SINGA-152 Worker class needs to take greater responsibility of its setting up

examples/cifar10/cudnn_hybrid.conf

@@ -1,10 +1,10 @@
 name: "cifar10-convnet"
-train_steps: 10000
+train_steps: 1000
 test_steps: 0
 test_freq: 200
 #validate_steps: 100
 #validate_freq: 300
-disp_freq: 200
+disp_freq: 50
 gpu: 0
 gpu: 1
 #debug: true
@@ -43,7 +43,6 @@ neuralnet {
       shape: 32
     }
     include: kTrain
-    partition_dim: 0
   }
 #  layer{
 #    name: "data"
@@ -73,7 +72,6 @@ neuralnet {
       shape: 32
     }
     include: kTest
-    partition_dim: 0
   }
 
   layer {

examples/cifar10/gpu.conf

@@ -0,0 +1,282 @@
+name: "cifar10-convnet"
+train_steps: 1000
+test_steps: 100
+test_freq: 200
+#validate_steps: 100
+#validate_freq: 300
+disp_freq: 50
+gpu:0
+
+
+#checkpoint_path: "examples/cifar10/checkpoint/step1000-worker0"
+train_one_batch {
+  alg: kBP
+}
+updater{
+  type: kSGD
+  weight_decay:0.004
+  momentum:0.9
+  learning_rate {
+    type: kFixedStep
+    fixedstep_conf:{
+      step:0
+      step:60000
+      step:65000
+      step_lr:0.001
+      step_lr:0.0001
+      step_lr:0.00001
+    }
+  }
+}
+neuralnet {
+  layer{
+    name: "data"
+    type: kRecordInput
+    store_conf {
+      backend: "kvfile"
+      path: "examples/cifar10/train_data.bin"
+      mean_file: "examples/cifar10/image_mean.bin"
+      batchsize: 100
+      #random_skip: 5000
+      shape: 3
+      shape: 32
+      shape: 32
+    }
+    include: kTrain
+  }
+#  layer{
+#    name: "data"
+#    type: kRecordInput
+#    store_conf {
+#      backend: "kvfile"
+#      path: "examples/cifar10/val_data.bin"
+#      mean_file: "examples/cifar10/image_mean.bin"
+#      batchsize: 64
+#      random_skip: 5000
+#      shape: 3
+#      shape: 32
+#      shape: 32
+#    }
+#    include: kVal
+#  }
+  layer{
+    name: "data"
+    type: kRecordInput
+    store_conf {
+      backend: "kvfile"
+      path: "examples/cifar10/test_data.bin"
+      mean_file: "examples/cifar10/image_mean.bin"
+      batchsize: 100
+      shape: 3
+      shape: 32
+      shape: 32
+    }
+    include: kTest
+  }
+
+  layer {
+    name: "conv1"
+    type: kCConvolution
+    srclayers: "data"
+    convolution_conf {
+      num_filters: 32
+      kernel: 5
+      stride: 1
+      pad:2
+    }
+    param {
+      name: "w1"
+      init {
+        type:kGaussian
+        std:0.0001
+      }
+    }
+    param {
+      name: "b1"
+      lr_scale:2.0
+      init {
+        type: kConstant
+        value:0
+      }
+    }
+  }
+
+  layer {
+    name: "pool1"
+    type: kCPooling
+    srclayers: "conv1"
+    pooling_conf {
+      pool: MAX
+      kernel: 3
+      stride: 2
+    }
+  }
+  layer {
+    name: "relu1"
+    type: kReLU
+    srclayers:"pool1"
+  }
+  layer {
+    name: "norm1"
+    type: kLRN
+    lrn_conf {
+      local_size: 3
+      alpha: 5e-05
+      beta: 0.75
+    }
+    srclayers:"relu1"
+  }
+  layer {
+    name: "conv2"
+    type: kCConvolution
+    srclayers: "norm1"
+    convolution_conf {
+      num_filters: 32
+      kernel: 5
+      stride: 1
+      pad:2
+    }
+    param {
+      name: "w2"
+      init {
+        type:kGaussian
+        std:0.01
+      }
+    }
+    param {
+      name: "b2"
+      lr_scale:2.0
+      init {
+        type: kConstant
+        value:0
+      }
+    }
+  }
+  layer {
+    name: "relu2"
+    type: kReLU
+    srclayers:"conv2"
+  }
+  layer {
+    name: "pool2"
+    type: kCPooling
+    srclayers: "relu2"
+    pooling_conf {
+      pool: AVG
+      kernel: 3
+      stride: 2
+    }
+  }
+  layer {
+    name: "norm2"
+    type: kLRN
+    lrn_conf {
+      local_size: 3
+      alpha: 5e-05
+      beta: 0.75
+    }
+    srclayers:"pool2"
+  }
+  layer {
+    name: "conv3"
+    type: kCConvolution
+    srclayers: "norm2"
+    convolution_conf {
+      num_filters: 64
+      kernel: 5
+      stride: 1
+      pad:2
+    }
+    param {
+      name: "w3"
+      init {
+        type:kGaussian
+        std:0.01
+      }
+    }
+    param {
+      name: "b3"
+      init {
+        type: kConstant
+        value:0
+      }
+    }
+  }
+  layer {
+    name: "relu3"
+    type: kReLU
+    srclayers:"conv3"
+  }
+  layer {
+    name: "pool3"
+    type: kCPooling
+    srclayers: "relu3"
+    pooling_conf {
+      pool: AVG
+      kernel: 3
+      stride: 2
+    }
+  }
+  layer {
+    name: "ip1"
+    type: kInnerProduct
+    srclayers:"pool3"
+    innerproduct_conf {
+      num_output: 10
+    }
+    param {
+      name: "w4"
+      wd_scale:250
+      init {
+        type:kGaussian
+        std:0.01
+      }
+    }
+    param {
+      name: "b4"
+      lr_scale:2.0
+      wd_scale:0
+      init {
+        type: kConstant
+        value:0
+      }
+    }
+  }
+#  layer {
+#   name : "softmax"
+#   type: kSoftmax
+#   srclayers: "ip1"
+#  }
+#
+#  layer {
+#   name : "argsort"
+#   type: kArgSort
+#   srclayers: "softmax"
+#  }
+  layer{
+    name: "loss"
+    type: kSoftmaxLoss
+    softmaxloss_conf{
+      topk:1
+    }
+    srclayers:"ip1"
+    srclayers: "data"
+  }
+# uncomment "softmax", "argsort", "output" layer and comment "loss" layer
+# to extract features from argsort
+#  layer {
+#    name : "output"
+#    type: kCSVOutput
+#    srclayers: "argsort"
+#    store_conf {
+#      path: "examples/cifar10/out.csv"
+#    }
+#  }
+}
+cluster {
+  nworker_groups: 1
+  nserver_groups: 1
+  nworkers_per_group: 1
+  nworkers_per_procs: 1
+  workspace: "examples/cifar10"
+}

examples/cifar10/hybrid.conf

@@ -4,7 +4,9 @@ test_steps: 0
 test_freq: 200
 #validate_steps: 100
 #validate_freq: 300
-disp_freq: 30
+disp_freq: 50
+
+
 #debug: true
 #checkpoint_path: "examples/cifar10/checkpoint/step1000-worker0"
 train_one_batch {

examples/cifar10/job.conf

@@ -5,6 +5,9 @@ test_freq: 200
 #validate_steps: 100
 #validate_freq: 300
 disp_freq: 50
+
+
+
 #checkpoint_path: "examples/cifar10/checkpoint/step1000-worker0"
 train_one_batch {
   alg: kBP