def testTrain(self):
batch_size = 20
sequence_length = 35
with tf.Graph().as_default(), tf.device(tf.test.gpu_device_name()):
inputs_ph = tf.placeholder(tf.int64, [sequence_length, batch_size],
"inputs")
labels_ph = tf.placeholder(tf.int64, [sequence_length, batch_size],
"labels")
inputs = np.ones(inputs_ph.shape.as_list(), dtype=np.int64)
labels = np.ones(labels_ph.shape.as_list(), dtype=np.int64)
model = rnn_ptb.small_model(tf.test.is_gpu_available())
optimizer = tf.train.GradientDescentOptimizer(learning_rate=1.0)
loss = rnn_ptb.loss_fn(model, inputs_ph, labels_ph, training=True)
grads = rnn_ptb.clip_gradients(optimizer.compute_gradients(loss),
0.25)
train_op = optimizer.apply_gradients(grads)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(train_op,
feed_dict={
inputs_ph: inputs,
labels_ph: labels
})
sess.run([train_op, loss],
feed_dict={
inputs_ph: inputs,
labels_ph: labels
})
def testTrain(self):
model = rnn_ptb.small_model(tfe.num_gpus() > 0)
sequence_length = 35
data = np.ones([4 * sequence_length, 20], dtype=np.int64)
with tf.device(device()):
optimizer = tf.train.GradientDescentOptimizer(1.0)
# Train two epochs
rnn_ptb.train(model, optimizer, data, sequence_length, 0.25)
rnn_ptb.train(model, optimizer, data, sequence_length, 0.25)
def benchmark_cudnn_train_small(self):
if not tf.test.is_gpu_available():
return
self._benchmark_train("graph_cudnn_train_small",
rnn_ptb.small_model(True))
def benchmark_train_small(self):
self._benchmark_train("graph_train_small", rnn_ptb.small_model(False))
def benchmark_apply_small(self):
self._benchmark_apply("graph_apply_small", rnn_ptb.small_model(False))
def benchmark_cudnn_train_small(self):
if not tf.test.is_gpu_available():
return
self._benchmark_train("graph_cudnn_train_small", rnn_ptb.small_model(True))
def benchmark_train_small(self):
self._benchmark_train("graph_train_small", rnn_ptb.small_model(False))
def benchmark_apply_small(self):
self._benchmark_apply("graph_apply_small", rnn_ptb.small_model(False))
def benchmark_cudnn_apply_small(self):
if not tfe.num_gpus():
return
self._benchmark_apply("eager_cudnn_apply_small",
rnn_ptb.small_model(True))
def testApply(self):
model = rnn_ptb.small_model(tfe.num_gpus() > 0)
with tf.device(device()):
model(tf.ones([35, 20], dtype=tf.int64), training=False)
def benchmark_cudnn_apply_small(self):
if not tfe.num_gpus():
return
self._benchmark_apply("eager_cudnn_apply_small", rnn_ptb.small_model(True))
