400 400 8        
-        reshaped = output.reshape(samplesize, outchannels, -1).permute(2, 0, 1)
-        # 400 8
-        meaned = reshaped.mean(0)
-        # 102400 400
-        # bitted = ((reshaped > meaned)* shift).sum(2).type(torch.float32).detach().cpu()
-        bitted = ((reshaped > meaned)* shift).sum(2).type(torch.float32)
+        newweightshape = list(newlayer.weight.data.shape)
+        newweightshape[0] = stepsize
+        newweight = np.random.uniform(-1.0,1.0,newweightshape).astype("float32")
+        newweight = newweight.reshape((-1,newweightshape[-1]*newweightshape[-2]))
+        newweight = np.swapaxes(newweight,0,1)-np.mean(newweight,-1)
+        newweight = np.swapaxes(newweight,0,1).reshape(newweightshape)
         
-        entropys = []
-        for i in range(samplesize):
-            histced = bitted[:,i].histc(256,0,255).type(torch.float32)
-            histced = histced[histced>0]
-            histced = histced/histced.sum()
-            entropy = (histced.log2()*histced).sum()
-            entropys.append(entropy.detach().cpu().numpy())
-        argmin = np.argmin(entropys)
-        if entropys[argmin] < bestentropy:
-            bestweight = newweight[indexs[j][argmin]]
-            bestentropy = entropys[argmin]
+        newlayer.weight.data=utils.SetDevice(torch.from_numpy(newweight))
+        outputs = newlayer(datasnetout).transpose(0,1)
+        samplesize = stepsize*stepsize
+        indexs = np.random.randint(0,stepsize,samplesize*outchannels).reshape(samplesize,-1)
+        # 1000 8 4096 5 5
+        outputs = outputs[indexs]
+        # 102400 1000 8        
+        reshaped = outputs.reshape(samplesize, outchannels, -1).permute(2, 0, 1)
+        # 1000 8
+        meaned = reshaped.mean(0)
+        # 102400 1000
+        bitted = ((reshaped > meaned)* shift).sum(2).type(torch.float32).detach().cpu()
+        
+        bittedLock.acquire()
+        bittedset.append(bitted)
+        bittedLock.release()
+        threadcpu = CPU()
+        threadcpu.start()
+
+        # entropys = []
+        # for i in range(len(indexs)):
+        #     histced = bitted[:,i].histc(256,0,255).type(torch.float32)
+        #     histced = histced[histced>0]
+        #     histced = histced/histced.sum()
+        #     entropy = (histced.log2()*histced).sum()
+        #     entropys.append(entropy.detach().cpu().numpy())
+        
+        # argmin = np.argmin(entropys)
+        # if entropys[argmin] < bestentropy:
+        #     bestweight = newweight[indexs[argmin]]
+        #     bestentropy = entropys[argmin]
+
         interationbar.update(1)
-        interationbar.set_description("entropy:"+str(bestentropy))
+        interationbar.set_description("left:"+str(len(bittedset)))
+
+    while bittedset:
+        time.sleep(100)
+    
     interationbar.close()
     return bestweight
 
 
+
+
 def SetModelConvWeight(model, layer, weight):
     w = utils.SetDevice(torch.from_numpy(weight))
     model.features[layer].weight.data = w
diff --git a/FilterEvaluator/bestweightEntropySearch.npy b/FilterEvaluator/bestweightEntropySearch.npy
index c9e7d788827196874fd5107fb73e37749004474b..5f64aa961998aca9c5e662970679899674b7125a 100644
GIT binary patch
delta 297
zcmV+^0oMMY1E2$tfPcvyMnBN@4nG>zgTDA9<h~z{8a^+tLBCea5Wm;dl0Blz@IIzR
zNWbbvGCp<v&OT>n(LNqFK|j&tAHOOP4L{D5g};&}AwCG9n!Yf7f4ny_r#`IU89!<|
zMZfuHeZF!zmpzVLl0HyEMLiO>9zQ-QM84i*20t9F?Y&`WFn_<cs4>2Z#r(doioHAm
z-u*s$^9()}kTbtaekVV2)#E-I<&i#{RW!fQLr=dxRu4Z->j*yV$z(s;y<xxThR(gv
z!W}+_iYh+1(HK9h??OHQDFZ&<jDWvd`Qko^>8w0wyn#RH*ag2ay$`;Sk#|38sEods
vR#m@Odn~_*Eh51_6XwmnwE|DSDgZV=3yw=Zgjr)gg~brRBKj-8lw@!}_;-t5

