andr3/gotch
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

chore(example/cifar): reconfigure optimizer inside epoch for loop

Browse Source
This commit is contained in:
sugarme 2020-07-10 21:24:21 +10:00
parent 18873a8957
commit c265ba007e
1 changed files with 32 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
example/cifar/main.go
View File

@ -109,20 +109,40 @@ func main() {
net := fastResnet(vs.Root())
optConfig := nn.NewSGDConfig(0.9, 0.0, 5e-4, true)
opt, err := optConfig.Build(vs, 0.01)
if err != nil {
log.Fatal(err)
}
// optConfig := nn.NewSGDConfig(0.9, 0.0, 5e-4, true)
// opt, err := optConfig.Build(vs, 0.01)
// if err != nil {
// log.Fatal(err)
// }
var lossVal float64
startTime := time.Now()
for epoch := 0; epoch < 150; epoch++ {
opt.SetLR(learningRate(epoch))
// opt.SetLR(learningRate(epoch))
optConfig := nn.NewSGDConfig(0.9, 0.0, 5e-4, true)
var opt nn.Optimizer
var err error
switch {
case epoch < 50:
opt, err = optConfig.Build(vs, 0.1)
if err != nil {
log.Fatal(err)
}
case epoch < 100:
opt, err = optConfig.Build(vs, 0.01)
if err != nil {
log.Fatal(err)
}
case epoch >= 100:
opt, err = optConfig.Build(vs, 0.001)
if err != nil {
log.Fatal(err)
}
}
iter := ts.MustNewIter2(ds.TrainImages, ds.TrainLabels, int64(64))
// iter := ts.MustNewIter2(ds.TrainImages, ds.TrainLabels, int64(512))
// iter := ts.MustNewIter2(ds.TrainImages, ds.TrainLabels, int64(64))
iter := ts.MustNewIter2(ds.TrainImages, ds.TrainLabels, int64(512))
iter.Shuffle()
// iter = iter.ToDevice(device)
@ -137,26 +157,21 @@ func main() {
bimages := vision.Augmentation(devicedData, true, 4, 8)
logits := net.ForwardT(bimages, true)
// logits := net.ForwardT(devicedData, true)
// logits := net.ForwardT(item.Data.MustTo(vs.Device(), true), false)
loss := logits.CrossEntropyForLogits(devicedLabel)
opt.BackwardStep(loss)
lossVal = loss.Values()[0]
// logits.MustDrop()
// item.Data.MustDrop()
// item.Label.MustDrop()
devicedData.MustDrop()
devicedLabel.MustDrop()
bimages.MustDrop()
loss.MustDrop()
}
testAcc := ts.BatchAccuracyForLogits(net, ds.TestImages, ds.TestLabels, vs.Device(), 512)
fmt.Printf("Epoch:\t %v\t Loss: \t %.3f \tAcc: %10.2f%%\n", epoch, lossVal, testAcc*100.0)
// fmt.Printf("Epoch: %10.0d\tLoss:%10.3f\n", epoch, lossVal)
// testAcc := ts.BatchAccuracyForLogits(net, ds.TestImages, ds.TestLabels, vs.Device(), 512)
// fmt.Printf("Epoch:\t %v\t Loss: \t %.3f \tAcc: %10.2f%%\n", epoch, lossVal, testAcc*100.0)
fmt.Printf("Epoch: %10.0d\tLoss:%10.3f\n", epoch, lossVal)
iter.Drop()
/*
@ -174,6 +189,6 @@ func main() {
}
testAcc := ts.BatchAccuracyForLogits(net, ds.TestImages, ds.TestLabels, vs.Device(), 512)
fmt.Printf("Loss: \t %.3f\t Accuracy: %10.2f%%\n", lossVal, testAcc*100.0)
fmt.Printf("Accuracy: %10.2f%%\n", testAcc*100.0)
fmt.Printf("Taken time:\t%.2f mins\n", time.Since(startTime).Minutes())
}