[example] alexnet on imagenet

example/imagenet/README.md

@@ -0,0 +1,19 @@
+# Training Neural Networks on Imagenet
+
+## Prepare Dataset
+
+TODO
+
+## Neural Networks
+
+- [alexnet.py](alexnet.py) : alexnet with 5 convolution layers followed by 3
+  fully connnected layers
+
+## Results
+
+Machine: Dual Xeon E5-2680 2.8GHz, Dual GTX 980, Ubuntu 14.0, GCC 4.8, MKL, CUDA
+7, CUDNN v3
+
+| | val accuracy | 1 x GTX 980 | 2 x GTX 980 |
+| --- | ---: | ---: | ---: | ---: |
+| `alexnet.py` | ? | ? | 400 img/sec |

example/imagenet/alexnet.py

@@ -0,0 +1,61 @@
+# pylint: skip-file
+from data import ilsvrc12_iterator
+import mxnet as mx
+import logging
+
+## define alexnet
+input_data = mx.symbol.Variable(name="data")
+# stage 1
+conv1 = mx.symbol.Convolution(
+    data=input_data, kernel=(11, 11), stride=(4, 4), num_filter=96)
+relu1 = mx.symbol.Activation(data=conv1, act_type="relu")
+pool1 = mx.symbol.Pooling(
+    data=relu1, pool_type="max", kernel=(3, 3), stride=(2,2))
+lrn1 = mx.symbol.LRN(data=pool1, alpha=0.0001, beta=0.75, knorm=1, nsize=5)
+# stage 2
+conv2 = mx.symbol.Convolution(
+    data=lrn1, kernel=(5, 5), pad=(2, 2), num_filter=256)
+relu2 = mx.symbol.Activation(data=conv2, act_type="relu")
+pool2 = mx.symbol.Pooling(data=relu2, kernel=(3, 3), stride=(2, 2))
+lrn2 = mx.symbol.LRN(data=pool2, alpha=0.0001, beta=0.75, knorm=1, nsize=5)
+# stage 3
+conv3 = mx.symbol.Convolution(
+    data=lrn2, kernel=(3, 3), pad=(1, 1), num_filter=384)
+relu3 = mx.symbol.Activation(data=conv3, act_type="relu")
+conv4 = mx.symbol.Convolution(
+    data=relu3, kernel=(3, 3), pad=(1, 1), num_filter=384)
+relu4 = mx.symbol.Activation(data=conv4, act_type="relu")
+conv5 = mx.symbol.Convolution(
+    data=relu4, kernel=(3, 3), pad=(1, 1), num_filter=256)
+relu5 = mx.symbol.Activation(data=conv5, act_type="relu")
+pool3 = mx.symbol.Pooling(data=relu5, kernel=(3, 3), stride=(2, 2))
+# stage 4
+flatten = mx.symbol.Flatten(data=pool3)
+fc1 = mx.symbol.FullyConnected(data=flatten, num_hidden=4096)
+relu6 = mx.symbol.Activation(data=fc1, act_type="relu")
+dropout1 = mx.symbol.Dropout(data=relu6, p=0.5)
+# stage 5
+fc2 = mx.symbol.FullyConnected(data=dropout1, num_hidden=4096)
+relu7 = mx.symbol.Activation(data=fc2, act_type="relu")
+dropout2 = mx.symbol.Dropout(data=relu7, p=0.5)
+# stage 6
+fc3 = mx.symbol.FullyConnected(data=dropout2, num_hidden=1000)
+softmax = mx.symbol.Softmax(data=fc3)
+
+
+## data
+train, val = ilsvrc12_iterator(batch_size=256, input_shape=(3,224,224))
+
+## train
+num_gpus = 2
+gpus = [mx.gpu(i) for i in range(num_gpus)]
+model = mx.model.FeedForward(
+    ctx           = gpus,
+    symbol        = softmax,
+    num_round     = 20,
+    learning_rate = 0.01,
+    momentum      = 0.9,
+    wd            = 0.00001)
+logging.basicConfig(level = logging.DEBUG)
+model.fit(X = train, eval_data = val,
+          epoch_end_callback = mx.callback.Speedometer(100))

example/imagenet/data.py

@@ -0,0 +1,28 @@
+# pylint: skip-file
+""" data iterator for imagnet"""
+import sys
+sys.path.insert(0, "../../python/")
+import mxnet as mx
+
+def ilsvrc12_iterator(batch_size, input_shape):
+    """return train and val iterators for imagenet"""
+    train_dataiter = mx.io.ImageRecordIter(
+        path_imgrec        = "data/ilsvrc12/train.rec",
+        mean_img           = "data/ilsvrc12/mean.bin",
+        rand_crop          = True,
+        rand_mirror        = True,
+        prefetch_buffer    = 4,
+        preprocess_threads = 4,
+        data_shape         = input_shape,
+        batch_size         = batch_size)
+    val_dataiter = mx.io.ImageRecordIter(
+        path_imgrec        = "data/ilsvrc12/val.rec",
+        mean_img           = "data/ilsvrc12/mean.bin",
+        rand_crop          = False,
+        rand_mirror        = False,
+        prefetch_buffer    = 4,
+        preprocess_threads = 4,
+        data_shape         = input_shape,
+        batch_size         = batch_size)
+
+    return (train_dataiter, val_dataiter)

example/mnist/README.md

@@ -17,7 +17,7 @@ Cortes, and Christopher J.C. Burges.
 
 Using 100 minibatch size and 20 data passes (not fine tuned.)
 
-Machine: Dual Xeon E5-2680 2.8GHz, Dual GTX 980, Ubuntu 14.0, GCC 4.8. Intel MKL, and CUDA 7.0
+Machine: Dual Xeon E5-2680 2.8GHz, Dual GTX 980, Ubuntu 14.0, GCC 4.8.
 
 | | val accuracy | 2 x E5-2680 | 1 x GTX 980 | 2 x GTX 980 |
 | --- | ---: | ---: | ---: | ---: |

example/mnist/lenet.py

@@ -4,7 +4,6 @@
 import logging
 
 ## define lenet
-
 # input
 data = mx.symbol.Variable('data')
 # first conv
@@ -27,19 +26,14 @@
 lenet = mx.symbol.Softmax(data=fc2)
 
 ## data
-
 train, val = mnist_iterator(batch_size=100, input_shape=(1,28,28))
 
 ## train
-
 logging.basicConfig(level=logging.DEBUG)
-
 # dev = [mx.gpu(i) for i in range(2)]
 dev = mx.gpu()
-
 model = mx.model.FeedForward(
     ctx = dev, symbol = lenet, num_round = 20,
     learning_rate = 0.01, momentum = 0.9, wd = 0.00001)
-
 model.fit(X=train, eval_data=val,
           epoch_end_callback=mx.callback.Speedometer(100))