def test_train(self):
nnet, weightTerms = function_fit.inference(self.x, self.hidden1, self.hidden2)
modError, modLoss = function_fit.loss(nnet, weightTerms, self.regLambda, self.y)
train_op = function_fit.training(modLoss, self.learningRate)
init_op = tf.initialize_all_variables()
with tf.Session() as sess:
sess.run(init_op)
print("loss=%r" % sess.run(modLoss, feed_dict={self.x:np.ones((3,1), dtype=np.float32),
self.y:np.ones((3,1), dtype=np.float32)}))
def mainAfterWithGraph(functionData, mod):
# Generate placeholders for the input and labels
mod.x_input = tf.placeholder(tf.float32, shape=(None, 1))
mod.y_output = tf.placeholder(tf.float32, shape=(None, 1))
# Build a Graph that computes predictions from the inference model.
mod = ffit.inference(mod, mod.x_input, FLAGS.LayerNodes,
FLAGS.actfn, FLAGS.regNorm)
# Add to the Graph the Ops for loss calculation.
mod.modelError, mod.loss = ffit.loss(mod.nnetModel, mod.W_regterms, FLAGS.reg,
mod.y_output)
# Add to the Graph the Ops that calculate and apply gradients.
mod = ffit.training(mod, mod.loss, FLAGS.learning_rate)
# Build the summary operation based on the TF collection of Summaries.
# mod.train_summary_op = tf.merge_all_summaries()
# Create a saver for writing training checkpoints.
# saver = tf.train.Saver()
# Create a session for running Ops on the Graph.
sess = tf.Session(config=tf.ConfigProto(intra_op_parallelism_threads=4))
# Run the Op to initialize the variables.
init = tf.initialize_all_variables()
sess.run(init)
# Instantiate a SummaryWriter to output summaries and the Graph.
# summary_writer = tf.train.SummaryWriter(FLAGS.train_dir,
# graph_def=sess.graph_def)
h5out = tfh5.TensorFlowH5Writer(FLAGS.summaryfile)
FLAGS.h5write(h5out.h5)
functionData.h5write(h5out.h5)
train_feed_dict = {mod.x_input: functionData.x_train,
mod.y_output:functionData.y_train}
test_feed_dict = {mod.x_input: functionData.x_test,
mod.y_output:functionData.y_test}
all_feed_dict = {mod.x_input: functionData.x_all,
mod.y_output:functionData.y_all}
# And then after everything is built, start the training loop.
# saver.restore(sess, FLAGS.train_dir)
mod.grad = None
mod.gradMag = None
for step in xrange(FLAGS.max_steps):
trainStep(step, mod, sess, h5out, train_feed_dict, test_feed_dict, all_feed_dict)
