Running 3 convolutional layer neural network and applying stronger mu…

…ltiprocessing.

dash/deep_learning_multilayer.py

@@ -76,7 +76,11 @@ def train_model(dataDirName, overwrite=False):
     a = []
     x, y_, keep_prob, y_conv = convnet_variables(imWidth, imWidthReduc, N, nLabels)
 
-    with tf.Session() as sess: # config=tf.ConfigProto(inter_op_parallelism_threads=1, intra_op_parallelism_threads=1)) as sess:
+    config = tf.ConfigProto()
+    config.intra_op_parallelism_threads = 44
+    config.inter_op_parallelism_threads = 44
+
+    with tf.Session(config=config) as sess: # config=tf.ConfigProto(inter_op_parallelism_threads=1, intra_op_parallelism_threads=1)) as sess:
         # TRAIN AND EVALUATE MODEL
         cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv + 1e-8), reduction_indices=[1]))
         train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
@@ -144,7 +148,7 @@ def train_model(dataDirName, overwrite=False):
 
 if __name__ == '__main__':
     t1 = time.time()
-    savedFilenames = train_model('data_files/', overwrite=False)
+    savedFilenames = train_model('data_files_zeroZ/', overwrite=False)
     t2 = time.time()
     print("time spent: {0:.2f}".format(t2 - t1))
 

dash/multilayer_convnet.py

@@ -70,7 +70,7 @@ def convnet_variables(imWidth, imWidthReduc, N, ntypes):
     h_fc2 = Layer3.connect_layers(h_pool3, 64, 2)
 
     # READOUT LAYER
-    keep_prob, h_fc2_drop = Layer3.dropout(h_fc1)
+    keep_prob, h_fc2_drop = Layer3.dropout(h_fc2)
     W_fc3, b_fc3 = Layer3.readout_layer()
     y_conv = tf.nn.softmax(tf.matmul(h_fc2_drop, W_fc3) + b_fc3)