def fake_train(self):
experience = (ssbm.SimpleStateAction *
self.learner.config.experience_length)()
experience = ssbm.prepareStateActions(experience)
experience['initial'] = util.deepMap(np.zeros,
self.learner.core.hidden_size)
experiences = [experience] * self.batch_size
# For more advanced usage, user can control the tracing steps and
# dumping steps. User can also run online profiling during training.
#
# Create options to profile time/memory as well as parameters.
builder = tf.profiler.ProfileOptionBuilder
opts = builder(builder.time_and_memory()).order_by('micros').build()
opts2 = tf.profiler.ProfileOptionBuilder.trainable_variables_parameter(
)
# Collect traces of steps 10~20, dump the whole profile (with traces of
# step 10~20) at step 20. The dumped profile can be used for further profiling
# with command line interface or Web UI.
with tf.contrib.tfprof.ProfileContext('/tmp/train_dir',
trace_steps=range(10, 20),
dump_steps=[20]) as pctx:
# Run online profiling with 'op' view and 'opts' options at step 15, 18, 20.
pctx.add_auto_profiling('op', opts, [15, 18, 20])
# Run online profiling with 'scope' view and 'opts2' options at step 20.
pctx.add_auto_profiling('scope', opts2, [20])
# High level API, such as slim, Estimator, etc.
count = 0
while count != self.sweep_limit:
self.learner.train(experiences, self.batch_steps)
count += 1
def train(self):
shape = self.experiences['action'].shape
data_size = shape[0]
batches = []
for i in range(0, data_size, self.batch_size):
batches.append(util.deepMap(lambda t: t[i:i+self.batch_size], self.experiences))
valid_batches = batches[:self.valid_batches]
train_batches = batches[self.valid_batches:]
for epoch in range(self.epochs):
print("Epoch", epoch)
start_time = time.time()
for batch in train_batches:
self.rl.train(batch, log=False, zipped=True)
print(time.time() - start_time)
for batch in valid_batches:
self.rl.train(batch, train=False, zipped=True)
self.rl.save()
import resource
print('Memory usage: %s (kb)' % resource.getrusage(resource.RUSAGE_SELF).ru_maxrss)
def scan(f, inputs, initial_state, axis=0):
inputs = util.deepIter(util.deepMap(lambda t: iter(tf.unstack(t, axis=axis)), inputs))
outputs = []
output = initial_state
for input_ in inputs:
output = f(output, input_)
outputs.append(output)
return util.deepZipWith(lambda *ts: tf.stack(ts, axis=axis), *outputs)