caffe example autoencoder fall into stagnation

[Nick Young]

when i run the     ~/caffe/examples/mnist/train_mnist_autoencoder_adadelta.sh  ,       and i just modified the solver prototxt a little bit as bellow：

～/caffe/examples/mnist/mnist_autoencoder_solver_adadelta.prototxt

net: "examples/mnist/mnist_autoencoder.prototxt"

test_state: { stage: 'test-on-train' }

test_iter: 500

test_state: { stage: 'test-on-test' }

test_iter: 100

test_interval: 500

test_compute_loss: true

base_lr: 1.0

lr_policy: "fixed"

momentum: 0.95

delta: 1e-8

display: 100

# max_iter: 65000

max_iter: 1000

weight_decay: 0.0005

snapshot: 10000

snapshot_prefix: "/home/nick/Desktop/mnist_autoencoder_adadelta_train"

# solver mode: CPU or GPU

solver_mode: CPU

type: "AdaDelta"

and it stopped after these ：

I1106 22:13:53.124932  4140 layer_factory.hpp:76] Creating layer data

I1106 22:13:53.124995  4140 net.cpp:106] Creating Layer data

I1106 22:13:53.125005  4140 net.cpp:411] data -> data

why ?

[Belhassen Bayar]

Hi Nick,

did you solve the problem. I encountered the same problem.

I would really appreciate it if you could share how you fixed it.

[Nick Young]

yes, i just solved it. just type "caffe autoencoder bug" in google and search, you will get the answer on the top.

inspired by the site, i just copy the "/examples/mnist/mnist_train_lmdb" to the same folder, so i got 3 lmdb folders :

 "/examples/mnist/mnist_test_lmdb"

 "/examples/mnist/mnist_train_lmdb"

 "/examples/mnist/mnist_train_lmdb(copy)"

and then i modified the network definition prototxt, just the 2nd layer as :

layer {

  name: "data"

  type: "Data"

  top: "data"

  include {

    phase: TEST

    stage: "test-on-train"

  }

  transform_param {

    scale: 0.0039215684

  }

  data_param {

    source: "examples/mnist/mnist_train_lmdb (copy)"

    batch_size: 100

    backend: LMDB

  }

}

then it worked.

it seems like a lock on  "/examples/mnist/mnist_train_lmdb".

在 2016年3月3日星期四 UTC+8上午3:14:25，Belhassen Bayar写道：