| """ |
| Reference: |
| |
| Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012. |
| """ |
| import mxnet as mx |
| import numpy as np |
| |
| def get_symbol(num_classes, **kwargs): |
| input_data = mx.symbol.Variable(name="data") |
| input_data = mx.symbol.Cast(data=input_data, dtype=np.float16) |
| # stage 1 |
| weight = mx.symbol.Variable(name='conv1_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='conv1_bias', dtype=np.float16) |
| conv1 = mx.symbol.Convolution(name='conv1', |
| data=input_data, weight=weight, bias=bias, kernel=(11, 11), stride=(4, 4), num_filter=96) |
| relu1 = mx.symbol.Activation(data=conv1, act_type="relu") |
| lrn1 = mx.symbol.LRN(data=relu1, alpha=0.0001, beta=0.75, knorm=2, nsize=5) |
| pool1 = mx.symbol.Pooling( |
| data=lrn1, pool_type="max", kernel=(3, 3), stride=(2,2)) |
| # stage 2 |
| weight = mx.symbol.Variable(name='conv2_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='conv2_bias', dtype=np.float16) |
| conv2 = mx.symbol.Convolution(name='conv2', |
| data=pool1, weight=weight, bias=bias, kernel=(5, 5), pad=(2, 2), num_filter=256) |
| relu2 = mx.symbol.Activation(data=conv2, act_type="relu") |
| lrn2 = mx.symbol.LRN(data=relu2, alpha=0.0001, beta=0.75, knorm=2, nsize=5) |
| pool2 = mx.symbol.Pooling(data=lrn2, kernel=(3, 3), stride=(2, 2), pool_type="max") |
| # stage 3 |
| weight = mx.symbol.Variable(name='conv3_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='conv3_bias', dtype=np.float16) |
| conv3 = mx.symbol.Convolution(name='conv3', |
| data=pool2, weight=weight, bias=bias, kernel=(3, 3), pad=(1, 1), num_filter=384) |
| relu3 = mx.symbol.Activation(data=conv3, act_type="relu") |
| weight = mx.symbol.Variable(name='conv4_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='conv4_bias', dtype=np.float16) |
| conv4 = mx.symbol.Convolution(name='conv4', |
| data=relu3, weight=weight, bias=bias, kernel=(3, 3), pad=(1, 1), num_filter=384) |
| relu4 = mx.symbol.Activation(data=conv4, act_type="relu") |
| weight = mx.symbol.Variable(name='conv5_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='conv5_bias', dtype=np.float16) |
| conv5 = mx.symbol.Convolution(name='conv5', |
| data=relu4, weight=weight, bias=bias, 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), pool_type="max") |
| # stage 4 |
| flatten = mx.symbol.Flatten(data=pool3) |
| weight = mx.symbol.Variable(name='fc1_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='fc1_bias', dtype=np.float16) |
| fc1 = mx.symbol.FullyConnected(name='fc1', data=flatten, weight=weight, bias=bias, |
| num_hidden=4096) |
| relu6 = mx.symbol.Activation(data=fc1, act_type="relu") |
| dropout1 = mx.symbol.Dropout(data=relu6, p=0.5) |
| # stage 5 |
| weight = mx.symbol.Variable(name='fc2_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='fc2_bias', dtype=np.float16) |
| fc2 = mx.symbol.FullyConnected(name='fc2', data=dropout1, weight=weight, bias=bias, |
| num_hidden=4096) |
| relu7 = mx.symbol.Activation(data=fc2, act_type="relu") |
| dropout2 = mx.symbol.Dropout(data=relu7, p=0.5) |
| # stage 6 |
| weight = mx.symbol.Variable(name='fc3_weight', dtype=np.float16) |
| bias = mx.symbol.Variable(name='fc3_bias', dtype=np.float16) |
| fc3 = mx.symbol.FullyConnected(name='fc3', data=dropout2, weight=weight, bias=bias, |
| num_hidden=num_classes) |
| label = mx.symbol.Variable(name='softmax_label') |
| label = mx.symbol.Cast(data=label, dtype=np.float16) |
| softmax = mx.symbol.SoftmaxOutput(data=fc3, name='softmax', label=label) |
| return softmax |
