Initial commit

layers/layer-params-170-quant.cfg

@@ -0,0 +1,234 @@
+[conv1a]
+epsW=0.0
+epsB=0.0
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0.00
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv1b]
+epsW=0.0
+epsB=0.0
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0.00
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv2a]
+epsW=0.01
+epsB=0.02
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0.00
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv2b]
+epsW=0.01
+epsB=0.02
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0.00
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv3a]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv3b]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv4a]
+epsW=0.01
+epsB=0.02
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv4b]
+epsW=0.01
+epsB=0.02
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv5a]
+epsW=0.01
+epsB=0.02
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[conv5b]
+epsW=0.01
+epsB=0.02
+momW=0.9
+momB=0.9
+wc=0.0005
+wball=0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[fc2048a]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[fc2048b]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[fc2048ba]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[fc2048bb]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[fc1000]
+epsW=0.01,0.01
+epsB=0.02
+momW=0.9,0.9
+momB=0.9
+wc=0.0005,0.0005
+wball=0,0
+schedW=dexp[tgtFactor=250,noiseStdev=0,numSteps=4]
+schedB=dexp[tgtFactor=10,noiseStdev=0,numSteps=2]
+
+[logprob]
+coeff=1
+topk=5
+
+[hs1a]
+enable=true
+
+[hs2a]
+enable=true
+
+[hs1b]
+enable=true
+
+[hs2b]
+enable=true
+
+[rnorm1a]
+scale=0.0001
+pow=0.75
+minDiv=2
+
+[rnorm1b]
+scale=0.0001
+pow=0.75
+minDiv=2
+
+[rnorm2a]
+scale=0.0001
+pow=0.75
+minDiv=2
+
+[rnorm2b]
+scale=0.0001
+pow=0.75
+minDiv=2
+
+[cnorm2a]
+scale=0.001
+pow=0.75
+
+[cnorm2b]
+scale=0.001
+pow=0.75
+
+[pool2a]
+quantF=half
+quantB=half
+
+[pool2b]
+quantF=half
+quantB=half
+
+[pool3a]
+quantF=half
+quantB=half
+
+[pool3b]
+quantF=half
+quantB=half
+
+[hs1a]
+quantF=half
+quantB=half
+
+[hs1b]
+quantF=half
+quantB=half
+
+[hs2a]
+quantF=half
+quantB=half
+
+[hs2b]
+quantF=half
+quantB=half
+
+# this is like #170 (def file 153) but uses half quantization on all communicated layers
+# logs/layers-170-quant.log
+# /nobackup/kriz/tmp/ConvNet__2013-02-21_22.33.24
+# guppy9
+# epoch 68: set color noise to 0 from 0.1
+# epoch 74: set epsw conv1 to 0 from 0.01
+#           logprob:  1.861123, 0.425700, 0.199970 
+# multiview logprob:  1.749848, 0.405240, 0.184160