Merge branch 'master' of https://github.com/awni/cinc17

entry/next.sh

@@ -23,5 +23,5 @@ set -o pipefail
 
 RECORD=$1
 
-# TODO actually eval record and write to answers.txt
-echo $RECORD
+printf "$RECORD," >> answers.txt
+python eval.py . $RECORD >> answers.txt

entry/prepare-entry.sh

@@ -5,16 +5,41 @@
 set -e
 set -o pipefail
 
-cp ../LICENSE LICENSE.txt
+mkdir entry && cd entry
+
+## Copy source code files
+echo `pwd`
+cp ../../LICENSE LICENSE.txt
+cp ../setup.sh .
+cp ../next.sh .
+cp ../AUTHORS.txt .
+cp ../dependencies.txt .
+for f in 'eval' 'loader' 'network'
+do
+ cp ../../${f}.py .
+done
+
+## Copy model files
+model_dir=$1
+for f in 'config.json' 'loader.pkl' 'model.data-*' 'model.index'
+do
+ cp $model_dir/$f .
+done
 
 echo "==== running entry script on validation set ===="
 validation=/deep/group/med/alivecor/sample2017/validation
 
-rm -f answers.txt
 for r in `cat $validation/RECORDS`; do
  echo $r
  ln -sf $validation/$r.hea .
  ln -sf $validation/$r.mat .
  ./next.sh $r
  rm $r.hea $r.mat
 done
+
+## Make zip
+cd ..
+zip -r entry.zip entry
+
+## Cleanup
+# rm -r entry

network.py

@@ -1,7 +1,6 @@
 
 import numpy as np
 import tensorflow as tf
-import tensorflow.contrib.layers as tfl
 
 MOMENTUM_INIT = 0.5
 
@@ -18,9 +17,9 @@ def init_inference(self, config):
  num_filters = layer['num_filters']
  kernel_size = layer['kernel_size']
  stride = layer['stride']
- acts = tfl.convolution2d(acts, num_outputs=num_filters,
- kernel_size=[kernel_size, 1],
- stride=stride)
+ acts = tf.contrib.layers.convolution2d(acts, num_outputs=num_filters,
+  kernel_size=[kernel_size, 1],
+  stride=stride)
 
  # Activations should emerge from the convolution with shape
  # [batch_size, time (subsampled), 1, num_channels]
@@ -39,7 +38,7 @@ def init_inference(self, config):
  # Reduce the time-dimension to make a single prediction
  acts = tf.reduce_mean(acts, axis=1)
 
- self.logits = tfl.fully_connected(acts, self.output_dim)
+ self.logits = tf.contrib.layers.fully_connected(acts, self.output_dim)
  self.probs = tf.nn.softmax(self.logits)
 
  def init_loss(self):