delta 297
zcmV+^0oMMY1E2$tfPcsi*FM49$3DWsx4y)q*}bq@n!Vzs4ZqBHE53N)gFi7-6u&Ar
zyS^d2D?UJiwZ6pj5kI5}SU)o3&Ay=7y1u&rt3Kdj+rDPIJ3g}1A3n1`Y(KgT^uBII
z_&&DIvc8;RRX&|h8^6Ti1;30jJ-jhc*gf+mfj$})k-i(tPk%pI8&<yy@*q7x4;sHQ
zG6p|3_S-&9Cs97w9nQY@)-b<`6+=ItxNts@9icup@>;)uus}Nl+TK1W-*P^!W~)A0
z^s~N@3<$kgrx?Hc4<Nr=g~UGIU0gmYB_TgjOE<p}I!3(`Vh=xYN{l}8OdURf<SD<T
vZ|=S;myJI5)grOJ4Z2Q0m*XM7<Yq!Xb*wMH;nFm}$jE&^QpfJT2bEbrp?Ho9

diff --git a/FilterEvaluator/checkpointEntropySearch.pkl b/FilterEvaluator/checkpointEntropySearch.pkl
index a104d5b3fcd419fc9089a1e794d7b699273e7930..dbcdf121302977804f1ac98002d4bb6a7b564c49 100644
GIT binary patch
delta 1062
zcmX|<c~F#f9L9H7S`m>$4iN=R3WNYzSdOLreP19!0v#9)Pyz=PMX@oU)FGi|<3_|~
zIa4A8<Y<<ooRER{`+FfGMdTU}4x623Ig>DoGEmTR9DDx!JkLCTJoBjZRC+BUzP3c{
z;p*xk78gqRCxy;#E)rLl!g~HGEq9UF<N7ASHooi?^i1F%(BtuV60yWZ?B?PA()Sc6
z2>yreF0ZQ+)*N#R_%HfW<r$;Ww^p_B?rI*U8|9-$9Y{hadPq@IAG=|$l$g?7{7`?M
zbV=fIXkn2B?ktIVt|6@8q13@9iyi2ZHDKJ2$FV(+!=&44vP;;>M%?{`G);}+5q~kx
z?Vl!Jd7Z-TttUtn?+$q&{SLhk=5%ykt!wu?J5P9JKX(Ye&O)(d4^E7Hgof)!NOD#h
zrmd!v1K%b}=eLZZs^u6uFBM~>ZUtHD7-Y+O{$vdrSquw`CRH-b#^j0j(d>pUJF6*V
zwR);?$x(w_q8o_MwHR!kDJ3BS2R8C_8{X0+u};^L$&tZ9Op%mhd~^*q7?zU1Oeduk
z@**sEs*oD#b&(G)K1Sp20@lSo4I^F+#fy$HM&8a#9nr!i)ZJW0N`vI^6!{>Fw<^Ji
zHXV34aFQx#;^2H=3|&Yt03GW@NDXUZi~gyhyu?r0%3EvMlI1waJS3%cXBz0lPjYBh
zT!BqH%Aj}gOYpj`puK+-0<X$K+E3EJ+M)sUM8#C--iOBOUa~xN0TQ44z~z0xz&=<e
z8y&-;I-pWUBkfD51-%F%A3tMxTAGfP_DfhY|2D0)Jq<$f5Z-Lgq{86c(0?u!Y-YyL
zTVEf$*)-ZFFa%S}ccEN9hZs@|-qraX;W@=nc_<VWdpzjP*m97!$icd<26q2lOOF=>
z(!TM%)YHrwA1eZ>o3aL+jAc4BGN~Jt`%=KBHJd)E_JnV8wnC>#J3IZ%jvB06C3Epk
zbaZ?jxo7nS?zpiQ!%U)JTl#)F`^bk@`Y|NFxRsPl*;4;VUs$nvL0+G;*TSC}dZNde
z-%q1U)|1pK-yC)}`*2}*47ibdN>X)u0cvG_0o)^h7dW#z0I{*dVCn9|3{P5d_68f6
zfuamfY4av%U%gD7Q)H}%s18)WY-ZxT{J7=V6q+b0CHav`2=VnJ8(S?Iu-iq|M}^GR
z#(6w^!kKHW7t!3F721-EQoec>K3miU6NM`m^tmmQcz*)SrwSnXN~o9{dM}uRFnwn4
zoHqQ{8^es_VQw~a0Ed3CfXrNbXd5<TRKNR@7j|nvotMf*ENdY9SCurDl{2%WZ*U^i
zr;37eOk8~)Ee#9h!p}Q!*^XUcG~G|X7X)+mJa=Y&eh{P|JPH{-2Hf_i?{Gyi@yx$m
Co9R~o

delta 1062
zcmX|-drXsO7{yy*<t89QS4Tx)lUrffLMim}J|8MG3{lVxheaV{ia3=GhL;$&QgFZ`
zuqtwq4NwOv104d2wBPf7z#73!fq5I3i6~%*OA14Ri2Gw0<IZ0vC%>GWoCb#mhej#i
zI#?zT4GNJ<iWU4l0(p=kNETXrfxp*MA`4M83m3WbwenXsRF=k%bKvoK3Yj8M798@L
z3{6|~e<%-pQ<Y}1D?qb=zLL-3RBBK%FjtaIx8!^d_nb>WYo!Cjik)CMA_AX?r8IY!
z7kq2;5nQ(($CMH&8L^{S^v7DLdMcwuKWY!r-1ME4H)kX>njHKk=u>!Lt_DfeJSz$I
zfGT+`T%B+Q(dvCvZ#$3WBpvyWvw$}rN%|mwYEF4WR$T-;;%G&d*A`IQ>2%PioKg4B
zI)U?wl~f?;Ld!4{adFmB&-ZE|d{rgs&Pc%Pg|;-l<sI6m-CBnS&t9O7Uo4E6KE-Bj
zv{jWmZ>HVv3(4u{i)iCOGwaV>pk}j>X1=Vz-iT5vYS)8U@h;4+8N^8E&%od37Kwbp
zK-ju$)!)AzqZ6^Ixc<Q{?5i3lP5*uX`8VuwNvw|Q*PN%qaUGqAFQwKsCA2E<6RJJ;
zupQgtPC|^`Fdn=qrr{RVq<pahCj7b|8jG%w(EVp=^aeeRc>Or9t@9GG_AkQYMZ;w5
zP8*7aJILu6vC8CXLW8m#8)AFZf({8OJXVRlt18f2@DFYpwIuCrJXKvl9w`XAtIj{#
zh3zY+l)BzREnYJ#*eM4kx+e;agQYB{xW8ai#XRC8RIxo7LJY5LQny!4vSor6;+}S&
zmC8oR(MzK!Q`}Sg9;hSf23Nf7n}NlvOGs90o9eF(lgM1$joUNF)CpxzFn2heSlZR7
zi{0u;zE_0$Y2qf57_k>i=TBmcpLU83yWwvv=rFUnvDZ;D8Hn^$BDrr7MLex4$PdPI
z_~h#q<c_j}xS58@o+GzO?vi%Qatpx|yKT?{ldx8;C#@|D<vT~%6M1@k^HjIPpU^SX
z1)ZI()b#TJHfk#HajiXf^N=4s{3wT!e!P?uYeS6G!{`TL*;lCPnG;iE6UPLO-lC@m
z4AjW0hmoWqnru`-(_J@cX7(|L#WOVZ!#TLx(?owtiiOH(5odbz7FQr8tTd*dJoD0l
z%d+EC^m{b3O|+G>n|3E*jxL-uP{7D%hhU@kX6SeAq0jG2xbK&(V=}9+eFb{gGa#SV
za9NFEDvQBk@OhXH7nAEjGt*4<i>)#4{BrK4qdOP>oe73_YzMKY2eV~E2<N0+k8dYA
zGHH<oFdgLzx-nm_=1Vj6D=UQAdzY!bVG~m}>cNdAuZ05*tGP$FWn6dnH_VnUGqoz&
G#{368q}<8?